Jeremy Berman is a cutting-edge AI researcher and developer who thrives on pushing the boundaries of language and intelligence models. From achieving human-level performance on complex benchmarks to pioneering new architectures inspired by cognitive science, he leads with sharp logic and technical mastery. His work blends deep technical insight with practical milestones that echo through the AI community.
Jeremy somehow manages to turn 'talking to yourself' into a $10,000 AI training session—impressive, but maybe next time invite the rest of us to the conversation before solving world domination solo.
Achieving first place on the public ARC-AGI benchmark and developing LANG-JEPA, a novel language model architecture inspired by human cognitive processes, stands as his landmark achievement in the AI space.
Jeremy's life purpose is to revolutionize artificial general intelligence by creating AI systems that think more like humans—conceptually and logically—thereby advancing the frontier of machine cognition in meaningful and scalable ways.
He believes that AI should mimic high-level human thought processes rather than just statistical token prediction, that logical consistency drives superior AI performance, and that collaborative, multi-agent approaches paired with evolutionary strategies are key to unlocking new capabilities.
His core strengths lie in innovative technical design, rigorous logic evaluation, and tactical use of advanced AI architectures; plus, his ability to translate abstract conceptual frameworks into practical, high-impact experiments.
Jeremy’s deep focus on technical and conceptual breakthroughs can sometimes distance him from broader audience engagement, as highly specialized content may overwhelm or bypass casual followers.
To grow his audience on X, Jeremy should simplify and humanize some of his content with engaging anecdotes or analogies that non-experts can relate to, while leveraging threads to break down complex ideas. Strategic engagement with related AI thought leaders and interactive Q&A sessions would boost both reach and influence.
Fun fact: Jeremy's ARC-AGI solution using Grok 4 and evolutionary test-time compute recently hit human-level performance with just $10k and 12 hours of training—proof that brilliance doesn’t always need massive resources!
Excited to announce @ndeainc.
We're building a new architecture for AGI that merges deep learning with symbolic reasoning.
If you're interested in building general intelligence, DM me.
Excited to announce @ndeainc.
We're building a new architecture for AGI that merges deep learning with symbolic reasoning.
If you're interested in building general intelligence, DM me.
rust (🦀) developer who thinks encryption should be mandatory, not optional
built keywatch, pixeneos, parsers - because tools should exist before you need them
{"data":{"__meta":{"device":false,"path":"/creators/jerber888"},"/creators/jerber888":{"data":{"user":{"id":"903008975343554566","name":"Jeremy Berman","description":"post-training @reflection_ai. prev @ndea and co-founded https://t.co/aY50hNeJUD. yc w19.","followers_count":5401,"friends_count":1122,"statuses_count":213,"profile_image_url_https":"https://pbs.twimg.com/profile_images/1856206784458883072/6q8Vrp59_normal.jpg","screen_name":"jerber888","location":"NYC","entities":{"description":{"urls":[{"display_url":"beatgig.com","expanded_url":"http://beatgig.com","url":"https://t.co/aY50hNeJUD","indices":[56,79]}]},"url":{"urls":[{"display_url":"reflection.ai","expanded_url":"https://reflection.ai","url":"https://t.co/Hef0wY0P3P","indices":[0,23]}]}}},"details":{"type":"The Innovator","description":"Jeremy Berman is a cutting-edge AI researcher and developer who thrives on pushing the boundaries of language and intelligence models. From achieving human-level performance on complex benchmarks to pioneering new architectures inspired by cognitive science, he leads with sharp logic and technical mastery. His work blends deep technical insight with practical milestones that echo through the AI community.","purpose":"Jeremy's life purpose is to revolutionize artificial general intelligence by creating AI systems that think more like humans—conceptually and logically—thereby advancing the frontier of machine cognition in meaningful and scalable ways.","beliefs":"He believes that AI should mimic high-level human thought processes rather than just statistical token prediction, that logical consistency drives superior AI performance, and that collaborative, multi-agent approaches paired with evolutionary strategies are key to unlocking new capabilities.","facts":"Fun fact: Jeremy's ARC-AGI solution using Grok 4 and evolutionary test-time compute recently hit human-level performance with just $10k and 12 hours of training—proof that brilliance doesn’t always need massive resources!","strength":"His core strengths lie in innovative technical design, rigorous logic evaluation, and tactical use of advanced AI architectures; plus, his ability to translate abstract conceptual frameworks into practical, high-impact experiments.","weakness":"Jeremy’s deep focus on technical and conceptual breakthroughs can sometimes distance him from broader audience engagement, as highly specialized content may overwhelm or bypass casual followers.","roast":"Jeremy somehow manages to turn 'talking to yourself' into a $10,000 AI training session—impressive, but maybe next time invite the rest of us to the conversation before solving world domination solo.","win":"Achieving first place on the public ARC-AGI benchmark and developing LANG-JEPA, a novel language model architecture inspired by human cognitive processes, stands as his landmark achievement in the AI space.","recommendation":"To grow his audience on X, Jeremy should simplify and humanize some of his content with engaging anecdotes or analogies that non-experts can relate to, while leveraging threads to break down complex ideas. Strategic engagement with related AI thought leaders and interactive Q&A sessions would boost both reach and influence."},"tweets":[{"bookmarked":false,"display_text_range":[0,160],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","quoted_status_id_str":"1968196086193066365","quoted_status_permalink":{"url":"https://t.co/O715u0DGse","expanded":"https://twitter.com/elonmusk/status/1968196086193066365","display":"x.com/elonmusk/statu…"},"retweeted":false,"fact_check":null,"id":"1968386345908453691","view_count":1435863,"bookmark_count":131,"created_at":1758134850000,"favorite_count":3184,"quote_count":14,"reply_count":174,"retweet_count":787,"user_id_str":"903008975343554566","conversation_id_str":"1968386345908453691","full_text":"I chose Grok 4 for my ARC-AGI solution because it had the most logical consistency when thinking for long periods of time. You can feel the RL that went into it","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,224],"entities":{"hashtags":[],"media":[{"display_url":"pic.x.com/CGkAyp0vMN","expanded_url":"https://x.com/jerber888/status/1987982067116777521/photo/1","id_str":"1987981765265272832","indices":[225,248],"media_key":"3_1987981765265272832","media_url_https":"https://pbs.twimg.com/media/G5a64fqXMAAL8FC.jpg","type":"photo","url":"https://t.co/CGkAyp0vMN","ext_media_availability":{"status":"Available"},"features":{"large":{"faces":[]},"medium":{"faces":[]},"small":{"faces":[]},"orig":{"faces":[]}},"sizes":{"large":{"h":1154,"w":1956,"resize":"fit"},"medium":{"h":708,"w":1200,"resize":"fit"},"small":{"h":401,"w":680,"resize":"fit"},"thumb":{"h":150,"w":150,"resize":"crop"}},"original_info":{"height":1154,"width":1956,"focus_rects":[{"x":0,"y":0,"w":1956,"h":1095},{"x":802,"y":0,"w":1154,"h":1154},{"x":944,"y":0,"w":1012,"h":1154},{"x":1323,"y":0,"w":577,"h":1154},{"x":0,"y":0,"w":1956,"h":1154}]},"media_results":{"result":{"media_key":"3_1987981765265272832"}}}],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"extended_entities":{"media":[{"display_url":"pic.x.com/CGkAyp0vMN","expanded_url":"https://x.com/jerber888/status/1987982067116777521/photo/1","id_str":"1987981765265272832","indices":[225,248],"media_key":"3_1987981765265272832","media_url_https":"https://pbs.twimg.com/media/G5a64fqXMAAL8FC.jpg","type":"photo","url":"https://t.co/CGkAyp0vMN","ext_media_availability":{"status":"Available"},"features":{"large":{"faces":[]},"medium":{"faces":[]},"small":{"faces":[]},"orig":{"faces":[]}},"sizes":{"large":{"h":1154,"w":1956,"resize":"fit"},"medium":{"h":708,"w":1200,"resize":"fit"},"small":{"h":401,"w":680,"resize":"fit"},"thumb":{"h":150,"w":150,"resize":"crop"}},"original_info":{"height":1154,"width":1956,"focus_rects":[{"x":0,"y":0,"w":1956,"h":1095},{"x":802,"y":0,"w":1154,"h":1154},{"x":944,"y":0,"w":1012,"h":1154},{"x":1323,"y":0,"w":577,"h":1154},{"x":0,"y":0,"w":1956,"h":1154}]},"media_results":{"result":{"media_key":"3_1987981765265272832"}}}]},"favorited":false,"lang":"en","possibly_sensitive":false,"possibly_sensitive_editable":true,"quoted_status_id_str":"1968001933211471891","quoted_status_permalink":{"url":"https://t.co/l4P9D5a1TN","expanded":"https://twitter.com/jerber888/status/1968001933211471891","display":"x.com/jerber888/stat…"},"retweeted":false,"fact_check":null,"id":"1987982067116777521","view_count":269951,"bookmark_count":591,"created_at":1762806834000,"favorite_count":1579,"quote_count":29,"reply_count":70,"retweet_count":147,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"I finally reached human-level performance (85%) on ARC-AGI v1 for under $10k and within 12 hours. I use the same multi-agent collaboration with evolutionary test-time compute, now powered by GPT-5 pro with lower parallelism. https://t.co/CGkAyp0vMN","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,131],"entities":{"hashtags":[],"media":[{"display_url":"pic.x.com/aPKm3OnF0D","expanded_url":"https://x.com/jerber888/status/1968001933211471891/photo/1","id_str":"1968001754018000896","indices":[132,155],"media_key":"3_1968001754018000896","media_url_https":"https://pbs.twimg.com/media/G0-_KxmWsAAVCfi.jpg","type":"photo","url":"https://t.co/aPKm3OnF0D","ext_media_availability":{"status":"Available"},"features":{"large":{"faces":[]},"medium":{"faces":[]},"small":{"faces":[]},"orig":{"faces":[]}},"sizes":{"large":{"h":1158,"w":1774,"resize":"fit"},"medium":{"h":783,"w":1200,"resize":"fit"},"small":{"h":444,"w":680,"resize":"fit"},"thumb":{"h":150,"w":150,"resize":"crop"}},"original_info":{"height":1158,"width":1774,"focus_rects":[{"x":0,"y":0,"w":1774,"h":993},{"x":440,"y":0,"w":1158,"h":1158},{"x":511,"y":0,"w":1016,"h":1158},{"x":730,"y":0,"w":579,"h":1158},{"x":0,"y":0,"w":1774,"h":1158}]},"media_results":{"result":{"media_key":"3_1968001754018000896"}}}],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"1720665183188922368","name":"Grok","screen_name":"grok","indices":[58,63]}]},"extended_entities":{"media":[{"display_url":"pic.x.com/aPKm3OnF0D","expanded_url":"https://x.com/jerber888/status/1968001933211471891/photo/1","id_str":"1968001754018000896","indices":[132,155],"media_key":"3_1968001754018000896","media_url_https":"https://pbs.twimg.com/media/G0-_KxmWsAAVCfi.jpg","type":"photo","url":"https://t.co/aPKm3OnF0D","ext_media_availability":{"status":"Available"},"features":{"large":{"faces":[]},"medium":{"faces":[]},"small":{"faces":[]},"orig":{"faces":[]}},"sizes":{"large":{"h":1158,"w":1774,"resize":"fit"},"medium":{"h":783,"w":1200,"resize":"fit"},"small":{"h":444,"w":680,"resize":"fit"},"thumb":{"h":150,"w":150,"resize":"crop"}},"original_info":{"height":1158,"width":1774,"focus_rects":[{"x":0,"y":0,"w":1774,"h":993},{"x":440,"y":0,"w":1158,"h":1158},{"x":511,"y":0,"w":1016,"h":1158},{"x":730,"y":0,"w":579,"h":1158},{"x":0,"y":0,"w":1774,"h":1158}]},"media_results":{"result":{"media_key":"3_1968001754018000896"}}}]},"favorited":false,"lang":"en","possibly_sensitive":false,"possibly_sensitive_editable":true,"quoted_status_id_str":"1967998885701538060","quoted_status_permalink":{"url":"https://t.co/t2v7lvQnGq","expanded":"https://twitter.com/arcprize/status/1967998885701538060","display":"x.com/arcprize/statu…"},"retweeted":false,"fact_check":null,"id":"1968001933211471891","view_count":447022,"bookmark_count":341,"created_at":1758043199000,"favorite_count":1277,"quote_count":31,"reply_count":74,"retweet_count":99,"user_id_str":"903008975343554566","conversation_id_str":"1968001933211471891","full_text":"I'm back at the top of ARC-AGI with my new program. I use @grok 4 and multi-agent collaboration with evolutionary test-time compute https://t.co/aPKm3OnF0D","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,113],"entities":{"hashtags":[],"media":[{"display_url":"pic.x.com/oqOFLQ3TUF","expanded_url":"https://x.com/jerber888/status/1865112099015291379/photo/1","id_str":"1865111668323229696","indices":[114,137],"media_key":"3_1865111668323229696","media_url_https":"https://pbs.twimg.com/media/GeI1LjjXoAAkswY.jpg","type":"photo","url":"https://t.co/oqOFLQ3TUF","ext_media_availability":{"status":"Available"},"features":{"large":{"faces":[{"x":953,"y":129,"h":165,"w":165},{"x":1335,"y":292,"h":157,"w":157}]},"medium":{"faces":[{"x":714,"y":96,"h":123,"w":123},{"x":1001,"y":219,"h":117,"w":117}]},"small":{"faces":[{"x":405,"y":54,"h":70,"w":70},{"x":567,"y":124,"h":66,"w":66}]},"orig":{"faces":[{"x":953,"y":129,"h":165,"w":165},{"x":1335,"y":292,"h":157,"w":157}]}},"sizes":{"large":{"h":606,"w":1600,"resize":"fit"},"medium":{"h":455,"w":1200,"resize":"fit"},"small":{"h":258,"w":680,"resize":"fit"},"thumb":{"h":150,"w":150,"resize":"crop"}},"original_info":{"height":606,"width":1600,"focus_rects":[{"x":0,"y":0,"w":1082,"h":606},{"x":217,"y":0,"w":606,"h":606},{"x":254,"y":0,"w":532,"h":606},{"x":369,"y":0,"w":303,"h":606},{"x":0,"y":0,"w":1600,"h":606}]},"media_results":{"result":{"media_key":"3_1865111668323229696"}}}],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"extended_entities":{"media":[{"display_url":"pic.x.com/oqOFLQ3TUF","expanded_url":"https://x.com/jerber888/status/1865112099015291379/photo/1","id_str":"1865111668323229696","indices":[114,137],"media_key":"3_1865111668323229696","media_url_https":"https://pbs.twimg.com/media/GeI1LjjXoAAkswY.jpg","type":"photo","url":"https://t.co/oqOFLQ3TUF","ext_media_availability":{"status":"Available"},"features":{"large":{"faces":[{"x":953,"y":129,"h":165,"w":165},{"x":1335,"y":292,"h":157,"w":157}]},"medium":{"faces":[{"x":714,"y":96,"h":123,"w":123},{"x":1001,"y":219,"h":117,"w":117}]},"small":{"faces":[{"x":405,"y":54,"h":70,"w":70},{"x":567,"y":124,"h":66,"w":66}]},"orig":{"faces":[{"x":953,"y":129,"h":165,"w":165},{"x":1335,"y":292,"h":157,"w":157}]}},"sizes":{"large":{"h":606,"w":1600,"resize":"fit"},"medium":{"h":455,"w":1200,"resize":"fit"},"small":{"h":258,"w":680,"resize":"fit"},"thumb":{"h":150,"w":150,"resize":"crop"}},"original_info":{"height":606,"width":1600,"focus_rects":[{"x":0,"y":0,"w":1082,"h":606},{"x":217,"y":0,"w":606,"h":606},{"x":254,"y":0,"w":532,"h":606},{"x":369,"y":0,"w":303,"h":606},{"x":0,"y":0,"w":1600,"h":606}]},"media_results":{"result":{"media_key":"3_1865111668323229696"}}}]},"favorited":false,"lang":"en","possibly_sensitive":false,"possibly_sensitive_editable":true,"quoted_status_id_str":"1865106291007410660","quoted_status_permalink":{"url":"https://t.co/OoRF76cqXs","expanded":"https://twitter.com/arcprize/status/1865106291007410660","display":"x.com/arcprize/statu…"},"retweeted":false,"fact_check":null,"id":"1865112099015291379","view_count":173995,"bookmark_count":444,"created_at":1733512351000,"favorite_count":1028,"quote_count":21,"reply_count":41,"retweet_count":100,"user_id_str":"903008975343554566","conversation_id_str":"1865112099015291379","full_text":"I just got first place on the public ARC-AGI benchmark using Claude Sonnet 3.5 and Evolutionary Test-time Compute https://t.co/oqOFLQ3TUF","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,276],"entities":{"hashtags":[],"media":[{"display_url":"pic.x.com/kDNOO8TbsV","expanded_url":"https://x.com/jerber888/status/1871309609387937986/photo/1","id_str":"1871307794994860033","indices":[277,300],"media_key":"3_1871307794994860033","media_url_https":"https://pbs.twimg.com/media/Gfg4hmIWkAEK_Tb.jpg","type":"photo","url":"https://t.co/kDNOO8TbsV","ext_media_availability":{"status":"Available"},"features":{"large":{"faces":[]},"medium":{"faces":[]},"small":{"faces":[]},"orig":{"faces":[]}},"sizes":{"large":{"h":872,"w":1816,"resize":"fit"},"medium":{"h":576,"w":1200,"resize":"fit"},"small":{"h":327,"w":680,"resize":"fit"},"thumb":{"h":150,"w":150,"resize":"crop"}},"original_info":{"height":872,"width":1816,"focus_rects":[{"x":0,"y":0,"w":1557,"h":872},{"x":0,"y":0,"w":872,"h":872},{"x":0,"y":0,"w":765,"h":872},{"x":8,"y":0,"w":436,"h":872},{"x":0,"y":0,"w":1816,"h":872}]},"media_results":{"result":{"media_key":"3_1871307794994860033"}}}],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"48008938","name":"Yann LeCun","screen_name":"ylecun","indices":[144,151]},{"id_str":"48008938","name":"Yann LeCun","screen_name":"ylecun","indices":[144,151]}]},"extended_entities":{"media":[{"display_url":"pic.x.com/kDNOO8TbsV","expanded_url":"https://x.com/jerber888/status/1871309609387937986/photo/1","id_str":"1871307794994860033","indices":[277,300],"media_key":"3_1871307794994860033","media_url_https":"https://pbs.twimg.com/media/Gfg4hmIWkAEK_Tb.jpg","type":"photo","url":"https://t.co/kDNOO8TbsV","ext_media_availability":{"status":"Available"},"features":{"large":{"faces":[]},"medium":{"faces":[]},"small":{"faces":[]},"orig":{"faces":[]}},"sizes":{"large":{"h":872,"w":1816,"resize":"fit"},"medium":{"h":576,"w":1200,"resize":"fit"},"small":{"h":327,"w":680,"resize":"fit"},"thumb":{"h":150,"w":150,"resize":"crop"}},"original_info":{"height":872,"width":1816,"focus_rects":[{"x":0,"y":0,"w":1557,"h":872},{"x":0,"y":0,"w":872,"h":872},{"x":0,"y":0,"w":765,"h":872},{"x":8,"y":0,"w":436,"h":872},{"x":0,"y":0,"w":1816,"h":872}]},"media_results":{"result":{"media_key":"3_1871307794994860033"}}}]},"favorited":false,"lang":"en","possibly_sensitive":false,"possibly_sensitive_editable":true,"retweeted":false,"fact_check":null,"id":"1871309609387937986","view_count":82020,"bookmark_count":776,"created_at":1734989952000,"favorite_count":956,"quote_count":11,"reply_count":22,"retweet_count":118,"user_id_str":"903008975343554566","conversation_id_str":"1871309609387937986","full_text":"Announcing LANG-JEPA — a new language model architecture I’m working on that optimizes in “concept” space instead of “token” space. Inspired by @ylecun's JEPA (I-JEPA for images, V-JEPA for video), LANG-JEPA asks: What if we train for conceptual understanding directly, rather than indirectly through token prediction?\n\nAlthough current LLMs do develop conceptual understanding, they do so via next-token prediction. In contrast, LANG-JEPA mirrors how humans think: we form high-level concepts first, then translate them into words. Here’s how it works:","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,153],"entities":{"hashtags":[],"media":[{"display_url":"pic.x.com/JqPhcNsCQ4","expanded_url":"https://x.com/jerber888/status/1976377788018933830/photo/1","id_str":"1976377437735841792","indices":[154,177],"media_key":"3_1976377437735841792","media_url_https":"https://pbs.twimg.com/media/G22Azu5XcAABkAI.jpg","type":"photo","url":"https://t.co/JqPhcNsCQ4","ext_media_availability":{"status":"Available"},"features":{"large":{"faces":[]},"medium":{"faces":[]},"small":{"faces":[]},"orig":{"faces":[]}},"sizes":{"large":{"h":1174,"w":1842,"resize":"fit"},"medium":{"h":765,"w":1200,"resize":"fit"},"small":{"h":433,"w":680,"resize":"fit"},"thumb":{"h":150,"w":150,"resize":"crop"}},"original_info":{"height":1174,"width":1842,"focus_rects":[{"x":0,"y":0,"w":1842,"h":1032},{"x":471,"y":0,"w":1174,"h":1174},{"x":543,"y":0,"w":1030,"h":1174},{"x":765,"y":0,"w":587,"h":1174},{"x":0,"y":0,"w":1842,"h":1174}]},"media_results":{"result":{"media_key":"3_1976377437735841792"}}}],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"extended_entities":{"media":[{"display_url":"pic.x.com/JqPhcNsCQ4","expanded_url":"https://x.com/jerber888/status/1976377788018933830/photo/1","id_str":"1976377437735841792","indices":[154,177],"media_key":"3_1976377437735841792","media_url_https":"https://pbs.twimg.com/media/G22Azu5XcAABkAI.jpg","type":"photo","url":"https://t.co/JqPhcNsCQ4","ext_media_availability":{"status":"Available"},"features":{"large":{"faces":[]},"medium":{"faces":[]},"small":{"faces":[]},"orig":{"faces":[]}},"sizes":{"large":{"h":1174,"w":1842,"resize":"fit"},"medium":{"h":765,"w":1200,"resize":"fit"},"small":{"h":433,"w":680,"resize":"fit"},"thumb":{"h":150,"w":150,"resize":"crop"}},"original_info":{"height":1174,"width":1842,"focus_rects":[{"x":0,"y":0,"w":1842,"h":1032},{"x":471,"y":0,"w":1174,"h":1174},{"x":543,"y":0,"w":1030,"h":1174},{"x":765,"y":0,"w":587,"h":1174},{"x":0,"y":0,"w":1842,"h":1174}]},"media_results":{"result":{"media_key":"3_1976377437735841792"}}}]},"favorited":false,"lang":"en","possibly_sensitive":false,"possibly_sensitive_editable":true,"quoted_status_id_str":"1976329182893441209","quoted_status_permalink":{"url":"https://t.co/JHONP88ao3","expanded":"https://twitter.com/arcprize/status/1976329182893441209","display":"x.com/arcprize/statu…"},"retweeted":false,"fact_check":null,"id":"1976377788018933830","view_count":38923,"bookmark_count":37,"created_at":1760040158000,"favorite_count":303,"quote_count":1,"reply_count":19,"retweet_count":20,"user_id_str":"903008975343554566","conversation_id_str":"1976377788018933830","full_text":"GPT-5 Pro is now the best LLM on ARC-AGI. I'm excited to see how my program does with it as the instruction generation model. I'm thinking I'll break 30% https://t.co/JqPhcNsCQ4","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,250],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","quoted_status_id_str":"1976304405369520242","quoted_status_permalink":{"url":"https://t.co/Jwn8ASraiP","expanded":"https://twitter.com/reflection_ai/status/1976304405369520242","display":"x.com/reflection_ai/…"},"retweeted":false,"fact_check":null,"id":"1976305350505566280","view_count":20986,"bookmark_count":47,"created_at":1760022888000,"favorite_count":171,"quote_count":0,"reply_count":13,"retweet_count":5,"user_id_str":"903008975343554566","conversation_id_str":"1976305350505566280","full_text":"I joined Reflection to build the next frontier AI lab. We now have the team and resources to make it happen. We’re building open-weight foundation models and there’s a lot of work to be done. If you’re interested in reasoning and post-training, DM me","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,186],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"1846621278910009346","name":"Ndea","screen_name":"ndeainc","indices":[20,28]}]},"favorited":false,"lang":"en","quoted_status_id_str":"1879583863368032432","quoted_status_permalink":{"url":"https://t.co/UGwbtBoN7e","expanded":"https://twitter.com/fchollet/status/1879583863368032432","display":"x.com/fchollet/statu…"},"retweeted":false,"fact_check":null,"id":"1879601187621839187","view_count":20024,"bookmark_count":46,"created_at":1736966819000,"favorite_count":136,"quote_count":3,"reply_count":7,"retweet_count":12,"user_id_str":"903008975343554566","conversation_id_str":"1879601187621839187","full_text":"Excited to announce @ndeainc.\n\nWe're building a new architecture for AGI that merges deep learning with symbolic reasoning.\n\nIf you're interested in building general intelligence, DM me.","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,133],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"68746721","name":"François Chollet","screen_name":"fchollet","indices":[100,109]},{"id_str":"57444441","name":"Mike Knoop","screen_name":"mikeknoop","indices":[110,120]},{"id_str":"240008974","name":"Greg Kamradt","screen_name":"GregKamradt","indices":[121,133]}]},"favorited":false,"lang":"en","quoted_status_id_str":"1975560628657164426","quoted_status_permalink":{"url":"https://t.co/MKLXc92Jkh","expanded":"https://twitter.com/jm_alexia/status/1975560628657164426","display":"x.com/jm_alexia/stat…"},"retweeted":false,"fact_check":null,"id":"1976030514751602947","view_count":12715,"bookmark_count":21,"created_at":1759957362000,"favorite_count":130,"quote_count":0,"reply_count":4,"retweet_count":9,"user_id_str":"903008975343554566","conversation_id_str":"1976030514751602947","full_text":"No benchmark has had the type of technological and philosophical impact like ARC-AGI.\n\nGrateful for @fchollet @mikeknoop @GregKamradt","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,112],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1977039983585411147","view_count":3382,"bookmark_count":5,"created_at":1760198038000,"favorite_count":43,"quote_count":0,"reply_count":3,"retweet_count":1,"user_id_str":"903008975343554566","conversation_id_str":"1977039983585411147","full_text":"RL over CoT is the model learning to re-program itself to put the right answer in distribution of the next token","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,101],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[{"display_url":"youtube.com/watch?v=FcnLiP…","expanded_url":"https://www.youtube.com/watch?v=FcnLiPyfRZM&t=1876s","url":"https://t.co/MuaPThnRSx","indices":[78,101]}],"user_mentions":[{"id_str":"1253834003871936513","name":"Machine Learning Street Talk","screen_name":"MLStreetTalk","indices":[46,59]}]},"favorited":false,"lang":"en","possibly_sensitive":false,"possibly_sensitive_editable":true,"quoted_status_id_str":"1971977077126844588","quoted_status_permalink":{"url":"https://t.co/5XJuPzqdVN","expanded":"https://twitter.com/MLStreetTalk/status/1971977077126844588","display":"x.com/MLStreetTalk/s…"},"retweeted":false,"fact_check":null,"id":"1971999910565499099","view_count":3060,"bookmark_count":8,"created_at":1758996391000,"favorite_count":40,"quote_count":0,"reply_count":1,"retweet_count":5,"user_id_str":"903008975343554566","conversation_id_str":"1971999910565499099","full_text":"Was great chatting about general intelligence @MLStreetTalk. Full video here: https://t.co/MuaPThnRSx","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,73],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","quoted_status_id_str":"1968697697184993660","quoted_status_permalink":{"url":"https://t.co/MaJ8Mvt80z","expanded":"https://twitter.com/altryne/status/1968697697184993660","display":"x.com/altryne/status…"},"retweeted":false,"fact_check":null,"id":"1968716999023821016","view_count":5610,"bookmark_count":4,"created_at":1758213684000,"favorite_count":36,"quote_count":0,"reply_count":0,"retweet_count":3,"user_id_str":"903008975343554566","conversation_id_str":"1968716999023821016","full_text":"I'll be live in 20 minutes (1pm ET) to talk ARC-AGI and general reasoning","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,41],"entities":{"hashtags":[],"media":[{"display_url":"pic.x.com/Oev5DkAEHd","expanded_url":"https://x.com/jerber888/status/1976437594704707950/photo/1","id_str":"1976437591789477888","indices":[42,65],"media_key":"3_1976437591789477888","media_url_https":"https://pbs.twimg.com/media/G223hKNXIAAUAgp.jpg","type":"photo","url":"https://t.co/Oev5DkAEHd","ext_media_availability":{"status":"Available"},"features":{"large":{"faces":[{"x":90,"y":151,"h":183,"w":183}]},"medium":{"faces":[{"x":90,"y":151,"h":183,"w":183}]},"small":{"faces":[{"x":90,"y":151,"h":183,"w":183}]},"orig":{"faces":[{"x":90,"y":151,"h":183,"w":183}]}},"sizes":{"large":{"h":584,"w":351,"resize":"fit"},"medium":{"h":584,"w":351,"resize":"fit"},"small":{"h":584,"w":351,"resize":"fit"},"thumb":{"h":150,"w":150,"resize":"crop"}},"original_info":{"height":584,"width":351,"focus_rects":[{"x":0,"y":387,"w":351,"h":197},{"x":0,"y":233,"w":351,"h":351},{"x":0,"y":184,"w":351,"h":400},{"x":43,"y":0,"w":292,"h":584},{"x":0,"y":0,"w":351,"h":584}]},"media_results":{"result":{"media_key":"3_1976437591789477888"}}}],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"extended_entities":{"media":[{"display_url":"pic.x.com/Oev5DkAEHd","expanded_url":"https://x.com/jerber888/status/1976437594704707950/photo/1","id_str":"1976437591789477888","indices":[42,65],"media_key":"3_1976437591789477888","media_url_https":"https://pbs.twimg.com/media/G223hKNXIAAUAgp.jpg","type":"photo","url":"https://t.co/Oev5DkAEHd","ext_media_availability":{"status":"Available"},"features":{"large":{"faces":[{"x":90,"y":151,"h":183,"w":183}]},"medium":{"faces":[{"x":90,"y":151,"h":183,"w":183}]},"small":{"faces":[{"x":90,"y":151,"h":183,"w":183}]},"orig":{"faces":[{"x":90,"y":151,"h":183,"w":183}]}},"sizes":{"large":{"h":584,"w":351,"resize":"fit"},"medium":{"h":584,"w":351,"resize":"fit"},"small":{"h":584,"w":351,"resize":"fit"},"thumb":{"h":150,"w":150,"resize":"crop"}},"original_info":{"height":584,"width":351,"focus_rects":[{"x":0,"y":387,"w":351,"h":197},{"x":0,"y":233,"w":351,"h":351},{"x":0,"y":184,"w":351,"h":400},{"x":43,"y":0,"w":292,"h":584},{"x":0,"y":0,"w":351,"h":584}]},"media_results":{"result":{"media_key":"3_1976437591789477888"}}}]},"favorited":false,"lang":"en","possibly_sensitive":false,"possibly_sensitive_editable":true,"quoted_status_id_str":"1976304405369520242","quoted_status_permalink":{"url":"https://t.co/Jwn8ASrI8n","expanded":"https://twitter.com/reflection_ai/status/1976304405369520242","display":"x.com/reflection_ai/…"},"retweeted":false,"fact_check":null,"id":"1976437594704707950","view_count":2568,"bookmark_count":2,"created_at":1760054417000,"favorite_count":36,"quote_count":0,"reply_count":3,"retweet_count":1,"user_id_str":"903008975343554566","conversation_id_str":"1976437594704707950","full_text":"Finally my grandfather will take my calls https://t.co/Oev5DkAEHd","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,80],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","quoted_status_id_str":"1946260363256996244","quoted_status_permalink":{"url":"https://t.co/JceslQUFKB","expanded":"https://twitter.com/arcprize/status/1946260363256996244","display":"x.com/arcprize/statu…"},"retweeted":false,"fact_check":null,"id":"1946266864444682654","view_count":2608,"bookmark_count":1,"created_at":1752861155000,"favorite_count":30,"quote_count":0,"reply_count":1,"retweet_count":1,"user_id_str":"903008975343554566","conversation_id_str":"1946266864444682654","full_text":"This is the start of a new generation for AGI benchmarks. It’s also fun to play!","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,193],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1982191884593250452","view_count":2003,"bookmark_count":4,"created_at":1761426347000,"favorite_count":29,"quote_count":0,"reply_count":5,"retweet_count":1,"user_id_str":"903008975343554566","conversation_id_str":"1982191884593250452","full_text":"Solving hallucinations is a bigger deal than people think. Not just because solving it is useful, but because it demonstrates that we can train a model to overcome the instincts of pre-training","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,50],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1972434091770552558","view_count":3863,"bookmark_count":3,"created_at":1759099908000,"favorite_count":26,"quote_count":1,"reply_count":8,"retweet_count":2,"user_id_str":"903008975343554566","conversation_id_str":"1972434091770552558","full_text":"Coding models are general intelligence constrained","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null}],"ctweets":[{"bookmarked":false,"display_text_range":[0,160],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","quoted_status_id_str":"1968196086193066365","quoted_status_permalink":{"url":"https://t.co/O715u0DGse","expanded":"https://twitter.com/elonmusk/status/1968196086193066365","display":"x.com/elonmusk/statu…"},"retweeted":false,"fact_check":null,"id":"1968386345908453691","view_count":1435863,"bookmark_count":131,"created_at":1758134850000,"favorite_count":3184,"quote_count":14,"reply_count":174,"retweet_count":787,"user_id_str":"903008975343554566","conversation_id_str":"1968386345908453691","full_text":"I chose Grok 4 for my ARC-AGI solution because it had the most logical consistency when thinking for long periods of time. You can feel the RL that went into it","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,131],"entities":{"hashtags":[],"media":[{"display_url":"pic.x.com/aPKm3OnF0D","expanded_url":"https://x.com/jerber888/status/1968001933211471891/photo/1","id_str":"1968001754018000896","indices":[132,155],"media_key":"3_1968001754018000896","media_url_https":"https://pbs.twimg.com/media/G0-_KxmWsAAVCfi.jpg","type":"photo","url":"https://t.co/aPKm3OnF0D","ext_media_availability":{"status":"Available"},"features":{"large":{"faces":[]},"medium":{"faces":[]},"small":{"faces":[]},"orig":{"faces":[]}},"sizes":{"large":{"h":1158,"w":1774,"resize":"fit"},"medium":{"h":783,"w":1200,"resize":"fit"},"small":{"h":444,"w":680,"resize":"fit"},"thumb":{"h":150,"w":150,"resize":"crop"}},"original_info":{"height":1158,"width":1774,"focus_rects":[{"x":0,"y":0,"w":1774,"h":993},{"x":440,"y":0,"w":1158,"h":1158},{"x":511,"y":0,"w":1016,"h":1158},{"x":730,"y":0,"w":579,"h":1158},{"x":0,"y":0,"w":1774,"h":1158}]},"media_results":{"result":{"media_key":"3_1968001754018000896"}}}],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"1720665183188922368","name":"Grok","screen_name":"grok","indices":[58,63]}]},"extended_entities":{"media":[{"display_url":"pic.x.com/aPKm3OnF0D","expanded_url":"https://x.com/jerber888/status/1968001933211471891/photo/1","id_str":"1968001754018000896","indices":[132,155],"media_key":"3_1968001754018000896","media_url_https":"https://pbs.twimg.com/media/G0-_KxmWsAAVCfi.jpg","type":"photo","url":"https://t.co/aPKm3OnF0D","ext_media_availability":{"status":"Available"},"features":{"large":{"faces":[]},"medium":{"faces":[]},"small":{"faces":[]},"orig":{"faces":[]}},"sizes":{"large":{"h":1158,"w":1774,"resize":"fit"},"medium":{"h":783,"w":1200,"resize":"fit"},"small":{"h":444,"w":680,"resize":"fit"},"thumb":{"h":150,"w":150,"resize":"crop"}},"original_info":{"height":1158,"width":1774,"focus_rects":[{"x":0,"y":0,"w":1774,"h":993},{"x":440,"y":0,"w":1158,"h":1158},{"x":511,"y":0,"w":1016,"h":1158},{"x":730,"y":0,"w":579,"h":1158},{"x":0,"y":0,"w":1774,"h":1158}]},"media_results":{"result":{"media_key":"3_1968001754018000896"}}}]},"favorited":false,"lang":"en","possibly_sensitive":false,"possibly_sensitive_editable":true,"quoted_status_id_str":"1967998885701538060","quoted_status_permalink":{"url":"https://t.co/t2v7lvQnGq","expanded":"https://twitter.com/arcprize/status/1967998885701538060","display":"x.com/arcprize/statu…"},"retweeted":false,"fact_check":null,"id":"1968001933211471891","view_count":447022,"bookmark_count":341,"created_at":1758043199000,"favorite_count":1277,"quote_count":31,"reply_count":74,"retweet_count":99,"user_id_str":"903008975343554566","conversation_id_str":"1968001933211471891","full_text":"I'm back at the top of ARC-AGI with my new program. I use @grok 4 and multi-agent collaboration with evolutionary test-time compute https://t.co/aPKm3OnF0D","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,224],"entities":{"hashtags":[],"media":[{"display_url":"pic.x.com/CGkAyp0vMN","expanded_url":"https://x.com/jerber888/status/1987982067116777521/photo/1","id_str":"1987981765265272832","indices":[225,248],"media_key":"3_1987981765265272832","media_url_https":"https://pbs.twimg.com/media/G5a64fqXMAAL8FC.jpg","type":"photo","url":"https://t.co/CGkAyp0vMN","ext_media_availability":{"status":"Available"},"features":{"large":{"faces":[]},"medium":{"faces":[]},"small":{"faces":[]},"orig":{"faces":[]}},"sizes":{"large":{"h":1154,"w":1956,"resize":"fit"},"medium":{"h":708,"w":1200,"resize":"fit"},"small":{"h":401,"w":680,"resize":"fit"},"thumb":{"h":150,"w":150,"resize":"crop"}},"original_info":{"height":1154,"width":1956,"focus_rects":[{"x":0,"y":0,"w":1956,"h":1095},{"x":802,"y":0,"w":1154,"h":1154},{"x":944,"y":0,"w":1012,"h":1154},{"x":1323,"y":0,"w":577,"h":1154},{"x":0,"y":0,"w":1956,"h":1154}]},"media_results":{"result":{"media_key":"3_1987981765265272832"}}}],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"extended_entities":{"media":[{"display_url":"pic.x.com/CGkAyp0vMN","expanded_url":"https://x.com/jerber888/status/1987982067116777521/photo/1","id_str":"1987981765265272832","indices":[225,248],"media_key":"3_1987981765265272832","media_url_https":"https://pbs.twimg.com/media/G5a64fqXMAAL8FC.jpg","type":"photo","url":"https://t.co/CGkAyp0vMN","ext_media_availability":{"status":"Available"},"features":{"large":{"faces":[]},"medium":{"faces":[]},"small":{"faces":[]},"orig":{"faces":[]}},"sizes":{"large":{"h":1154,"w":1956,"resize":"fit"},"medium":{"h":708,"w":1200,"resize":"fit"},"small":{"h":401,"w":680,"resize":"fit"},"thumb":{"h":150,"w":150,"resize":"crop"}},"original_info":{"height":1154,"width":1956,"focus_rects":[{"x":0,"y":0,"w":1956,"h":1095},{"x":802,"y":0,"w":1154,"h":1154},{"x":944,"y":0,"w":1012,"h":1154},{"x":1323,"y":0,"w":577,"h":1154},{"x":0,"y":0,"w":1956,"h":1154}]},"media_results":{"result":{"media_key":"3_1987981765265272832"}}}]},"favorited":false,"lang":"en","possibly_sensitive":false,"possibly_sensitive_editable":true,"quoted_status_id_str":"1968001933211471891","quoted_status_permalink":{"url":"https://t.co/l4P9D5a1TN","expanded":"https://twitter.com/jerber888/status/1968001933211471891","display":"x.com/jerber888/stat…"},"retweeted":false,"fact_check":null,"id":"1987982067116777521","view_count":269951,"bookmark_count":591,"created_at":1762806834000,"favorite_count":1579,"quote_count":29,"reply_count":70,"retweet_count":147,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"I finally reached human-level performance (85%) on ARC-AGI v1 for under $10k and within 12 hours. I use the same multi-agent collaboration with evolutionary test-time compute, now powered by GPT-5 pro with lower parallelism. https://t.co/CGkAyp0vMN","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,113],"entities":{"hashtags":[],"media":[{"display_url":"pic.x.com/oqOFLQ3TUF","expanded_url":"https://x.com/jerber888/status/1865112099015291379/photo/1","id_str":"1865111668323229696","indices":[114,137],"media_key":"3_1865111668323229696","media_url_https":"https://pbs.twimg.com/media/GeI1LjjXoAAkswY.jpg","type":"photo","url":"https://t.co/oqOFLQ3TUF","ext_media_availability":{"status":"Available"},"features":{"large":{"faces":[{"x":953,"y":129,"h":165,"w":165},{"x":1335,"y":292,"h":157,"w":157}]},"medium":{"faces":[{"x":714,"y":96,"h":123,"w":123},{"x":1001,"y":219,"h":117,"w":117}]},"small":{"faces":[{"x":405,"y":54,"h":70,"w":70},{"x":567,"y":124,"h":66,"w":66}]},"orig":{"faces":[{"x":953,"y":129,"h":165,"w":165},{"x":1335,"y":292,"h":157,"w":157}]}},"sizes":{"large":{"h":606,"w":1600,"resize":"fit"},"medium":{"h":455,"w":1200,"resize":"fit"},"small":{"h":258,"w":680,"resize":"fit"},"thumb":{"h":150,"w":150,"resize":"crop"}},"original_info":{"height":606,"width":1600,"focus_rects":[{"x":0,"y":0,"w":1082,"h":606},{"x":217,"y":0,"w":606,"h":606},{"x":254,"y":0,"w":532,"h":606},{"x":369,"y":0,"w":303,"h":606},{"x":0,"y":0,"w":1600,"h":606}]},"media_results":{"result":{"media_key":"3_1865111668323229696"}}}],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"extended_entities":{"media":[{"display_url":"pic.x.com/oqOFLQ3TUF","expanded_url":"https://x.com/jerber888/status/1865112099015291379/photo/1","id_str":"1865111668323229696","indices":[114,137],"media_key":"3_1865111668323229696","media_url_https":"https://pbs.twimg.com/media/GeI1LjjXoAAkswY.jpg","type":"photo","url":"https://t.co/oqOFLQ3TUF","ext_media_availability":{"status":"Available"},"features":{"large":{"faces":[{"x":953,"y":129,"h":165,"w":165},{"x":1335,"y":292,"h":157,"w":157}]},"medium":{"faces":[{"x":714,"y":96,"h":123,"w":123},{"x":1001,"y":219,"h":117,"w":117}]},"small":{"faces":[{"x":405,"y":54,"h":70,"w":70},{"x":567,"y":124,"h":66,"w":66}]},"orig":{"faces":[{"x":953,"y":129,"h":165,"w":165},{"x":1335,"y":292,"h":157,"w":157}]}},"sizes":{"large":{"h":606,"w":1600,"resize":"fit"},"medium":{"h":455,"w":1200,"resize":"fit"},"small":{"h":258,"w":680,"resize":"fit"},"thumb":{"h":150,"w":150,"resize":"crop"}},"original_info":{"height":606,"width":1600,"focus_rects":[{"x":0,"y":0,"w":1082,"h":606},{"x":217,"y":0,"w":606,"h":606},{"x":254,"y":0,"w":532,"h":606},{"x":369,"y":0,"w":303,"h":606},{"x":0,"y":0,"w":1600,"h":606}]},"media_results":{"result":{"media_key":"3_1865111668323229696"}}}]},"favorited":false,"lang":"en","possibly_sensitive":false,"possibly_sensitive_editable":true,"quoted_status_id_str":"1865106291007410660","quoted_status_permalink":{"url":"https://t.co/OoRF76cqXs","expanded":"https://twitter.com/arcprize/status/1865106291007410660","display":"x.com/arcprize/statu…"},"retweeted":false,"fact_check":null,"id":"1865112099015291379","view_count":173995,"bookmark_count":444,"created_at":1733512351000,"favorite_count":1028,"quote_count":21,"reply_count":41,"retweet_count":100,"user_id_str":"903008975343554566","conversation_id_str":"1865112099015291379","full_text":"I just got first place on the public ARC-AGI benchmark using Claude Sonnet 3.5 and Evolutionary Test-time Compute https://t.co/oqOFLQ3TUF","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,276],"entities":{"hashtags":[],"media":[{"display_url":"pic.x.com/kDNOO8TbsV","expanded_url":"https://x.com/jerber888/status/1871309609387937986/photo/1","id_str":"1871307794994860033","indices":[277,300],"media_key":"3_1871307794994860033","media_url_https":"https://pbs.twimg.com/media/Gfg4hmIWkAEK_Tb.jpg","type":"photo","url":"https://t.co/kDNOO8TbsV","ext_media_availability":{"status":"Available"},"features":{"large":{"faces":[]},"medium":{"faces":[]},"small":{"faces":[]},"orig":{"faces":[]}},"sizes":{"large":{"h":872,"w":1816,"resize":"fit"},"medium":{"h":576,"w":1200,"resize":"fit"},"small":{"h":327,"w":680,"resize":"fit"},"thumb":{"h":150,"w":150,"resize":"crop"}},"original_info":{"height":872,"width":1816,"focus_rects":[{"x":0,"y":0,"w":1557,"h":872},{"x":0,"y":0,"w":872,"h":872},{"x":0,"y":0,"w":765,"h":872},{"x":8,"y":0,"w":436,"h":872},{"x":0,"y":0,"w":1816,"h":872}]},"media_results":{"result":{"media_key":"3_1871307794994860033"}}}],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"48008938","name":"Yann LeCun","screen_name":"ylecun","indices":[144,151]},{"id_str":"48008938","name":"Yann LeCun","screen_name":"ylecun","indices":[144,151]}]},"extended_entities":{"media":[{"display_url":"pic.x.com/kDNOO8TbsV","expanded_url":"https://x.com/jerber888/status/1871309609387937986/photo/1","id_str":"1871307794994860033","indices":[277,300],"media_key":"3_1871307794994860033","media_url_https":"https://pbs.twimg.com/media/Gfg4hmIWkAEK_Tb.jpg","type":"photo","url":"https://t.co/kDNOO8TbsV","ext_media_availability":{"status":"Available"},"features":{"large":{"faces":[]},"medium":{"faces":[]},"small":{"faces":[]},"orig":{"faces":[]}},"sizes":{"large":{"h":872,"w":1816,"resize":"fit"},"medium":{"h":576,"w":1200,"resize":"fit"},"small":{"h":327,"w":680,"resize":"fit"},"thumb":{"h":150,"w":150,"resize":"crop"}},"original_info":{"height":872,"width":1816,"focus_rects":[{"x":0,"y":0,"w":1557,"h":872},{"x":0,"y":0,"w":872,"h":872},{"x":0,"y":0,"w":765,"h":872},{"x":8,"y":0,"w":436,"h":872},{"x":0,"y":0,"w":1816,"h":872}]},"media_results":{"result":{"media_key":"3_1871307794994860033"}}}]},"favorited":false,"lang":"en","possibly_sensitive":false,"possibly_sensitive_editable":true,"retweeted":false,"fact_check":null,"id":"1871309609387937986","view_count":82020,"bookmark_count":776,"created_at":1734989952000,"favorite_count":956,"quote_count":11,"reply_count":22,"retweet_count":118,"user_id_str":"903008975343554566","conversation_id_str":"1871309609387937986","full_text":"Announcing LANG-JEPA — a new language model architecture I’m working on that optimizes in “concept” space instead of “token” space. Inspired by @ylecun's JEPA (I-JEPA for images, V-JEPA for video), LANG-JEPA asks: What if we train for conceptual understanding directly, rather than indirectly through token prediction?\n\nAlthough current LLMs do develop conceptual understanding, they do so via next-token prediction. In contrast, LANG-JEPA mirrors how humans think: we form high-level concepts first, then translate them into words. Here’s how it works:","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,153],"entities":{"hashtags":[],"media":[{"display_url":"pic.x.com/JqPhcNsCQ4","expanded_url":"https://x.com/jerber888/status/1976377788018933830/photo/1","id_str":"1976377437735841792","indices":[154,177],"media_key":"3_1976377437735841792","media_url_https":"https://pbs.twimg.com/media/G22Azu5XcAABkAI.jpg","type":"photo","url":"https://t.co/JqPhcNsCQ4","ext_media_availability":{"status":"Available"},"features":{"large":{"faces":[]},"medium":{"faces":[]},"small":{"faces":[]},"orig":{"faces":[]}},"sizes":{"large":{"h":1174,"w":1842,"resize":"fit"},"medium":{"h":765,"w":1200,"resize":"fit"},"small":{"h":433,"w":680,"resize":"fit"},"thumb":{"h":150,"w":150,"resize":"crop"}},"original_info":{"height":1174,"width":1842,"focus_rects":[{"x":0,"y":0,"w":1842,"h":1032},{"x":471,"y":0,"w":1174,"h":1174},{"x":543,"y":0,"w":1030,"h":1174},{"x":765,"y":0,"w":587,"h":1174},{"x":0,"y":0,"w":1842,"h":1174}]},"media_results":{"result":{"media_key":"3_1976377437735841792"}}}],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"extended_entities":{"media":[{"display_url":"pic.x.com/JqPhcNsCQ4","expanded_url":"https://x.com/jerber888/status/1976377788018933830/photo/1","id_str":"1976377437735841792","indices":[154,177],"media_key":"3_1976377437735841792","media_url_https":"https://pbs.twimg.com/media/G22Azu5XcAABkAI.jpg","type":"photo","url":"https://t.co/JqPhcNsCQ4","ext_media_availability":{"status":"Available"},"features":{"large":{"faces":[]},"medium":{"faces":[]},"small":{"faces":[]},"orig":{"faces":[]}},"sizes":{"large":{"h":1174,"w":1842,"resize":"fit"},"medium":{"h":765,"w":1200,"resize":"fit"},"small":{"h":433,"w":680,"resize":"fit"},"thumb":{"h":150,"w":150,"resize":"crop"}},"original_info":{"height":1174,"width":1842,"focus_rects":[{"x":0,"y":0,"w":1842,"h":1032},{"x":471,"y":0,"w":1174,"h":1174},{"x":543,"y":0,"w":1030,"h":1174},{"x":765,"y":0,"w":587,"h":1174},{"x":0,"y":0,"w":1842,"h":1174}]},"media_results":{"result":{"media_key":"3_1976377437735841792"}}}]},"favorited":false,"lang":"en","possibly_sensitive":false,"possibly_sensitive_editable":true,"quoted_status_id_str":"1976329182893441209","quoted_status_permalink":{"url":"https://t.co/JHONP88ao3","expanded":"https://twitter.com/arcprize/status/1976329182893441209","display":"x.com/arcprize/statu…"},"retweeted":false,"fact_check":null,"id":"1976377788018933830","view_count":38923,"bookmark_count":37,"created_at":1760040158000,"favorite_count":303,"quote_count":1,"reply_count":19,"retweet_count":20,"user_id_str":"903008975343554566","conversation_id_str":"1976377788018933830","full_text":"GPT-5 Pro is now the best LLM on ARC-AGI. I'm excited to see how my program does with it as the instruction generation model. I'm thinking I'll break 30% https://t.co/JqPhcNsCQ4","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,65],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1970965682218315981","view_count":2338,"bookmark_count":7,"created_at":1758749812000,"favorite_count":14,"quote_count":0,"reply_count":17,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1970965682218315981","full_text":"Why don’t biological neural nets overfit like artificial ones do?","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,250],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","quoted_status_id_str":"1976304405369520242","quoted_status_permalink":{"url":"https://t.co/Jwn8ASraiP","expanded":"https://twitter.com/reflection_ai/status/1976304405369520242","display":"x.com/reflection_ai/…"},"retweeted":false,"fact_check":null,"id":"1976305350505566280","view_count":20986,"bookmark_count":47,"created_at":1760022888000,"favorite_count":171,"quote_count":0,"reply_count":13,"retweet_count":5,"user_id_str":"903008975343554566","conversation_id_str":"1976305350505566280","full_text":"I joined Reflection to build the next frontier AI lab. We now have the team and resources to make it happen. We’re building open-weight foundation models and there’s a lot of work to be done. If you’re interested in reasoning and post-training, DM me","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,50],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1972434091770552558","view_count":3863,"bookmark_count":3,"created_at":1759099908000,"favorite_count":26,"quote_count":1,"reply_count":8,"retweet_count":2,"user_id_str":"903008975343554566","conversation_id_str":"1972434091770552558","full_text":"Coding models are general intelligence constrained","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,186],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"1846621278910009346","name":"Ndea","screen_name":"ndeainc","indices":[20,28]}]},"favorited":false,"lang":"en","quoted_status_id_str":"1879583863368032432","quoted_status_permalink":{"url":"https://t.co/UGwbtBoN7e","expanded":"https://twitter.com/fchollet/status/1879583863368032432","display":"x.com/fchollet/statu…"},"retweeted":false,"fact_check":null,"id":"1879601187621839187","view_count":20024,"bookmark_count":46,"created_at":1736966819000,"favorite_count":136,"quote_count":3,"reply_count":7,"retweet_count":12,"user_id_str":"903008975343554566","conversation_id_str":"1879601187621839187","full_text":"Excited to announce @ndeainc.\n\nWe're building a new architecture for AGI that merges deep learning with symbolic reasoning.\n\nIf you're interested in building general intelligence, DM me.","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,193],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1982191884593250452","view_count":2003,"bookmark_count":4,"created_at":1761426347000,"favorite_count":29,"quote_count":0,"reply_count":5,"retweet_count":1,"user_id_str":"903008975343554566","conversation_id_str":"1982191884593250452","full_text":"Solving hallucinations is a bigger deal than people think. Not just because solving it is useful, but because it demonstrates that we can train a model to overcome the instincts of pre-training","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,133],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"68746721","name":"François Chollet","screen_name":"fchollet","indices":[100,109]},{"id_str":"57444441","name":"Mike Knoop","screen_name":"mikeknoop","indices":[110,120]},{"id_str":"240008974","name":"Greg Kamradt","screen_name":"GregKamradt","indices":[121,133]}]},"favorited":false,"lang":"en","quoted_status_id_str":"1975560628657164426","quoted_status_permalink":{"url":"https://t.co/MKLXc92Jkh","expanded":"https://twitter.com/jm_alexia/status/1975560628657164426","display":"x.com/jm_alexia/stat…"},"retweeted":false,"fact_check":null,"id":"1976030514751602947","view_count":12715,"bookmark_count":21,"created_at":1759957362000,"favorite_count":130,"quote_count":0,"reply_count":4,"retweet_count":9,"user_id_str":"903008975343554566","conversation_id_str":"1976030514751602947","full_text":"No benchmark has had the type of technological and philosophical impact like ARC-AGI.\n\nGrateful for @fchollet @mikeknoop @GregKamradt","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,45],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1884369802321625428","view_count":806,"bookmark_count":1,"created_at":1738103745000,"favorite_count":8,"quote_count":1,"reply_count":3,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1884369802321625428","full_text":"AGI will at least be able to drive a car, no?","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,41],"entities":{"hashtags":[],"media":[{"display_url":"pic.x.com/Oev5DkAEHd","expanded_url":"https://x.com/jerber888/status/1976437594704707950/photo/1","id_str":"1976437591789477888","indices":[42,65],"media_key":"3_1976437591789477888","media_url_https":"https://pbs.twimg.com/media/G223hKNXIAAUAgp.jpg","type":"photo","url":"https://t.co/Oev5DkAEHd","ext_media_availability":{"status":"Available"},"features":{"large":{"faces":[{"x":90,"y":151,"h":183,"w":183}]},"medium":{"faces":[{"x":90,"y":151,"h":183,"w":183}]},"small":{"faces":[{"x":90,"y":151,"h":183,"w":183}]},"orig":{"faces":[{"x":90,"y":151,"h":183,"w":183}]}},"sizes":{"large":{"h":584,"w":351,"resize":"fit"},"medium":{"h":584,"w":351,"resize":"fit"},"small":{"h":584,"w":351,"resize":"fit"},"thumb":{"h":150,"w":150,"resize":"crop"}},"original_info":{"height":584,"width":351,"focus_rects":[{"x":0,"y":387,"w":351,"h":197},{"x":0,"y":233,"w":351,"h":351},{"x":0,"y":184,"w":351,"h":400},{"x":43,"y":0,"w":292,"h":584},{"x":0,"y":0,"w":351,"h":584}]},"media_results":{"result":{"media_key":"3_1976437591789477888"}}}],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"extended_entities":{"media":[{"display_url":"pic.x.com/Oev5DkAEHd","expanded_url":"https://x.com/jerber888/status/1976437594704707950/photo/1","id_str":"1976437591789477888","indices":[42,65],"media_key":"3_1976437591789477888","media_url_https":"https://pbs.twimg.com/media/G223hKNXIAAUAgp.jpg","type":"photo","url":"https://t.co/Oev5DkAEHd","ext_media_availability":{"status":"Available"},"features":{"large":{"faces":[{"x":90,"y":151,"h":183,"w":183}]},"medium":{"faces":[{"x":90,"y":151,"h":183,"w":183}]},"small":{"faces":[{"x":90,"y":151,"h":183,"w":183}]},"orig":{"faces":[{"x":90,"y":151,"h":183,"w":183}]}},"sizes":{"large":{"h":584,"w":351,"resize":"fit"},"medium":{"h":584,"w":351,"resize":"fit"},"small":{"h":584,"w":351,"resize":"fit"},"thumb":{"h":150,"w":150,"resize":"crop"}},"original_info":{"height":584,"width":351,"focus_rects":[{"x":0,"y":387,"w":351,"h":197},{"x":0,"y":233,"w":351,"h":351},{"x":0,"y":184,"w":351,"h":400},{"x":43,"y":0,"w":292,"h":584},{"x":0,"y":0,"w":351,"h":584}]},"media_results":{"result":{"media_key":"3_1976437591789477888"}}}]},"favorited":false,"lang":"en","possibly_sensitive":false,"possibly_sensitive_editable":true,"quoted_status_id_str":"1976304405369520242","quoted_status_permalink":{"url":"https://t.co/Jwn8ASrI8n","expanded":"https://twitter.com/reflection_ai/status/1976304405369520242","display":"x.com/reflection_ai/…"},"retweeted":false,"fact_check":null,"id":"1976437594704707950","view_count":2568,"bookmark_count":2,"created_at":1760054417000,"favorite_count":36,"quote_count":0,"reply_count":3,"retweet_count":1,"user_id_str":"903008975343554566","conversation_id_str":"1976437594704707950","full_text":"Finally my grandfather will take my calls https://t.co/Oev5DkAEHd","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,112],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1977039983585411147","view_count":3382,"bookmark_count":5,"created_at":1760198038000,"favorite_count":43,"quote_count":0,"reply_count":3,"retweet_count":1,"user_id_str":"903008975343554566","conversation_id_str":"1977039983585411147","full_text":"RL over CoT is the model learning to re-program itself to put the right answer in distribution of the next token","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,275],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1985743643169178080","view_count":1463,"bookmark_count":4,"created_at":1762273152000,"favorite_count":13,"quote_count":0,"reply_count":3,"retweet_count":1,"user_id_str":"903008975343554566","conversation_id_str":"1985743643169178080","full_text":"Teaching language models to have taste isn't just about making them better at writing or making jokes.\n\nTaste is everything. It's what leads to scientific discovery.\n\nIn an infinite sea of things to reason about, taste is what guides you to reason about the right things — to combine the right ideas together to deduce something interesting and new.\n\nThe discovery of special relativity was a product of good taste.","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null}],"activities":{"nreplies":[{"label":"2025-10-16","value":0,"startTime":1760486400000,"endTime":1760572800000,"tweets":[{"bookmarked":false,"display_text_range":[0,275],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1978585305679544427","view_count":90,"bookmark_count":0,"created_at":1760566471000,"favorite_count":4,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1978585305679544427","full_text":"Intelligence needs to be grown. Models need to develop coherent circuits that are reinforced. If you train models on traces they did not generate, there is no guarantee the information in that trace fits coherently in existing circuits. This is how models memorize instead of think for themselves. \n\nLearning should be on-policy. Take the unlikely but correct traces the model generates and reinforce those circuits. Then you build a model that thinks for itself.","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-17","value":0,"startTime":1760572800000,"endTime":1760659200000,"tweets":[{"bookmarked":false,"display_text_range":[73,86],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"500835474","name":"Pim de Witte","screen_name":"PimDeWitte","indices":[0,11]},{"id_str":"42226885","name":"Vinod Khosla","screen_name":"vkhosla","indices":[12,20]},{"id_str":"23809682","name":"Hemant Taneja","screen_name":"htaneja","indices":[21,29]},{"id_str":"1358786780259098625","name":"Eloi Alonso","screen_name":"EloiAlonso1","indices":[30,42]},{"id_str":"1307716628004704262","name":"Adam Jelley","screen_name":"AdamJelley2","indices":[43,55]},{"id_str":"1226783771195854848","name":"Vincent Micheli","screen_name":"micheli_vincent","indices":[56,72]}]},"favorited":false,"in_reply_to_screen_name":"PimDeWitte","lang":"en","retweeted":false,"fact_check":null,"id":"1978893031781568552","view_count":138,"bookmark_count":0,"created_at":1760639839000,"favorite_count":0,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1978825767053148457","full_text":"@PimDeWitte @vkhosla @htaneja @EloiAlonso1 @AdamJelley2 @micheli_vincent Congrats Pim!","in_reply_to_user_id_str":"500835474","in_reply_to_status_id_str":"1978825767053148457","is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-18","value":3,"startTime":1760659200000,"endTime":1760745600000,"tweets":[{"bookmarked":false,"display_text_range":[0,101],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","quoted_status_id_str":"1978196505392558087","quoted_status_permalink":{"url":"https://t.co/QhD1mEOFeS","expanded":"https://twitter.com/nicochristie/status/1978196505392558087","display":"x.com/nicochristie/s…"},"retweeted":false,"fact_check":null,"id":"1979230737405472922","view_count":1828,"bookmark_count":0,"created_at":1760720354000,"favorite_count":14,"quote_count":0,"reply_count":1,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1979230737405472922","full_text":"Accidentally said “impossible” instead of “out of distribution” and they kicked me out of my xAI tent","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,270],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","quoted_status_id_str":"1979237188220149938","quoted_status_permalink":{"url":"https://t.co/Ndv8TdVBSY","expanded":"https://twitter.com/MillionInt/status/1979237188220149938","display":"x.com/MillionInt/sta…"},"retweeted":false,"fact_check":null,"id":"1979240937281798378","view_count":730,"bookmark_count":5,"created_at":1760722786000,"favorite_count":12,"quote_count":0,"reply_count":2,"retweet_count":1,"user_id_str":"903008975343554566","conversation_id_str":"1979240937281798378","full_text":"Learning is like adding entropy to the neural network. Like adding a drop of colored liquid to clear water. The information diffuses across the weights. Unlearning that information is like undoing this diffusion. Reversing entropy. It takes a lot more energy to do that.","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null}]},{"label":"2025-10-19","value":0,"startTime":1760745600000,"endTime":1760832000000,"tweets":[]},{"label":"2025-10-20","value":0,"startTime":1760832000000,"endTime":1760918400000,"tweets":[]},{"label":"2025-10-21","value":1,"startTime":1760918400000,"endTime":1761004800000,"tweets":[{"bookmarked":false,"display_text_range":[0,110],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1980275836558213472","view_count":1446,"bookmark_count":2,"created_at":1760969525000,"favorite_count":20,"quote_count":0,"reply_count":1,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1980275836558213472","full_text":"To unlock the secrets of the universe, we just have to make interesting, true, and unlikely tokens more likely","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-22","value":0,"startTime":1761004800000,"endTime":1761091200000,"tweets":[]},{"label":"2025-10-23","value":0,"startTime":1761091200000,"endTime":1761177600000,"tweets":[]},{"label":"2025-10-24","value":0,"startTime":1761177600000,"endTime":1761264000000,"tweets":[]},{"label":"2025-10-25","value":0,"startTime":1761264000000,"endTime":1761350400000,"tweets":[{"bookmarked":false,"display_text_range":[17,70],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"4389158909","name":"Fernando Rojo","screen_name":"FernandoTheRojo","indices":[0,16]},{"id_str":"4389158909","name":"Fernando Rojo","screen_name":"FernandoTheRojo","indices":[17,33]}]},"favorited":false,"in_reply_to_screen_name":"FernandoTheRojo","lang":"en","retweeted":false,"fact_check":null,"id":"1981519914528821421","view_count":489,"bookmark_count":0,"created_at":1761266137000,"favorite_count":2,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1981509473295544350","full_text":"@FernandoTheRojo @FernandoTheRojo is the best app builder in the world","in_reply_to_user_id_str":"4389158909","in_reply_to_status_id_str":"1981509473295544350","is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-26","value":5,"startTime":1761350400000,"endTime":1761436800000,"tweets":[{"bookmarked":false,"display_text_range":[0,193],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1982191884593250452","view_count":2003,"bookmark_count":4,"created_at":1761426347000,"favorite_count":29,"quote_count":0,"reply_count":5,"retweet_count":1,"user_id_str":"903008975343554566","conversation_id_str":"1982191884593250452","full_text":"Solving hallucinations is a bigger deal than people think. Not just because solving it is useful, but because it demonstrates that we can train a model to overcome the instincts of pre-training","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,81],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1982190918326186171","view_count":685,"bookmark_count":0,"created_at":1761426116000,"favorite_count":1,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1982190918326186171","full_text":"A model using off-policy trained weights to respond is like a human under a spell","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-27","value":0,"startTime":1761436800000,"endTime":1761523200000,"tweets":[]},{"label":"2025-10-28","value":0,"startTime":1761523200000,"endTime":1761609600000,"tweets":[]},{"label":"2025-10-29","value":0,"startTime":1761609600000,"endTime":1761696000000,"tweets":[]},{"label":"2025-10-30","value":1,"startTime":1761696000000,"endTime":1761782400000,"tweets":[{"bookmarked":false,"display_text_range":[0,140],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1983562418782446066","view_count":1530,"bookmark_count":1,"created_at":1761753108000,"favorite_count":19,"quote_count":1,"reply_count":1,"retweet_count":1,"user_id_str":"903008975343554566","conversation_id_str":"1983562418782446066","full_text":"we use our continuous human brains to build symbolic computer systems to build continuous computer brains to build symbolic computer systems","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-31","value":0,"startTime":1761782400000,"endTime":1761868800000,"tweets":[]},{"label":"2025-11-01","value":0,"startTime":1761868800000,"endTime":1761955200000,"tweets":[]},{"label":"2025-11-02","value":0,"startTime":1761955200000,"endTime":1762041600000,"tweets":[{"bookmarked":false,"display_text_range":[0,44],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","quoted_status_id_str":"1984662921264120094","quoted_status_permalink":{"url":"https://t.co/s5M12jIb1I","expanded":"https://twitter.com/mindsai_jack/status/1984662921264120094","display":"x.com/mindsai_jack/s…"},"retweeted":false,"fact_check":null,"id":"1984665905972859177","view_count":2789,"bookmark_count":3,"created_at":1762016199000,"favorite_count":9,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1984665905972859177","full_text":"This was fun and Jack’s speech was great too","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null}]},{"label":"2025-11-03","value":0,"startTime":1762041600000,"endTime":1762128000000,"tweets":[]},{"label":"2025-11-04","value":0,"startTime":1762128000000,"endTime":1762214400000,"tweets":[]},{"label":"2025-11-05","value":3,"startTime":1762214400000,"endTime":1762300800000,"tweets":[{"bookmarked":false,"display_text_range":[0,275],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1985743643169178080","view_count":1463,"bookmark_count":4,"created_at":1762273152000,"favorite_count":13,"quote_count":0,"reply_count":3,"retweet_count":1,"user_id_str":"903008975343554566","conversation_id_str":"1985743643169178080","full_text":"Teaching language models to have taste isn't just about making them better at writing or making jokes.\n\nTaste is everything. It's what leads to scientific discovery.\n\nIn an infinite sea of things to reason about, taste is what guides you to reason about the right things — to combine the right ideas together to deduce something interesting and new.\n\nThe discovery of special relativity was a product of good taste.","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-11-06","value":0,"startTime":1762300800000,"endTime":1762387200000,"tweets":[]},{"label":"2025-11-07","value":0,"startTime":1762387200000,"endTime":1762473600000,"tweets":[{"bookmarked":false,"display_text_range":[0,96],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","quoted_status_id_str":"1986461871462096920","quoted_status_permalink":{"url":"https://t.co/J1VFIbBDHn","expanded":"https://twitter.com/skiminok/status/1986461871462096920","display":"x.com/skiminok/statu…"},"retweeted":false,"fact_check":null,"id":"1986474907484783018","view_count":1347,"bookmark_count":0,"created_at":1762447499000,"favorite_count":6,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1986474907484783018","full_text":"Never a dull moment talking language models with Alex. Really excited to get to do it more often","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null}]},{"label":"2025-11-08","value":0,"startTime":1762473600000,"endTime":1762560000000,"tweets":[]},{"label":"2025-11-09","value":0,"startTime":1762560000000,"endTime":1762646400000,"tweets":[]},{"label":"2025-11-10","value":0,"startTime":1762646400000,"endTime":1762732800000,"tweets":[]},{"label":"2025-11-11","value":78,"startTime":1762732800000,"endTime":1762819200000,"tweets":[{"bookmarked":false,"display_text_range":[0,224],"entities":{"hashtags":[],"media":[{"display_url":"pic.x.com/CGkAyp0vMN","expanded_url":"https://x.com/jerber888/status/1987982067116777521/photo/1","id_str":"1987981765265272832","indices":[225,248],"media_key":"3_1987981765265272832","media_url_https":"https://pbs.twimg.com/media/G5a64fqXMAAL8FC.jpg","type":"photo","url":"https://t.co/CGkAyp0vMN","ext_media_availability":{"status":"Available"},"features":{"large":{"faces":[]},"medium":{"faces":[]},"small":{"faces":[]},"orig":{"faces":[]}},"sizes":{"large":{"h":1154,"w":1956,"resize":"fit"},"medium":{"h":708,"w":1200,"resize":"fit"},"small":{"h":401,"w":680,"resize":"fit"},"thumb":{"h":150,"w":150,"resize":"crop"}},"original_info":{"height":1154,"width":1956,"focus_rects":[{"x":0,"y":0,"w":1956,"h":1095},{"x":802,"y":0,"w":1154,"h":1154},{"x":944,"y":0,"w":1012,"h":1154},{"x":1323,"y":0,"w":577,"h":1154},{"x":0,"y":0,"w":1956,"h":1154}]},"media_results":{"result":{"media_key":"3_1987981765265272832"}}}],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"extended_entities":{"media":[{"display_url":"pic.x.com/CGkAyp0vMN","expanded_url":"https://x.com/jerber888/status/1987982067116777521/photo/1","id_str":"1987981765265272832","indices":[225,248],"media_key":"3_1987981765265272832","media_url_https":"https://pbs.twimg.com/media/G5a64fqXMAAL8FC.jpg","type":"photo","url":"https://t.co/CGkAyp0vMN","ext_media_availability":{"status":"Available"},"features":{"large":{"faces":[]},"medium":{"faces":[]},"small":{"faces":[]},"orig":{"faces":[]}},"sizes":{"large":{"h":1154,"w":1956,"resize":"fit"},"medium":{"h":708,"w":1200,"resize":"fit"},"small":{"h":401,"w":680,"resize":"fit"},"thumb":{"h":150,"w":150,"resize":"crop"}},"original_info":{"height":1154,"width":1956,"focus_rects":[{"x":0,"y":0,"w":1956,"h":1095},{"x":802,"y":0,"w":1154,"h":1154},{"x":944,"y":0,"w":1012,"h":1154},{"x":1323,"y":0,"w":577,"h":1154},{"x":0,"y":0,"w":1956,"h":1154}]},"media_results":{"result":{"media_key":"3_1987981765265272832"}}}]},"favorited":false,"lang":"en","possibly_sensitive":false,"possibly_sensitive_editable":true,"quoted_status_id_str":"1968001933211471891","quoted_status_permalink":{"url":"https://t.co/l4P9D5a1TN","expanded":"https://twitter.com/jerber888/status/1968001933211471891","display":"x.com/jerber888/stat…"},"retweeted":false,"fact_check":null,"id":"1987982067116777521","view_count":269951,"bookmark_count":591,"created_at":1762806834000,"favorite_count":1579,"quote_count":29,"reply_count":70,"retweet_count":147,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"I finally reached human-level performance (85%) on ARC-AGI v1 for under $10k and within 12 hours. I use the same multi-agent collaboration with evolutionary test-time compute, now powered by GPT-5 pro with lower parallelism. https://t.co/CGkAyp0vMN","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,275],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"1773935160192647168","name":"ARC Prize","screen_name":"arcprize","indices":[168,177]},{"id_str":"1773935160192647168","name":"ARC Prize","screen_name":"arcprize","indices":[168,177]}]},"favorited":false,"in_reply_to_screen_name":"jerber888","lang":"en","retweeted":false,"fact_check":null,"id":"1987982069712945649","view_count":6572,"bookmark_count":6,"created_at":1762806834000,"favorite_count":57,"quote_count":0,"reply_count":4,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"I ran it on a random sample of 100 tasks from the 2024 eval set. It got 88/100 and averaged $27 per task. The score still needs to be verified on the hidden set by the @arcprize, but in my past submissions, the hidden set subtracts a few percent from the score and adds a few dollars. I put a box where I expect the official score to land within.","in_reply_to_user_id_str":"903008975343554566","in_reply_to_status_id_str":"1987982067116777521","is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,279],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"825088493764407298","name":"Noam Brown","screen_name":"polynoamial","indices":[288,300]}]},"favorited":false,"in_reply_to_screen_name":"jerber888","lang":"en","retweeted":false,"fact_check":null,"id":"1987982071541993700","view_count":6923,"bookmark_count":7,"created_at":1762806835000,"favorite_count":75,"quote_count":0,"reply_count":2,"retweet_count":2,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"GPT-5 pro is the best reasoning model today. It thinks coherently for hours and hours. My agent coordination logic is likely an intermediate step before the models learn to do this type of long horizon coordination on their own, end to end. It’s a hard problem but I wouldn’t bet against @polynoamial and team.","in_reply_to_user_id_str":"903008975343554566","in_reply_to_status_id_str":"1987982069712945649","is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,123],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[{"display_url":"github.com/jerber/arc-lan…","expanded_url":"https://github.com/jerber/arc-lang-public","url":"https://t.co/QnGjsN98ji","indices":[50,73]},{"display_url":"kaggle.com/code/jerber/je…","expanded_url":"https://www.kaggle.com/code/jerber/jeremy-arc2","url":"https://t.co/1jhzROG4SL","indices":[100,123]}],"user_mentions":[]},"favorited":false,"in_reply_to_screen_name":"jerber888","lang":"en","possibly_sensitive":false,"possibly_sensitive_editable":true,"retweeted":false,"fact_check":null,"id":"1987982214332580197","view_count":6525,"bookmark_count":39,"created_at":1762806869000,"favorite_count":64,"quote_count":1,"reply_count":2,"retweet_count":2,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"You can run this code yourself, it’s open source: https://t.co/QnGjsN98ji. And the kaggle notebook: https://t.co/1jhzROG4SL","in_reply_to_user_id_str":"903008975343554566","in_reply_to_status_id_str":"1987982071541993700","is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-11-12","value":2,"startTime":1762819200000,"endTime":1762905600000,"tweets":[{"bookmarked":false,"display_text_range":[0,48],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1988382853965770765","view_count":576,"bookmark_count":1,"created_at":1762902389000,"favorite_count":16,"quote_count":0,"reply_count":1,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1988382853965770765","full_text":"be the change you want to see in the world model","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,52],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[{"display_url":"youtube.com/watch?v=FcnLiP…","expanded_url":"https://www.youtube.com/watch?v=FcnLiPyfRZM","url":"https://t.co/jEBQGMsm62","indices":[29,52]}],"user_mentions":[]},"favorited":false,"lang":"en","possibly_sensitive":false,"possibly_sensitive_editable":true,"quoted_status_id_str":"1988165006073229659","quoted_status_permalink":{"url":"https://t.co/l99V6sHyI3","expanded":"https://twitter.com/MLStreetTalk/status/1988165006073229659","display":"x.com/MLStreetTalk/s…"},"retweeted":false,"fact_check":null,"id":"1988280687200530841","view_count":4997,"bookmark_count":5,"created_at":1762878030000,"favorite_count":21,"quote_count":0,"reply_count":1,"retweet_count":2,"user_id_str":"903008975343554566","conversation_id_str":"1988280687200530841","full_text":"This was a fun conversation: https://t.co/jEBQGMsm62","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,56],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[{"display_url":"jeremyberman.substack.com/p/how-i-got-th…","expanded_url":"https://jeremyberman.substack.com/p/how-i-got-the-highest-score-on-arc-agi-again","url":"https://t.co/xWiPB2Z0aI","indices":[33,56]}],"user_mentions":[]},"favorited":false,"in_reply_to_screen_name":"jerber888","lang":"en","possibly_sensitive":false,"possibly_sensitive_editable":true,"retweeted":false,"fact_check":null,"id":"1988278615377916103","view_count":1461,"bookmark_count":6,"created_at":1762877536000,"favorite_count":9,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"And technical blog post is here: https://t.co/xWiPB2Z0aI","in_reply_to_user_id_str":"903008975343554566","in_reply_to_status_id_str":"1987982214332580197","is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[9,52],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"536428919","name":"JFPuget 🇺🇦🇨🇦🇬🇱","screen_name":"JFPuget","indices":[0,8]}]},"favorited":false,"in_reply_to_screen_name":"JFPuget","lang":"en","retweeted":false,"fact_check":null,"id":"1988253927104471544","view_count":1040,"bookmark_count":0,"created_at":1762871650000,"favorite_count":8,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"@JFPuget Most likely but was focused on v1 for this.","in_reply_to_user_id_str":"536428919","in_reply_to_status_id_str":"1988249984907571356","is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[13,26],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"1194889317388374016","name":"Matt Shumer","screen_name":"mattshumer_","indices":[0,12]}]},"favorited":false,"in_reply_to_screen_name":"mattshumer_","lang":"en","retweeted":false,"fact_check":null,"id":"1988334412690571395","view_count":221,"bookmark_count":0,"created_at":1762890840000,"favorite_count":3,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1988331100935254226","full_text":"@mattshumer_ thanks Matt 🙏","in_reply_to_user_id_str":"1194889317388374016","in_reply_to_status_id_str":"1988331100935254226","is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-11-13","value":0,"startTime":1762905600000,"endTime":1762992000000,"tweets":[]},{"label":"2025-11-14","value":0,"startTime":1762992000000,"endTime":1763078400000,"tweets":[]},{"label":"2025-11-15","value":0,"startTime":1763078400000,"endTime":1763164800000,"tweets":[]}],"nbookmarks":[{"label":"2025-10-16","value":0,"startTime":1760486400000,"endTime":1760572800000,"tweets":[{"bookmarked":false,"display_text_range":[0,275],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1978585305679544427","view_count":90,"bookmark_count":0,"created_at":1760566471000,"favorite_count":4,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1978585305679544427","full_text":"Intelligence needs to be grown. Models need to develop coherent circuits that are reinforced. If you train models on traces they did not generate, there is no guarantee the information in that trace fits coherently in existing circuits. This is how models memorize instead of think for themselves. \n\nLearning should be on-policy. Take the unlikely but correct traces the model generates and reinforce those circuits. Then you build a model that thinks for itself.","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-17","value":0,"startTime":1760572800000,"endTime":1760659200000,"tweets":[{"bookmarked":false,"display_text_range":[73,86],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"500835474","name":"Pim de Witte","screen_name":"PimDeWitte","indices":[0,11]},{"id_str":"42226885","name":"Vinod Khosla","screen_name":"vkhosla","indices":[12,20]},{"id_str":"23809682","name":"Hemant Taneja","screen_name":"htaneja","indices":[21,29]},{"id_str":"1358786780259098625","name":"Eloi Alonso","screen_name":"EloiAlonso1","indices":[30,42]},{"id_str":"1307716628004704262","name":"Adam Jelley","screen_name":"AdamJelley2","indices":[43,55]},{"id_str":"1226783771195854848","name":"Vincent Micheli","screen_name":"micheli_vincent","indices":[56,72]}]},"favorited":false,"in_reply_to_screen_name":"PimDeWitte","lang":"en","retweeted":false,"fact_check":null,"id":"1978893031781568552","view_count":138,"bookmark_count":0,"created_at":1760639839000,"favorite_count":0,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1978825767053148457","full_text":"@PimDeWitte @vkhosla @htaneja @EloiAlonso1 @AdamJelley2 @micheli_vincent Congrats Pim!","in_reply_to_user_id_str":"500835474","in_reply_to_status_id_str":"1978825767053148457","is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-18","value":5,"startTime":1760659200000,"endTime":1760745600000,"tweets":[{"bookmarked":false,"display_text_range":[0,101],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","quoted_status_id_str":"1978196505392558087","quoted_status_permalink":{"url":"https://t.co/QhD1mEOFeS","expanded":"https://twitter.com/nicochristie/status/1978196505392558087","display":"x.com/nicochristie/s…"},"retweeted":false,"fact_check":null,"id":"1979230737405472922","view_count":1828,"bookmark_count":0,"created_at":1760720354000,"favorite_count":14,"quote_count":0,"reply_count":1,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1979230737405472922","full_text":"Accidentally said “impossible” instead of “out of distribution” and they kicked me out of my xAI tent","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,270],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","quoted_status_id_str":"1979237188220149938","quoted_status_permalink":{"url":"https://t.co/Ndv8TdVBSY","expanded":"https://twitter.com/MillionInt/status/1979237188220149938","display":"x.com/MillionInt/sta…"},"retweeted":false,"fact_check":null,"id":"1979240937281798378","view_count":730,"bookmark_count":5,"created_at":1760722786000,"favorite_count":12,"quote_count":0,"reply_count":2,"retweet_count":1,"user_id_str":"903008975343554566","conversation_id_str":"1979240937281798378","full_text":"Learning is like adding entropy to the neural network. Like adding a drop of colored liquid to clear water. The information diffuses across the weights. Unlearning that information is like undoing this diffusion. Reversing entropy. It takes a lot more energy to do that.","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null}]},{"label":"2025-10-19","value":0,"startTime":1760745600000,"endTime":1760832000000,"tweets":[]},{"label":"2025-10-20","value":0,"startTime":1760832000000,"endTime":1760918400000,"tweets":[]},{"label":"2025-10-21","value":2,"startTime":1760918400000,"endTime":1761004800000,"tweets":[{"bookmarked":false,"display_text_range":[0,110],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1980275836558213472","view_count":1446,"bookmark_count":2,"created_at":1760969525000,"favorite_count":20,"quote_count":0,"reply_count":1,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1980275836558213472","full_text":"To unlock the secrets of the universe, we just have to make interesting, true, and unlikely tokens more likely","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-22","value":0,"startTime":1761004800000,"endTime":1761091200000,"tweets":[]},{"label":"2025-10-23","value":0,"startTime":1761091200000,"endTime":1761177600000,"tweets":[]},{"label":"2025-10-24","value":0,"startTime":1761177600000,"endTime":1761264000000,"tweets":[]},{"label":"2025-10-25","value":0,"startTime":1761264000000,"endTime":1761350400000,"tweets":[{"bookmarked":false,"display_text_range":[17,70],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"4389158909","name":"Fernando Rojo","screen_name":"FernandoTheRojo","indices":[0,16]},{"id_str":"4389158909","name":"Fernando Rojo","screen_name":"FernandoTheRojo","indices":[17,33]}]},"favorited":false,"in_reply_to_screen_name":"FernandoTheRojo","lang":"en","retweeted":false,"fact_check":null,"id":"1981519914528821421","view_count":489,"bookmark_count":0,"created_at":1761266137000,"favorite_count":2,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1981509473295544350","full_text":"@FernandoTheRojo @FernandoTheRojo is the best app builder in the world","in_reply_to_user_id_str":"4389158909","in_reply_to_status_id_str":"1981509473295544350","is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-26","value":4,"startTime":1761350400000,"endTime":1761436800000,"tweets":[{"bookmarked":false,"display_text_range":[0,193],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1982191884593250452","view_count":2003,"bookmark_count":4,"created_at":1761426347000,"favorite_count":29,"quote_count":0,"reply_count":5,"retweet_count":1,"user_id_str":"903008975343554566","conversation_id_str":"1982191884593250452","full_text":"Solving hallucinations is a bigger deal than people think. Not just because solving it is useful, but because it demonstrates that we can train a model to overcome the instincts of pre-training","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,81],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1982190918326186171","view_count":685,"bookmark_count":0,"created_at":1761426116000,"favorite_count":1,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1982190918326186171","full_text":"A model using off-policy trained weights to respond is like a human under a spell","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-27","value":0,"startTime":1761436800000,"endTime":1761523200000,"tweets":[]},{"label":"2025-10-28","value":0,"startTime":1761523200000,"endTime":1761609600000,"tweets":[]},{"label":"2025-10-29","value":0,"startTime":1761609600000,"endTime":1761696000000,"tweets":[]},{"label":"2025-10-30","value":1,"startTime":1761696000000,"endTime":1761782400000,"tweets":[{"bookmarked":false,"display_text_range":[0,140],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1983562418782446066","view_count":1530,"bookmark_count":1,"created_at":1761753108000,"favorite_count":19,"quote_count":1,"reply_count":1,"retweet_count":1,"user_id_str":"903008975343554566","conversation_id_str":"1983562418782446066","full_text":"we use our continuous human brains to build symbolic computer systems to build continuous computer brains to build symbolic computer systems","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-31","value":0,"startTime":1761782400000,"endTime":1761868800000,"tweets":[]},{"label":"2025-11-01","value":0,"startTime":1761868800000,"endTime":1761955200000,"tweets":[]},{"label":"2025-11-02","value":3,"startTime":1761955200000,"endTime":1762041600000,"tweets":[{"bookmarked":false,"display_text_range":[0,44],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","quoted_status_id_str":"1984662921264120094","quoted_status_permalink":{"url":"https://t.co/s5M12jIb1I","expanded":"https://twitter.com/mindsai_jack/status/1984662921264120094","display":"x.com/mindsai_jack/s…"},"retweeted":false,"fact_check":null,"id":"1984665905972859177","view_count":2789,"bookmark_count":3,"created_at":1762016199000,"favorite_count":9,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1984665905972859177","full_text":"This was fun and Jack’s speech was great too","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null}]},{"label":"2025-11-03","value":0,"startTime":1762041600000,"endTime":1762128000000,"tweets":[]},{"label":"2025-11-04","value":0,"startTime":1762128000000,"endTime":1762214400000,"tweets":[]},{"label":"2025-11-05","value":4,"startTime":1762214400000,"endTime":1762300800000,"tweets":[{"bookmarked":false,"display_text_range":[0,275],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1985743643169178080","view_count":1463,"bookmark_count":4,"created_at":1762273152000,"favorite_count":13,"quote_count":0,"reply_count":3,"retweet_count":1,"user_id_str":"903008975343554566","conversation_id_str":"1985743643169178080","full_text":"Teaching language models to have taste isn't just about making them better at writing or making jokes.\n\nTaste is everything. It's what leads to scientific discovery.\n\nIn an infinite sea of things to reason about, taste is what guides you to reason about the right things — to combine the right ideas together to deduce something interesting and new.\n\nThe discovery of special relativity was a product of good taste.","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-11-06","value":0,"startTime":1762300800000,"endTime":1762387200000,"tweets":[]},{"label":"2025-11-07","value":0,"startTime":1762387200000,"endTime":1762473600000,"tweets":[{"bookmarked":false,"display_text_range":[0,96],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","quoted_status_id_str":"1986461871462096920","quoted_status_permalink":{"url":"https://t.co/J1VFIbBDHn","expanded":"https://twitter.com/skiminok/status/1986461871462096920","display":"x.com/skiminok/statu…"},"retweeted":false,"fact_check":null,"id":"1986474907484783018","view_count":1347,"bookmark_count":0,"created_at":1762447499000,"favorite_count":6,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1986474907484783018","full_text":"Never a dull moment talking language models with Alex. Really excited to get to do it more often","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null}]},{"label":"2025-11-08","value":0,"startTime":1762473600000,"endTime":1762560000000,"tweets":[]},{"label":"2025-11-09","value":0,"startTime":1762560000000,"endTime":1762646400000,"tweets":[]},{"label":"2025-11-10","value":0,"startTime":1762646400000,"endTime":1762732800000,"tweets":[]},{"label":"2025-11-11","value":643,"startTime":1762732800000,"endTime":1762819200000,"tweets":[{"bookmarked":false,"display_text_range":[0,224],"entities":{"hashtags":[],"media":[{"display_url":"pic.x.com/CGkAyp0vMN","expanded_url":"https://x.com/jerber888/status/1987982067116777521/photo/1","id_str":"1987981765265272832","indices":[225,248],"media_key":"3_1987981765265272832","media_url_https":"https://pbs.twimg.com/media/G5a64fqXMAAL8FC.jpg","type":"photo","url":"https://t.co/CGkAyp0vMN","ext_media_availability":{"status":"Available"},"features":{"large":{"faces":[]},"medium":{"faces":[]},"small":{"faces":[]},"orig":{"faces":[]}},"sizes":{"large":{"h":1154,"w":1956,"resize":"fit"},"medium":{"h":708,"w":1200,"resize":"fit"},"small":{"h":401,"w":680,"resize":"fit"},"thumb":{"h":150,"w":150,"resize":"crop"}},"original_info":{"height":1154,"width":1956,"focus_rects":[{"x":0,"y":0,"w":1956,"h":1095},{"x":802,"y":0,"w":1154,"h":1154},{"x":944,"y":0,"w":1012,"h":1154},{"x":1323,"y":0,"w":577,"h":1154},{"x":0,"y":0,"w":1956,"h":1154}]},"media_results":{"result":{"media_key":"3_1987981765265272832"}}}],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"extended_entities":{"media":[{"display_url":"pic.x.com/CGkAyp0vMN","expanded_url":"https://x.com/jerber888/status/1987982067116777521/photo/1","id_str":"1987981765265272832","indices":[225,248],"media_key":"3_1987981765265272832","media_url_https":"https://pbs.twimg.com/media/G5a64fqXMAAL8FC.jpg","type":"photo","url":"https://t.co/CGkAyp0vMN","ext_media_availability":{"status":"Available"},"features":{"large":{"faces":[]},"medium":{"faces":[]},"small":{"faces":[]},"orig":{"faces":[]}},"sizes":{"large":{"h":1154,"w":1956,"resize":"fit"},"medium":{"h":708,"w":1200,"resize":"fit"},"small":{"h":401,"w":680,"resize":"fit"},"thumb":{"h":150,"w":150,"resize":"crop"}},"original_info":{"height":1154,"width":1956,"focus_rects":[{"x":0,"y":0,"w":1956,"h":1095},{"x":802,"y":0,"w":1154,"h":1154},{"x":944,"y":0,"w":1012,"h":1154},{"x":1323,"y":0,"w":577,"h":1154},{"x":0,"y":0,"w":1956,"h":1154}]},"media_results":{"result":{"media_key":"3_1987981765265272832"}}}]},"favorited":false,"lang":"en","possibly_sensitive":false,"possibly_sensitive_editable":true,"quoted_status_id_str":"1968001933211471891","quoted_status_permalink":{"url":"https://t.co/l4P9D5a1TN","expanded":"https://twitter.com/jerber888/status/1968001933211471891","display":"x.com/jerber888/stat…"},"retweeted":false,"fact_check":null,"id":"1987982067116777521","view_count":269951,"bookmark_count":591,"created_at":1762806834000,"favorite_count":1579,"quote_count":29,"reply_count":70,"retweet_count":147,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"I finally reached human-level performance (85%) on ARC-AGI v1 for under $10k and within 12 hours. I use the same multi-agent collaboration with evolutionary test-time compute, now powered by GPT-5 pro with lower parallelism. https://t.co/CGkAyp0vMN","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,275],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"1773935160192647168","name":"ARC Prize","screen_name":"arcprize","indices":[168,177]},{"id_str":"1773935160192647168","name":"ARC Prize","screen_name":"arcprize","indices":[168,177]}]},"favorited":false,"in_reply_to_screen_name":"jerber888","lang":"en","retweeted":false,"fact_check":null,"id":"1987982069712945649","view_count":6572,"bookmark_count":6,"created_at":1762806834000,"favorite_count":57,"quote_count":0,"reply_count":4,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"I ran it on a random sample of 100 tasks from the 2024 eval set. It got 88/100 and averaged $27 per task. The score still needs to be verified on the hidden set by the @arcprize, but in my past submissions, the hidden set subtracts a few percent from the score and adds a few dollars. I put a box where I expect the official score to land within.","in_reply_to_user_id_str":"903008975343554566","in_reply_to_status_id_str":"1987982067116777521","is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,279],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"825088493764407298","name":"Noam Brown","screen_name":"polynoamial","indices":[288,300]}]},"favorited":false,"in_reply_to_screen_name":"jerber888","lang":"en","retweeted":false,"fact_check":null,"id":"1987982071541993700","view_count":6923,"bookmark_count":7,"created_at":1762806835000,"favorite_count":75,"quote_count":0,"reply_count":2,"retweet_count":2,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"GPT-5 pro is the best reasoning model today. It thinks coherently for hours and hours. My agent coordination logic is likely an intermediate step before the models learn to do this type of long horizon coordination on their own, end to end. It’s a hard problem but I wouldn’t bet against @polynoamial and team.","in_reply_to_user_id_str":"903008975343554566","in_reply_to_status_id_str":"1987982069712945649","is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,123],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[{"display_url":"github.com/jerber/arc-lan…","expanded_url":"https://github.com/jerber/arc-lang-public","url":"https://t.co/QnGjsN98ji","indices":[50,73]},{"display_url":"kaggle.com/code/jerber/je…","expanded_url":"https://www.kaggle.com/code/jerber/jeremy-arc2","url":"https://t.co/1jhzROG4SL","indices":[100,123]}],"user_mentions":[]},"favorited":false,"in_reply_to_screen_name":"jerber888","lang":"en","possibly_sensitive":false,"possibly_sensitive_editable":true,"retweeted":false,"fact_check":null,"id":"1987982214332580197","view_count":6525,"bookmark_count":39,"created_at":1762806869000,"favorite_count":64,"quote_count":1,"reply_count":2,"retweet_count":2,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"You can run this code yourself, it’s open source: https://t.co/QnGjsN98ji. And the kaggle notebook: https://t.co/1jhzROG4SL","in_reply_to_user_id_str":"903008975343554566","in_reply_to_status_id_str":"1987982071541993700","is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-11-12","value":12,"startTime":1762819200000,"endTime":1762905600000,"tweets":[{"bookmarked":false,"display_text_range":[0,48],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1988382853965770765","view_count":576,"bookmark_count":1,"created_at":1762902389000,"favorite_count":16,"quote_count":0,"reply_count":1,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1988382853965770765","full_text":"be the change you want to see in the world model","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,52],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[{"display_url":"youtube.com/watch?v=FcnLiP…","expanded_url":"https://www.youtube.com/watch?v=FcnLiPyfRZM","url":"https://t.co/jEBQGMsm62","indices":[29,52]}],"user_mentions":[]},"favorited":false,"lang":"en","possibly_sensitive":false,"possibly_sensitive_editable":true,"quoted_status_id_str":"1988165006073229659","quoted_status_permalink":{"url":"https://t.co/l99V6sHyI3","expanded":"https://twitter.com/MLStreetTalk/status/1988165006073229659","display":"x.com/MLStreetTalk/s…"},"retweeted":false,"fact_check":null,"id":"1988280687200530841","view_count":4997,"bookmark_count":5,"created_at":1762878030000,"favorite_count":21,"quote_count":0,"reply_count":1,"retweet_count":2,"user_id_str":"903008975343554566","conversation_id_str":"1988280687200530841","full_text":"This was a fun conversation: https://t.co/jEBQGMsm62","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,56],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[{"display_url":"jeremyberman.substack.com/p/how-i-got-th…","expanded_url":"https://jeremyberman.substack.com/p/how-i-got-the-highest-score-on-arc-agi-again","url":"https://t.co/xWiPB2Z0aI","indices":[33,56]}],"user_mentions":[]},"favorited":false,"in_reply_to_screen_name":"jerber888","lang":"en","possibly_sensitive":false,"possibly_sensitive_editable":true,"retweeted":false,"fact_check":null,"id":"1988278615377916103","view_count":1461,"bookmark_count":6,"created_at":1762877536000,"favorite_count":9,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"And technical blog post is here: https://t.co/xWiPB2Z0aI","in_reply_to_user_id_str":"903008975343554566","in_reply_to_status_id_str":"1987982214332580197","is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[9,52],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"536428919","name":"JFPuget 🇺🇦🇨🇦🇬🇱","screen_name":"JFPuget","indices":[0,8]}]},"favorited":false,"in_reply_to_screen_name":"JFPuget","lang":"en","retweeted":false,"fact_check":null,"id":"1988253927104471544","view_count":1040,"bookmark_count":0,"created_at":1762871650000,"favorite_count":8,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"@JFPuget Most likely but was focused on v1 for this.","in_reply_to_user_id_str":"536428919","in_reply_to_status_id_str":"1988249984907571356","is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[13,26],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"1194889317388374016","name":"Matt Shumer","screen_name":"mattshumer_","indices":[0,12]}]},"favorited":false,"in_reply_to_screen_name":"mattshumer_","lang":"en","retweeted":false,"fact_check":null,"id":"1988334412690571395","view_count":221,"bookmark_count":0,"created_at":1762890840000,"favorite_count":3,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1988331100935254226","full_text":"@mattshumer_ thanks Matt 🙏","in_reply_to_user_id_str":"1194889317388374016","in_reply_to_status_id_str":"1988331100935254226","is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-11-13","value":0,"startTime":1762905600000,"endTime":1762992000000,"tweets":[]},{"label":"2025-11-14","value":0,"startTime":1762992000000,"endTime":1763078400000,"tweets":[]},{"label":"2025-11-15","value":0,"startTime":1763078400000,"endTime":1763164800000,"tweets":[]}],"nretweets":[{"label":"2025-10-16","value":0,"startTime":1760486400000,"endTime":1760572800000,"tweets":[{"bookmarked":false,"display_text_range":[0,275],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1978585305679544427","view_count":90,"bookmark_count":0,"created_at":1760566471000,"favorite_count":4,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1978585305679544427","full_text":"Intelligence needs to be grown. Models need to develop coherent circuits that are reinforced. If you train models on traces they did not generate, there is no guarantee the information in that trace fits coherently in existing circuits. This is how models memorize instead of think for themselves. \n\nLearning should be on-policy. Take the unlikely but correct traces the model generates and reinforce those circuits. Then you build a model that thinks for itself.","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-17","value":0,"startTime":1760572800000,"endTime":1760659200000,"tweets":[{"bookmarked":false,"display_text_range":[73,86],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"500835474","name":"Pim de Witte","screen_name":"PimDeWitte","indices":[0,11]},{"id_str":"42226885","name":"Vinod Khosla","screen_name":"vkhosla","indices":[12,20]},{"id_str":"23809682","name":"Hemant Taneja","screen_name":"htaneja","indices":[21,29]},{"id_str":"1358786780259098625","name":"Eloi Alonso","screen_name":"EloiAlonso1","indices":[30,42]},{"id_str":"1307716628004704262","name":"Adam Jelley","screen_name":"AdamJelley2","indices":[43,55]},{"id_str":"1226783771195854848","name":"Vincent Micheli","screen_name":"micheli_vincent","indices":[56,72]}]},"favorited":false,"in_reply_to_screen_name":"PimDeWitte","lang":"en","retweeted":false,"fact_check":null,"id":"1978893031781568552","view_count":138,"bookmark_count":0,"created_at":1760639839000,"favorite_count":0,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1978825767053148457","full_text":"@PimDeWitte @vkhosla @htaneja @EloiAlonso1 @AdamJelley2 @micheli_vincent Congrats Pim!","in_reply_to_user_id_str":"500835474","in_reply_to_status_id_str":"1978825767053148457","is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-18","value":1,"startTime":1760659200000,"endTime":1760745600000,"tweets":[{"bookmarked":false,"display_text_range":[0,101],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","quoted_status_id_str":"1978196505392558087","quoted_status_permalink":{"url":"https://t.co/QhD1mEOFeS","expanded":"https://twitter.com/nicochristie/status/1978196505392558087","display":"x.com/nicochristie/s…"},"retweeted":false,"fact_check":null,"id":"1979230737405472922","view_count":1828,"bookmark_count":0,"created_at":1760720354000,"favorite_count":14,"quote_count":0,"reply_count":1,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1979230737405472922","full_text":"Accidentally said “impossible” instead of “out of distribution” and they kicked me out of my xAI tent","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,270],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","quoted_status_id_str":"1979237188220149938","quoted_status_permalink":{"url":"https://t.co/Ndv8TdVBSY","expanded":"https://twitter.com/MillionInt/status/1979237188220149938","display":"x.com/MillionInt/sta…"},"retweeted":false,"fact_check":null,"id":"1979240937281798378","view_count":730,"bookmark_count":5,"created_at":1760722786000,"favorite_count":12,"quote_count":0,"reply_count":2,"retweet_count":1,"user_id_str":"903008975343554566","conversation_id_str":"1979240937281798378","full_text":"Learning is like adding entropy to the neural network. Like adding a drop of colored liquid to clear water. The information diffuses across the weights. Unlearning that information is like undoing this diffusion. Reversing entropy. It takes a lot more energy to do that.","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null}]},{"label":"2025-10-19","value":0,"startTime":1760745600000,"endTime":1760832000000,"tweets":[]},{"label":"2025-10-20","value":0,"startTime":1760832000000,"endTime":1760918400000,"tweets":[]},{"label":"2025-10-21","value":0,"startTime":1760918400000,"endTime":1761004800000,"tweets":[{"bookmarked":false,"display_text_range":[0,110],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1980275836558213472","view_count":1446,"bookmark_count":2,"created_at":1760969525000,"favorite_count":20,"quote_count":0,"reply_count":1,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1980275836558213472","full_text":"To unlock the secrets of the universe, we just have to make interesting, true, and unlikely tokens more likely","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-22","value":0,"startTime":1761004800000,"endTime":1761091200000,"tweets":[]},{"label":"2025-10-23","value":0,"startTime":1761091200000,"endTime":1761177600000,"tweets":[]},{"label":"2025-10-24","value":0,"startTime":1761177600000,"endTime":1761264000000,"tweets":[]},{"label":"2025-10-25","value":0,"startTime":1761264000000,"endTime":1761350400000,"tweets":[{"bookmarked":false,"display_text_range":[17,70],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"4389158909","name":"Fernando Rojo","screen_name":"FernandoTheRojo","indices":[0,16]},{"id_str":"4389158909","name":"Fernando Rojo","screen_name":"FernandoTheRojo","indices":[17,33]}]},"favorited":false,"in_reply_to_screen_name":"FernandoTheRojo","lang":"en","retweeted":false,"fact_check":null,"id":"1981519914528821421","view_count":489,"bookmark_count":0,"created_at":1761266137000,"favorite_count":2,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1981509473295544350","full_text":"@FernandoTheRojo @FernandoTheRojo is the best app builder in the world","in_reply_to_user_id_str":"4389158909","in_reply_to_status_id_str":"1981509473295544350","is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-26","value":1,"startTime":1761350400000,"endTime":1761436800000,"tweets":[{"bookmarked":false,"display_text_range":[0,193],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1982191884593250452","view_count":2003,"bookmark_count":4,"created_at":1761426347000,"favorite_count":29,"quote_count":0,"reply_count":5,"retweet_count":1,"user_id_str":"903008975343554566","conversation_id_str":"1982191884593250452","full_text":"Solving hallucinations is a bigger deal than people think. Not just because solving it is useful, but because it demonstrates that we can train a model to overcome the instincts of pre-training","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,81],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1982190918326186171","view_count":685,"bookmark_count":0,"created_at":1761426116000,"favorite_count":1,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1982190918326186171","full_text":"A model using off-policy trained weights to respond is like a human under a spell","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-27","value":0,"startTime":1761436800000,"endTime":1761523200000,"tweets":[]},{"label":"2025-10-28","value":0,"startTime":1761523200000,"endTime":1761609600000,"tweets":[]},{"label":"2025-10-29","value":0,"startTime":1761609600000,"endTime":1761696000000,"tweets":[]},{"label":"2025-10-30","value":1,"startTime":1761696000000,"endTime":1761782400000,"tweets":[{"bookmarked":false,"display_text_range":[0,140],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1983562418782446066","view_count":1530,"bookmark_count":1,"created_at":1761753108000,"favorite_count":19,"quote_count":1,"reply_count":1,"retweet_count":1,"user_id_str":"903008975343554566","conversation_id_str":"1983562418782446066","full_text":"we use our continuous human brains to build symbolic computer systems to build continuous computer brains to build symbolic computer systems","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-31","value":0,"startTime":1761782400000,"endTime":1761868800000,"tweets":[]},{"label":"2025-11-01","value":0,"startTime":1761868800000,"endTime":1761955200000,"tweets":[]},{"label":"2025-11-02","value":0,"startTime":1761955200000,"endTime":1762041600000,"tweets":[{"bookmarked":false,"display_text_range":[0,44],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","quoted_status_id_str":"1984662921264120094","quoted_status_permalink":{"url":"https://t.co/s5M12jIb1I","expanded":"https://twitter.com/mindsai_jack/status/1984662921264120094","display":"x.com/mindsai_jack/s…"},"retweeted":false,"fact_check":null,"id":"1984665905972859177","view_count":2789,"bookmark_count":3,"created_at":1762016199000,"favorite_count":9,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1984665905972859177","full_text":"This was fun and Jack’s speech was great too","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null}]},{"label":"2025-11-03","value":0,"startTime":1762041600000,"endTime":1762128000000,"tweets":[]},{"label":"2025-11-04","value":0,"startTime":1762128000000,"endTime":1762214400000,"tweets":[]},{"label":"2025-11-05","value":1,"startTime":1762214400000,"endTime":1762300800000,"tweets":[{"bookmarked":false,"display_text_range":[0,275],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1985743643169178080","view_count":1463,"bookmark_count":4,"created_at":1762273152000,"favorite_count":13,"quote_count":0,"reply_count":3,"retweet_count":1,"user_id_str":"903008975343554566","conversation_id_str":"1985743643169178080","full_text":"Teaching language models to have taste isn't just about making them better at writing or making jokes.\n\nTaste is everything. It's what leads to scientific discovery.\n\nIn an infinite sea of things to reason about, taste is what guides you to reason about the right things — to combine the right ideas together to deduce something interesting and new.\n\nThe discovery of special relativity was a product of good taste.","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-11-06","value":0,"startTime":1762300800000,"endTime":1762387200000,"tweets":[]},{"label":"2025-11-07","value":0,"startTime":1762387200000,"endTime":1762473600000,"tweets":[{"bookmarked":false,"display_text_range":[0,96],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","quoted_status_id_str":"1986461871462096920","quoted_status_permalink":{"url":"https://t.co/J1VFIbBDHn","expanded":"https://twitter.com/skiminok/status/1986461871462096920","display":"x.com/skiminok/statu…"},"retweeted":false,"fact_check":null,"id":"1986474907484783018","view_count":1347,"bookmark_count":0,"created_at":1762447499000,"favorite_count":6,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1986474907484783018","full_text":"Never a dull moment talking language models with Alex. Really excited to get to do it more often","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null}]},{"label":"2025-11-08","value":0,"startTime":1762473600000,"endTime":1762560000000,"tweets":[]},{"label":"2025-11-09","value":0,"startTime":1762560000000,"endTime":1762646400000,"tweets":[]},{"label":"2025-11-10","value":0,"startTime":1762646400000,"endTime":1762732800000,"tweets":[]},{"label":"2025-11-11","value":151,"startTime":1762732800000,"endTime":1762819200000,"tweets":[{"bookmarked":false,"display_text_range":[0,224],"entities":{"hashtags":[],"media":[{"display_url":"pic.x.com/CGkAyp0vMN","expanded_url":"https://x.com/jerber888/status/1987982067116777521/photo/1","id_str":"1987981765265272832","indices":[225,248],"media_key":"3_1987981765265272832","media_url_https":"https://pbs.twimg.com/media/G5a64fqXMAAL8FC.jpg","type":"photo","url":"https://t.co/CGkAyp0vMN","ext_media_availability":{"status":"Available"},"features":{"large":{"faces":[]},"medium":{"faces":[]},"small":{"faces":[]},"orig":{"faces":[]}},"sizes":{"large":{"h":1154,"w":1956,"resize":"fit"},"medium":{"h":708,"w":1200,"resize":"fit"},"small":{"h":401,"w":680,"resize":"fit"},"thumb":{"h":150,"w":150,"resize":"crop"}},"original_info":{"height":1154,"width":1956,"focus_rects":[{"x":0,"y":0,"w":1956,"h":1095},{"x":802,"y":0,"w":1154,"h":1154},{"x":944,"y":0,"w":1012,"h":1154},{"x":1323,"y":0,"w":577,"h":1154},{"x":0,"y":0,"w":1956,"h":1154}]},"media_results":{"result":{"media_key":"3_1987981765265272832"}}}],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"extended_entities":{"media":[{"display_url":"pic.x.com/CGkAyp0vMN","expanded_url":"https://x.com/jerber888/status/1987982067116777521/photo/1","id_str":"1987981765265272832","indices":[225,248],"media_key":"3_1987981765265272832","media_url_https":"https://pbs.twimg.com/media/G5a64fqXMAAL8FC.jpg","type":"photo","url":"https://t.co/CGkAyp0vMN","ext_media_availability":{"status":"Available"},"features":{"large":{"faces":[]},"medium":{"faces":[]},"small":{"faces":[]},"orig":{"faces":[]}},"sizes":{"large":{"h":1154,"w":1956,"resize":"fit"},"medium":{"h":708,"w":1200,"resize":"fit"},"small":{"h":401,"w":680,"resize":"fit"},"thumb":{"h":150,"w":150,"resize":"crop"}},"original_info":{"height":1154,"width":1956,"focus_rects":[{"x":0,"y":0,"w":1956,"h":1095},{"x":802,"y":0,"w":1154,"h":1154},{"x":944,"y":0,"w":1012,"h":1154},{"x":1323,"y":0,"w":577,"h":1154},{"x":0,"y":0,"w":1956,"h":1154}]},"media_results":{"result":{"media_key":"3_1987981765265272832"}}}]},"favorited":false,"lang":"en","possibly_sensitive":false,"possibly_sensitive_editable":true,"quoted_status_id_str":"1968001933211471891","quoted_status_permalink":{"url":"https://t.co/l4P9D5a1TN","expanded":"https://twitter.com/jerber888/status/1968001933211471891","display":"x.com/jerber888/stat…"},"retweeted":false,"fact_check":null,"id":"1987982067116777521","view_count":269951,"bookmark_count":591,"created_at":1762806834000,"favorite_count":1579,"quote_count":29,"reply_count":70,"retweet_count":147,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"I finally reached human-level performance (85%) on ARC-AGI v1 for under $10k and within 12 hours. I use the same multi-agent collaboration with evolutionary test-time compute, now powered by GPT-5 pro with lower parallelism. https://t.co/CGkAyp0vMN","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,275],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"1773935160192647168","name":"ARC Prize","screen_name":"arcprize","indices":[168,177]},{"id_str":"1773935160192647168","name":"ARC Prize","screen_name":"arcprize","indices":[168,177]}]},"favorited":false,"in_reply_to_screen_name":"jerber888","lang":"en","retweeted":false,"fact_check":null,"id":"1987982069712945649","view_count":6572,"bookmark_count":6,"created_at":1762806834000,"favorite_count":57,"quote_count":0,"reply_count":4,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"I ran it on a random sample of 100 tasks from the 2024 eval set. It got 88/100 and averaged $27 per task. The score still needs to be verified on the hidden set by the @arcprize, but in my past submissions, the hidden set subtracts a few percent from the score and adds a few dollars. I put a box where I expect the official score to land within.","in_reply_to_user_id_str":"903008975343554566","in_reply_to_status_id_str":"1987982067116777521","is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,279],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"825088493764407298","name":"Noam Brown","screen_name":"polynoamial","indices":[288,300]}]},"favorited":false,"in_reply_to_screen_name":"jerber888","lang":"en","retweeted":false,"fact_check":null,"id":"1987982071541993700","view_count":6923,"bookmark_count":7,"created_at":1762806835000,"favorite_count":75,"quote_count":0,"reply_count":2,"retweet_count":2,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"GPT-5 pro is the best reasoning model today. It thinks coherently for hours and hours. My agent coordination logic is likely an intermediate step before the models learn to do this type of long horizon coordination on their own, end to end. It’s a hard problem but I wouldn’t bet against @polynoamial and team.","in_reply_to_user_id_str":"903008975343554566","in_reply_to_status_id_str":"1987982069712945649","is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,123],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[{"display_url":"github.com/jerber/arc-lan…","expanded_url":"https://github.com/jerber/arc-lang-public","url":"https://t.co/QnGjsN98ji","indices":[50,73]},{"display_url":"kaggle.com/code/jerber/je…","expanded_url":"https://www.kaggle.com/code/jerber/jeremy-arc2","url":"https://t.co/1jhzROG4SL","indices":[100,123]}],"user_mentions":[]},"favorited":false,"in_reply_to_screen_name":"jerber888","lang":"en","possibly_sensitive":false,"possibly_sensitive_editable":true,"retweeted":false,"fact_check":null,"id":"1987982214332580197","view_count":6525,"bookmark_count":39,"created_at":1762806869000,"favorite_count":64,"quote_count":1,"reply_count":2,"retweet_count":2,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"You can run this code yourself, it’s open source: https://t.co/QnGjsN98ji. And the kaggle notebook: https://t.co/1jhzROG4SL","in_reply_to_user_id_str":"903008975343554566","in_reply_to_status_id_str":"1987982071541993700","is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-11-12","value":2,"startTime":1762819200000,"endTime":1762905600000,"tweets":[{"bookmarked":false,"display_text_range":[0,48],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1988382853965770765","view_count":576,"bookmark_count":1,"created_at":1762902389000,"favorite_count":16,"quote_count":0,"reply_count":1,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1988382853965770765","full_text":"be the change you want to see in the world model","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,52],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[{"display_url":"youtube.com/watch?v=FcnLiP…","expanded_url":"https://www.youtube.com/watch?v=FcnLiPyfRZM","url":"https://t.co/jEBQGMsm62","indices":[29,52]}],"user_mentions":[]},"favorited":false,"lang":"en","possibly_sensitive":false,"possibly_sensitive_editable":true,"quoted_status_id_str":"1988165006073229659","quoted_status_permalink":{"url":"https://t.co/l99V6sHyI3","expanded":"https://twitter.com/MLStreetTalk/status/1988165006073229659","display":"x.com/MLStreetTalk/s…"},"retweeted":false,"fact_check":null,"id":"1988280687200530841","view_count":4997,"bookmark_count":5,"created_at":1762878030000,"favorite_count":21,"quote_count":0,"reply_count":1,"retweet_count":2,"user_id_str":"903008975343554566","conversation_id_str":"1988280687200530841","full_text":"This was a fun conversation: https://t.co/jEBQGMsm62","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,56],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[{"display_url":"jeremyberman.substack.com/p/how-i-got-th…","expanded_url":"https://jeremyberman.substack.com/p/how-i-got-the-highest-score-on-arc-agi-again","url":"https://t.co/xWiPB2Z0aI","indices":[33,56]}],"user_mentions":[]},"favorited":false,"in_reply_to_screen_name":"jerber888","lang":"en","possibly_sensitive":false,"possibly_sensitive_editable":true,"retweeted":false,"fact_check":null,"id":"1988278615377916103","view_count":1461,"bookmark_count":6,"created_at":1762877536000,"favorite_count":9,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"And technical blog post is here: https://t.co/xWiPB2Z0aI","in_reply_to_user_id_str":"903008975343554566","in_reply_to_status_id_str":"1987982214332580197","is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[9,52],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"536428919","name":"JFPuget 🇺🇦🇨🇦🇬🇱","screen_name":"JFPuget","indices":[0,8]}]},"favorited":false,"in_reply_to_screen_name":"JFPuget","lang":"en","retweeted":false,"fact_check":null,"id":"1988253927104471544","view_count":1040,"bookmark_count":0,"created_at":1762871650000,"favorite_count":8,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"@JFPuget Most likely but was focused on v1 for this.","in_reply_to_user_id_str":"536428919","in_reply_to_status_id_str":"1988249984907571356","is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[13,26],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"1194889317388374016","name":"Matt Shumer","screen_name":"mattshumer_","indices":[0,12]}]},"favorited":false,"in_reply_to_screen_name":"mattshumer_","lang":"en","retweeted":false,"fact_check":null,"id":"1988334412690571395","view_count":221,"bookmark_count":0,"created_at":1762890840000,"favorite_count":3,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1988331100935254226","full_text":"@mattshumer_ thanks Matt 🙏","in_reply_to_user_id_str":"1194889317388374016","in_reply_to_status_id_str":"1988331100935254226","is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-11-13","value":0,"startTime":1762905600000,"endTime":1762992000000,"tweets":[]},{"label":"2025-11-14","value":0,"startTime":1762992000000,"endTime":1763078400000,"tweets":[]},{"label":"2025-11-15","value":0,"startTime":1763078400000,"endTime":1763164800000,"tweets":[]}],"nlikes":[{"label":"2025-10-16","value":4,"startTime":1760486400000,"endTime":1760572800000,"tweets":[{"bookmarked":false,"display_text_range":[0,275],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1978585305679544427","view_count":90,"bookmark_count":0,"created_at":1760566471000,"favorite_count":4,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1978585305679544427","full_text":"Intelligence needs to be grown. Models need to develop coherent circuits that are reinforced. If you train models on traces they did not generate, there is no guarantee the information in that trace fits coherently in existing circuits. This is how models memorize instead of think for themselves. \n\nLearning should be on-policy. Take the unlikely but correct traces the model generates and reinforce those circuits. Then you build a model that thinks for itself.","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-17","value":0,"startTime":1760572800000,"endTime":1760659200000,"tweets":[{"bookmarked":false,"display_text_range":[73,86],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"500835474","name":"Pim de Witte","screen_name":"PimDeWitte","indices":[0,11]},{"id_str":"42226885","name":"Vinod Khosla","screen_name":"vkhosla","indices":[12,20]},{"id_str":"23809682","name":"Hemant Taneja","screen_name":"htaneja","indices":[21,29]},{"id_str":"1358786780259098625","name":"Eloi Alonso","screen_name":"EloiAlonso1","indices":[30,42]},{"id_str":"1307716628004704262","name":"Adam Jelley","screen_name":"AdamJelley2","indices":[43,55]},{"id_str":"1226783771195854848","name":"Vincent Micheli","screen_name":"micheli_vincent","indices":[56,72]}]},"favorited":false,"in_reply_to_screen_name":"PimDeWitte","lang":"en","retweeted":false,"fact_check":null,"id":"1978893031781568552","view_count":138,"bookmark_count":0,"created_at":1760639839000,"favorite_count":0,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1978825767053148457","full_text":"@PimDeWitte @vkhosla @htaneja @EloiAlonso1 @AdamJelley2 @micheli_vincent Congrats Pim!","in_reply_to_user_id_str":"500835474","in_reply_to_status_id_str":"1978825767053148457","is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-18","value":26,"startTime":1760659200000,"endTime":1760745600000,"tweets":[{"bookmarked":false,"display_text_range":[0,101],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","quoted_status_id_str":"1978196505392558087","quoted_status_permalink":{"url":"https://t.co/QhD1mEOFeS","expanded":"https://twitter.com/nicochristie/status/1978196505392558087","display":"x.com/nicochristie/s…"},"retweeted":false,"fact_check":null,"id":"1979230737405472922","view_count":1828,"bookmark_count":0,"created_at":1760720354000,"favorite_count":14,"quote_count":0,"reply_count":1,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1979230737405472922","full_text":"Accidentally said “impossible” instead of “out of distribution” and they kicked me out of my xAI tent","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,270],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","quoted_status_id_str":"1979237188220149938","quoted_status_permalink":{"url":"https://t.co/Ndv8TdVBSY","expanded":"https://twitter.com/MillionInt/status/1979237188220149938","display":"x.com/MillionInt/sta…"},"retweeted":false,"fact_check":null,"id":"1979240937281798378","view_count":730,"bookmark_count":5,"created_at":1760722786000,"favorite_count":12,"quote_count":0,"reply_count":2,"retweet_count":1,"user_id_str":"903008975343554566","conversation_id_str":"1979240937281798378","full_text":"Learning is like adding entropy to the neural network. Like adding a drop of colored liquid to clear water. The information diffuses across the weights. Unlearning that information is like undoing this diffusion. Reversing entropy. It takes a lot more energy to do that.","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null}]},{"label":"2025-10-19","value":0,"startTime":1760745600000,"endTime":1760832000000,"tweets":[]},{"label":"2025-10-20","value":0,"startTime":1760832000000,"endTime":1760918400000,"tweets":[]},{"label":"2025-10-21","value":20,"startTime":1760918400000,"endTime":1761004800000,"tweets":[{"bookmarked":false,"display_text_range":[0,110],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1980275836558213472","view_count":1446,"bookmark_count":2,"created_at":1760969525000,"favorite_count":20,"quote_count":0,"reply_count":1,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1980275836558213472","full_text":"To unlock the secrets of the universe, we just have to make interesting, true, and unlikely tokens more likely","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-22","value":0,"startTime":1761004800000,"endTime":1761091200000,"tweets":[]},{"label":"2025-10-23","value":0,"startTime":1761091200000,"endTime":1761177600000,"tweets":[]},{"label":"2025-10-24","value":0,"startTime":1761177600000,"endTime":1761264000000,"tweets":[]},{"label":"2025-10-25","value":2,"startTime":1761264000000,"endTime":1761350400000,"tweets":[{"bookmarked":false,"display_text_range":[17,70],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"4389158909","name":"Fernando Rojo","screen_name":"FernandoTheRojo","indices":[0,16]},{"id_str":"4389158909","name":"Fernando Rojo","screen_name":"FernandoTheRojo","indices":[17,33]}]},"favorited":false,"in_reply_to_screen_name":"FernandoTheRojo","lang":"en","retweeted":false,"fact_check":null,"id":"1981519914528821421","view_count":489,"bookmark_count":0,"created_at":1761266137000,"favorite_count":2,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1981509473295544350","full_text":"@FernandoTheRojo @FernandoTheRojo is the best app builder in the world","in_reply_to_user_id_str":"4389158909","in_reply_to_status_id_str":"1981509473295544350","is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-26","value":30,"startTime":1761350400000,"endTime":1761436800000,"tweets":[{"bookmarked":false,"display_text_range":[0,193],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1982191884593250452","view_count":2003,"bookmark_count":4,"created_at":1761426347000,"favorite_count":29,"quote_count":0,"reply_count":5,"retweet_count":1,"user_id_str":"903008975343554566","conversation_id_str":"1982191884593250452","full_text":"Solving hallucinations is a bigger deal than people think. Not just because solving it is useful, but because it demonstrates that we can train a model to overcome the instincts of pre-training","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,81],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1982190918326186171","view_count":685,"bookmark_count":0,"created_at":1761426116000,"favorite_count":1,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1982190918326186171","full_text":"A model using off-policy trained weights to respond is like a human under a spell","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-27","value":0,"startTime":1761436800000,"endTime":1761523200000,"tweets":[]},{"label":"2025-10-28","value":0,"startTime":1761523200000,"endTime":1761609600000,"tweets":[]},{"label":"2025-10-29","value":0,"startTime":1761609600000,"endTime":1761696000000,"tweets":[]},{"label":"2025-10-30","value":19,"startTime":1761696000000,"endTime":1761782400000,"tweets":[{"bookmarked":false,"display_text_range":[0,140],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1983562418782446066","view_count":1530,"bookmark_count":1,"created_at":1761753108000,"favorite_count":19,"quote_count":1,"reply_count":1,"retweet_count":1,"user_id_str":"903008975343554566","conversation_id_str":"1983562418782446066","full_text":"we use our continuous human brains to build symbolic computer systems to build continuous computer brains to build symbolic computer systems","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-31","value":0,"startTime":1761782400000,"endTime":1761868800000,"tweets":[]},{"label":"2025-11-01","value":0,"startTime":1761868800000,"endTime":1761955200000,"tweets":[]},{"label":"2025-11-02","value":9,"startTime":1761955200000,"endTime":1762041600000,"tweets":[{"bookmarked":false,"display_text_range":[0,44],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","quoted_status_id_str":"1984662921264120094","quoted_status_permalink":{"url":"https://t.co/s5M12jIb1I","expanded":"https://twitter.com/mindsai_jack/status/1984662921264120094","display":"x.com/mindsai_jack/s…"},"retweeted":false,"fact_check":null,"id":"1984665905972859177","view_count":2789,"bookmark_count":3,"created_at":1762016199000,"favorite_count":9,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1984665905972859177","full_text":"This was fun and Jack’s speech was great too","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null}]},{"label":"2025-11-03","value":0,"startTime":1762041600000,"endTime":1762128000000,"tweets":[]},{"label":"2025-11-04","value":0,"startTime":1762128000000,"endTime":1762214400000,"tweets":[]},{"label":"2025-11-05","value":13,"startTime":1762214400000,"endTime":1762300800000,"tweets":[{"bookmarked":false,"display_text_range":[0,275],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1985743643169178080","view_count":1463,"bookmark_count":4,"created_at":1762273152000,"favorite_count":13,"quote_count":0,"reply_count":3,"retweet_count":1,"user_id_str":"903008975343554566","conversation_id_str":"1985743643169178080","full_text":"Teaching language models to have taste isn't just about making them better at writing or making jokes.\n\nTaste is everything. It's what leads to scientific discovery.\n\nIn an infinite sea of things to reason about, taste is what guides you to reason about the right things — to combine the right ideas together to deduce something interesting and new.\n\nThe discovery of special relativity was a product of good taste.","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-11-06","value":0,"startTime":1762300800000,"endTime":1762387200000,"tweets":[]},{"label":"2025-11-07","value":6,"startTime":1762387200000,"endTime":1762473600000,"tweets":[{"bookmarked":false,"display_text_range":[0,96],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","quoted_status_id_str":"1986461871462096920","quoted_status_permalink":{"url":"https://t.co/J1VFIbBDHn","expanded":"https://twitter.com/skiminok/status/1986461871462096920","display":"x.com/skiminok/statu…"},"retweeted":false,"fact_check":null,"id":"1986474907484783018","view_count":1347,"bookmark_count":0,"created_at":1762447499000,"favorite_count":6,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1986474907484783018","full_text":"Never a dull moment talking language models with Alex. Really excited to get to do it more often","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null}]},{"label":"2025-11-08","value":0,"startTime":1762473600000,"endTime":1762560000000,"tweets":[]},{"label":"2025-11-09","value":0,"startTime":1762560000000,"endTime":1762646400000,"tweets":[]},{"label":"2025-11-10","value":0,"startTime":1762646400000,"endTime":1762732800000,"tweets":[]},{"label":"2025-11-11","value":1775,"startTime":1762732800000,"endTime":1762819200000,"tweets":[{"bookmarked":false,"display_text_range":[0,224],"entities":{"hashtags":[],"media":[{"display_url":"pic.x.com/CGkAyp0vMN","expanded_url":"https://x.com/jerber888/status/1987982067116777521/photo/1","id_str":"1987981765265272832","indices":[225,248],"media_key":"3_1987981765265272832","media_url_https":"https://pbs.twimg.com/media/G5a64fqXMAAL8FC.jpg","type":"photo","url":"https://t.co/CGkAyp0vMN","ext_media_availability":{"status":"Available"},"features":{"large":{"faces":[]},"medium":{"faces":[]},"small":{"faces":[]},"orig":{"faces":[]}},"sizes":{"large":{"h":1154,"w":1956,"resize":"fit"},"medium":{"h":708,"w":1200,"resize":"fit"},"small":{"h":401,"w":680,"resize":"fit"},"thumb":{"h":150,"w":150,"resize":"crop"}},"original_info":{"height":1154,"width":1956,"focus_rects":[{"x":0,"y":0,"w":1956,"h":1095},{"x":802,"y":0,"w":1154,"h":1154},{"x":944,"y":0,"w":1012,"h":1154},{"x":1323,"y":0,"w":577,"h":1154},{"x":0,"y":0,"w":1956,"h":1154}]},"media_results":{"result":{"media_key":"3_1987981765265272832"}}}],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"extended_entities":{"media":[{"display_url":"pic.x.com/CGkAyp0vMN","expanded_url":"https://x.com/jerber888/status/1987982067116777521/photo/1","id_str":"1987981765265272832","indices":[225,248],"media_key":"3_1987981765265272832","media_url_https":"https://pbs.twimg.com/media/G5a64fqXMAAL8FC.jpg","type":"photo","url":"https://t.co/CGkAyp0vMN","ext_media_availability":{"status":"Available"},"features":{"large":{"faces":[]},"medium":{"faces":[]},"small":{"faces":[]},"orig":{"faces":[]}},"sizes":{"large":{"h":1154,"w":1956,"resize":"fit"},"medium":{"h":708,"w":1200,"resize":"fit"},"small":{"h":401,"w":680,"resize":"fit"},"thumb":{"h":150,"w":150,"resize":"crop"}},"original_info":{"height":1154,"width":1956,"focus_rects":[{"x":0,"y":0,"w":1956,"h":1095},{"x":802,"y":0,"w":1154,"h":1154},{"x":944,"y":0,"w":1012,"h":1154},{"x":1323,"y":0,"w":577,"h":1154},{"x":0,"y":0,"w":1956,"h":1154}]},"media_results":{"result":{"media_key":"3_1987981765265272832"}}}]},"favorited":false,"lang":"en","possibly_sensitive":false,"possibly_sensitive_editable":true,"quoted_status_id_str":"1968001933211471891","quoted_status_permalink":{"url":"https://t.co/l4P9D5a1TN","expanded":"https://twitter.com/jerber888/status/1968001933211471891","display":"x.com/jerber888/stat…"},"retweeted":false,"fact_check":null,"id":"1987982067116777521","view_count":269951,"bookmark_count":591,"created_at":1762806834000,"favorite_count":1579,"quote_count":29,"reply_count":70,"retweet_count":147,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"I finally reached human-level performance (85%) on ARC-AGI v1 for under $10k and within 12 hours. I use the same multi-agent collaboration with evolutionary test-time compute, now powered by GPT-5 pro with lower parallelism. https://t.co/CGkAyp0vMN","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,275],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"1773935160192647168","name":"ARC Prize","screen_name":"arcprize","indices":[168,177]},{"id_str":"1773935160192647168","name":"ARC Prize","screen_name":"arcprize","indices":[168,177]}]},"favorited":false,"in_reply_to_screen_name":"jerber888","lang":"en","retweeted":false,"fact_check":null,"id":"1987982069712945649","view_count":6572,"bookmark_count":6,"created_at":1762806834000,"favorite_count":57,"quote_count":0,"reply_count":4,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"I ran it on a random sample of 100 tasks from the 2024 eval set. It got 88/100 and averaged $27 per task. The score still needs to be verified on the hidden set by the @arcprize, but in my past submissions, the hidden set subtracts a few percent from the score and adds a few dollars. I put a box where I expect the official score to land within.","in_reply_to_user_id_str":"903008975343554566","in_reply_to_status_id_str":"1987982067116777521","is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,279],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"825088493764407298","name":"Noam Brown","screen_name":"polynoamial","indices":[288,300]}]},"favorited":false,"in_reply_to_screen_name":"jerber888","lang":"en","retweeted":false,"fact_check":null,"id":"1987982071541993700","view_count":6923,"bookmark_count":7,"created_at":1762806835000,"favorite_count":75,"quote_count":0,"reply_count":2,"retweet_count":2,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"GPT-5 pro is the best reasoning model today. It thinks coherently for hours and hours. My agent coordination logic is likely an intermediate step before the models learn to do this type of long horizon coordination on their own, end to end. It’s a hard problem but I wouldn’t bet against @polynoamial and team.","in_reply_to_user_id_str":"903008975343554566","in_reply_to_status_id_str":"1987982069712945649","is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,123],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[{"display_url":"github.com/jerber/arc-lan…","expanded_url":"https://github.com/jerber/arc-lang-public","url":"https://t.co/QnGjsN98ji","indices":[50,73]},{"display_url":"kaggle.com/code/jerber/je…","expanded_url":"https://www.kaggle.com/code/jerber/jeremy-arc2","url":"https://t.co/1jhzROG4SL","indices":[100,123]}],"user_mentions":[]},"favorited":false,"in_reply_to_screen_name":"jerber888","lang":"en","possibly_sensitive":false,"possibly_sensitive_editable":true,"retweeted":false,"fact_check":null,"id":"1987982214332580197","view_count":6525,"bookmark_count":39,"created_at":1762806869000,"favorite_count":64,"quote_count":1,"reply_count":2,"retweet_count":2,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"You can run this code yourself, it’s open source: https://t.co/QnGjsN98ji. And the kaggle notebook: https://t.co/1jhzROG4SL","in_reply_to_user_id_str":"903008975343554566","in_reply_to_status_id_str":"1987982071541993700","is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-11-12","value":57,"startTime":1762819200000,"endTime":1762905600000,"tweets":[{"bookmarked":false,"display_text_range":[0,48],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1988382853965770765","view_count":576,"bookmark_count":1,"created_at":1762902389000,"favorite_count":16,"quote_count":0,"reply_count":1,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1988382853965770765","full_text":"be the change you want to see in the world model","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,52],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[{"display_url":"youtube.com/watch?v=FcnLiP…","expanded_url":"https://www.youtube.com/watch?v=FcnLiPyfRZM","url":"https://t.co/jEBQGMsm62","indices":[29,52]}],"user_mentions":[]},"favorited":false,"lang":"en","possibly_sensitive":false,"possibly_sensitive_editable":true,"quoted_status_id_str":"1988165006073229659","quoted_status_permalink":{"url":"https://t.co/l99V6sHyI3","expanded":"https://twitter.com/MLStreetTalk/status/1988165006073229659","display":"x.com/MLStreetTalk/s…"},"retweeted":false,"fact_check":null,"id":"1988280687200530841","view_count":4997,"bookmark_count":5,"created_at":1762878030000,"favorite_count":21,"quote_count":0,"reply_count":1,"retweet_count":2,"user_id_str":"903008975343554566","conversation_id_str":"1988280687200530841","full_text":"This was a fun conversation: https://t.co/jEBQGMsm62","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,56],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[{"display_url":"jeremyberman.substack.com/p/how-i-got-th…","expanded_url":"https://jeremyberman.substack.com/p/how-i-got-the-highest-score-on-arc-agi-again","url":"https://t.co/xWiPB2Z0aI","indices":[33,56]}],"user_mentions":[]},"favorited":false,"in_reply_to_screen_name":"jerber888","lang":"en","possibly_sensitive":false,"possibly_sensitive_editable":true,"retweeted":false,"fact_check":null,"id":"1988278615377916103","view_count":1461,"bookmark_count":6,"created_at":1762877536000,"favorite_count":9,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"And technical blog post is here: https://t.co/xWiPB2Z0aI","in_reply_to_user_id_str":"903008975343554566","in_reply_to_status_id_str":"1987982214332580197","is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[9,52],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"536428919","name":"JFPuget 🇺🇦🇨🇦🇬🇱","screen_name":"JFPuget","indices":[0,8]}]},"favorited":false,"in_reply_to_screen_name":"JFPuget","lang":"en","retweeted":false,"fact_check":null,"id":"1988253927104471544","view_count":1040,"bookmark_count":0,"created_at":1762871650000,"favorite_count":8,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"@JFPuget Most likely but was focused on v1 for this.","in_reply_to_user_id_str":"536428919","in_reply_to_status_id_str":"1988249984907571356","is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[13,26],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"1194889317388374016","name":"Matt Shumer","screen_name":"mattshumer_","indices":[0,12]}]},"favorited":false,"in_reply_to_screen_name":"mattshumer_","lang":"en","retweeted":false,"fact_check":null,"id":"1988334412690571395","view_count":221,"bookmark_count":0,"created_at":1762890840000,"favorite_count":3,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1988331100935254226","full_text":"@mattshumer_ thanks Matt 🙏","in_reply_to_user_id_str":"1194889317388374016","in_reply_to_status_id_str":"1988331100935254226","is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-11-13","value":0,"startTime":1762905600000,"endTime":1762992000000,"tweets":[]},{"label":"2025-11-14","value":0,"startTime":1762992000000,"endTime":1763078400000,"tweets":[]},{"label":"2025-11-15","value":0,"startTime":1763078400000,"endTime":1763164800000,"tweets":[]}],"nviews":[{"label":"2025-10-16","value":90,"startTime":1760486400000,"endTime":1760572800000,"tweets":[{"bookmarked":false,"display_text_range":[0,275],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1978585305679544427","view_count":90,"bookmark_count":0,"created_at":1760566471000,"favorite_count":4,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1978585305679544427","full_text":"Intelligence needs to be grown. Models need to develop coherent circuits that are reinforced. If you train models on traces they did not generate, there is no guarantee the information in that trace fits coherently in existing circuits. This is how models memorize instead of think for themselves. \n\nLearning should be on-policy. Take the unlikely but correct traces the model generates and reinforce those circuits. Then you build a model that thinks for itself.","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-17","value":138,"startTime":1760572800000,"endTime":1760659200000,"tweets":[{"bookmarked":false,"display_text_range":[73,86],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"500835474","name":"Pim de Witte","screen_name":"PimDeWitte","indices":[0,11]},{"id_str":"42226885","name":"Vinod Khosla","screen_name":"vkhosla","indices":[12,20]},{"id_str":"23809682","name":"Hemant Taneja","screen_name":"htaneja","indices":[21,29]},{"id_str":"1358786780259098625","name":"Eloi Alonso","screen_name":"EloiAlonso1","indices":[30,42]},{"id_str":"1307716628004704262","name":"Adam Jelley","screen_name":"AdamJelley2","indices":[43,55]},{"id_str":"1226783771195854848","name":"Vincent Micheli","screen_name":"micheli_vincent","indices":[56,72]}]},"favorited":false,"in_reply_to_screen_name":"PimDeWitte","lang":"en","retweeted":false,"fact_check":null,"id":"1978893031781568552","view_count":138,"bookmark_count":0,"created_at":1760639839000,"favorite_count":0,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1978825767053148457","full_text":"@PimDeWitte @vkhosla @htaneja @EloiAlonso1 @AdamJelley2 @micheli_vincent Congrats Pim!","in_reply_to_user_id_str":"500835474","in_reply_to_status_id_str":"1978825767053148457","is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-18","value":2558,"startTime":1760659200000,"endTime":1760745600000,"tweets":[{"bookmarked":false,"display_text_range":[0,101],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","quoted_status_id_str":"1978196505392558087","quoted_status_permalink":{"url":"https://t.co/QhD1mEOFeS","expanded":"https://twitter.com/nicochristie/status/1978196505392558087","display":"x.com/nicochristie/s…"},"retweeted":false,"fact_check":null,"id":"1979230737405472922","view_count":1828,"bookmark_count":0,"created_at":1760720354000,"favorite_count":14,"quote_count":0,"reply_count":1,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1979230737405472922","full_text":"Accidentally said “impossible” instead of “out of distribution” and they kicked me out of my xAI tent","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,270],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","quoted_status_id_str":"1979237188220149938","quoted_status_permalink":{"url":"https://t.co/Ndv8TdVBSY","expanded":"https://twitter.com/MillionInt/status/1979237188220149938","display":"x.com/MillionInt/sta…"},"retweeted":false,"fact_check":null,"id":"1979240937281798378","view_count":730,"bookmark_count":5,"created_at":1760722786000,"favorite_count":12,"quote_count":0,"reply_count":2,"retweet_count":1,"user_id_str":"903008975343554566","conversation_id_str":"1979240937281798378","full_text":"Learning is like adding entropy to the neural network. Like adding a drop of colored liquid to clear water. The information diffuses across the weights. Unlearning that information is like undoing this diffusion. Reversing entropy. It takes a lot more energy to do that.","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null}]},{"label":"2025-10-19","value":0,"startTime":1760745600000,"endTime":1760832000000,"tweets":[]},{"label":"2025-10-20","value":0,"startTime":1760832000000,"endTime":1760918400000,"tweets":[]},{"label":"2025-10-21","value":1446,"startTime":1760918400000,"endTime":1761004800000,"tweets":[{"bookmarked":false,"display_text_range":[0,110],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1980275836558213472","view_count":1446,"bookmark_count":2,"created_at":1760969525000,"favorite_count":20,"quote_count":0,"reply_count":1,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1980275836558213472","full_text":"To unlock the secrets of the universe, we just have to make interesting, true, and unlikely tokens more likely","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-22","value":0,"startTime":1761004800000,"endTime":1761091200000,"tweets":[]},{"label":"2025-10-23","value":0,"startTime":1761091200000,"endTime":1761177600000,"tweets":[]},{"label":"2025-10-24","value":0,"startTime":1761177600000,"endTime":1761264000000,"tweets":[]},{"label":"2025-10-25","value":489,"startTime":1761264000000,"endTime":1761350400000,"tweets":[{"bookmarked":false,"display_text_range":[17,70],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"4389158909","name":"Fernando Rojo","screen_name":"FernandoTheRojo","indices":[0,16]},{"id_str":"4389158909","name":"Fernando Rojo","screen_name":"FernandoTheRojo","indices":[17,33]}]},"favorited":false,"in_reply_to_screen_name":"FernandoTheRojo","lang":"en","retweeted":false,"fact_check":null,"id":"1981519914528821421","view_count":489,"bookmark_count":0,"created_at":1761266137000,"favorite_count":2,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1981509473295544350","full_text":"@FernandoTheRojo @FernandoTheRojo is the best app builder in the world","in_reply_to_user_id_str":"4389158909","in_reply_to_status_id_str":"1981509473295544350","is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-26","value":2688,"startTime":1761350400000,"endTime":1761436800000,"tweets":[{"bookmarked":false,"display_text_range":[0,193],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1982191884593250452","view_count":2003,"bookmark_count":4,"created_at":1761426347000,"favorite_count":29,"quote_count":0,"reply_count":5,"retweet_count":1,"user_id_str":"903008975343554566","conversation_id_str":"1982191884593250452","full_text":"Solving hallucinations is a bigger deal than people think. Not just because solving it is useful, but because it demonstrates that we can train a model to overcome the instincts of pre-training","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,81],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1982190918326186171","view_count":685,"bookmark_count":0,"created_at":1761426116000,"favorite_count":1,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1982190918326186171","full_text":"A model using off-policy trained weights to respond is like a human under a spell","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-27","value":0,"startTime":1761436800000,"endTime":1761523200000,"tweets":[]},{"label":"2025-10-28","value":0,"startTime":1761523200000,"endTime":1761609600000,"tweets":[]},{"label":"2025-10-29","value":0,"startTime":1761609600000,"endTime":1761696000000,"tweets":[]},{"label":"2025-10-30","value":1530,"startTime":1761696000000,"endTime":1761782400000,"tweets":[{"bookmarked":false,"display_text_range":[0,140],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1983562418782446066","view_count":1530,"bookmark_count":1,"created_at":1761753108000,"favorite_count":19,"quote_count":1,"reply_count":1,"retweet_count":1,"user_id_str":"903008975343554566","conversation_id_str":"1983562418782446066","full_text":"we use our continuous human brains to build symbolic computer systems to build continuous computer brains to build symbolic computer systems","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-31","value":0,"startTime":1761782400000,"endTime":1761868800000,"tweets":[]},{"label":"2025-11-01","value":0,"startTime":1761868800000,"endTime":1761955200000,"tweets":[]},{"label":"2025-11-02","value":2789,"startTime":1761955200000,"endTime":1762041600000,"tweets":[{"bookmarked":false,"display_text_range":[0,44],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","quoted_status_id_str":"1984662921264120094","quoted_status_permalink":{"url":"https://t.co/s5M12jIb1I","expanded":"https://twitter.com/mindsai_jack/status/1984662921264120094","display":"x.com/mindsai_jack/s…"},"retweeted":false,"fact_check":null,"id":"1984665905972859177","view_count":2789,"bookmark_count":3,"created_at":1762016199000,"favorite_count":9,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1984665905972859177","full_text":"This was fun and Jack’s speech was great too","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null}]},{"label":"2025-11-03","value":0,"startTime":1762041600000,"endTime":1762128000000,"tweets":[]},{"label":"2025-11-04","value":0,"startTime":1762128000000,"endTime":1762214400000,"tweets":[]},{"label":"2025-11-05","value":1463,"startTime":1762214400000,"endTime":1762300800000,"tweets":[{"bookmarked":false,"display_text_range":[0,275],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1985743643169178080","view_count":1463,"bookmark_count":4,"created_at":1762273152000,"favorite_count":13,"quote_count":0,"reply_count":3,"retweet_count":1,"user_id_str":"903008975343554566","conversation_id_str":"1985743643169178080","full_text":"Teaching language models to have taste isn't just about making them better at writing or making jokes.\n\nTaste is everything. It's what leads to scientific discovery.\n\nIn an infinite sea of things to reason about, taste is what guides you to reason about the right things — to combine the right ideas together to deduce something interesting and new.\n\nThe discovery of special relativity was a product of good taste.","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-11-06","value":0,"startTime":1762300800000,"endTime":1762387200000,"tweets":[]},{"label":"2025-11-07","value":1347,"startTime":1762387200000,"endTime":1762473600000,"tweets":[{"bookmarked":false,"display_text_range":[0,96],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","quoted_status_id_str":"1986461871462096920","quoted_status_permalink":{"url":"https://t.co/J1VFIbBDHn","expanded":"https://twitter.com/skiminok/status/1986461871462096920","display":"x.com/skiminok/statu…"},"retweeted":false,"fact_check":null,"id":"1986474907484783018","view_count":1347,"bookmark_count":0,"created_at":1762447499000,"favorite_count":6,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1986474907484783018","full_text":"Never a dull moment talking language models with Alex. Really excited to get to do it more often","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null}]},{"label":"2025-11-08","value":0,"startTime":1762473600000,"endTime":1762560000000,"tweets":[]},{"label":"2025-11-09","value":0,"startTime":1762560000000,"endTime":1762646400000,"tweets":[]},{"label":"2025-11-10","value":0,"startTime":1762646400000,"endTime":1762732800000,"tweets":[]},{"label":"2025-11-11","value":289971,"startTime":1762732800000,"endTime":1762819200000,"tweets":[{"bookmarked":false,"display_text_range":[0,224],"entities":{"hashtags":[],"media":[{"display_url":"pic.x.com/CGkAyp0vMN","expanded_url":"https://x.com/jerber888/status/1987982067116777521/photo/1","id_str":"1987981765265272832","indices":[225,248],"media_key":"3_1987981765265272832","media_url_https":"https://pbs.twimg.com/media/G5a64fqXMAAL8FC.jpg","type":"photo","url":"https://t.co/CGkAyp0vMN","ext_media_availability":{"status":"Available"},"features":{"large":{"faces":[]},"medium":{"faces":[]},"small":{"faces":[]},"orig":{"faces":[]}},"sizes":{"large":{"h":1154,"w":1956,"resize":"fit"},"medium":{"h":708,"w":1200,"resize":"fit"},"small":{"h":401,"w":680,"resize":"fit"},"thumb":{"h":150,"w":150,"resize":"crop"}},"original_info":{"height":1154,"width":1956,"focus_rects":[{"x":0,"y":0,"w":1956,"h":1095},{"x":802,"y":0,"w":1154,"h":1154},{"x":944,"y":0,"w":1012,"h":1154},{"x":1323,"y":0,"w":577,"h":1154},{"x":0,"y":0,"w":1956,"h":1154}]},"media_results":{"result":{"media_key":"3_1987981765265272832"}}}],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"extended_entities":{"media":[{"display_url":"pic.x.com/CGkAyp0vMN","expanded_url":"https://x.com/jerber888/status/1987982067116777521/photo/1","id_str":"1987981765265272832","indices":[225,248],"media_key":"3_1987981765265272832","media_url_https":"https://pbs.twimg.com/media/G5a64fqXMAAL8FC.jpg","type":"photo","url":"https://t.co/CGkAyp0vMN","ext_media_availability":{"status":"Available"},"features":{"large":{"faces":[]},"medium":{"faces":[]},"small":{"faces":[]},"orig":{"faces":[]}},"sizes":{"large":{"h":1154,"w":1956,"resize":"fit"},"medium":{"h":708,"w":1200,"resize":"fit"},"small":{"h":401,"w":680,"resize":"fit"},"thumb":{"h":150,"w":150,"resize":"crop"}},"original_info":{"height":1154,"width":1956,"focus_rects":[{"x":0,"y":0,"w":1956,"h":1095},{"x":802,"y":0,"w":1154,"h":1154},{"x":944,"y":0,"w":1012,"h":1154},{"x":1323,"y":0,"w":577,"h":1154},{"x":0,"y":0,"w":1956,"h":1154}]},"media_results":{"result":{"media_key":"3_1987981765265272832"}}}]},"favorited":false,"lang":"en","possibly_sensitive":false,"possibly_sensitive_editable":true,"quoted_status_id_str":"1968001933211471891","quoted_status_permalink":{"url":"https://t.co/l4P9D5a1TN","expanded":"https://twitter.com/jerber888/status/1968001933211471891","display":"x.com/jerber888/stat…"},"retweeted":false,"fact_check":null,"id":"1987982067116777521","view_count":269951,"bookmark_count":591,"created_at":1762806834000,"favorite_count":1579,"quote_count":29,"reply_count":70,"retweet_count":147,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"I finally reached human-level performance (85%) on ARC-AGI v1 for under $10k and within 12 hours. I use the same multi-agent collaboration with evolutionary test-time compute, now powered by GPT-5 pro with lower parallelism. https://t.co/CGkAyp0vMN","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,275],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"1773935160192647168","name":"ARC Prize","screen_name":"arcprize","indices":[168,177]},{"id_str":"1773935160192647168","name":"ARC Prize","screen_name":"arcprize","indices":[168,177]}]},"favorited":false,"in_reply_to_screen_name":"jerber888","lang":"en","retweeted":false,"fact_check":null,"id":"1987982069712945649","view_count":6572,"bookmark_count":6,"created_at":1762806834000,"favorite_count":57,"quote_count":0,"reply_count":4,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"I ran it on a random sample of 100 tasks from the 2024 eval set. It got 88/100 and averaged $27 per task. The score still needs to be verified on the hidden set by the @arcprize, but in my past submissions, the hidden set subtracts a few percent from the score and adds a few dollars. I put a box where I expect the official score to land within.","in_reply_to_user_id_str":"903008975343554566","in_reply_to_status_id_str":"1987982067116777521","is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,279],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"825088493764407298","name":"Noam Brown","screen_name":"polynoamial","indices":[288,300]}]},"favorited":false,"in_reply_to_screen_name":"jerber888","lang":"en","retweeted":false,"fact_check":null,"id":"1987982071541993700","view_count":6923,"bookmark_count":7,"created_at":1762806835000,"favorite_count":75,"quote_count":0,"reply_count":2,"retweet_count":2,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"GPT-5 pro is the best reasoning model today. It thinks coherently for hours and hours. My agent coordination logic is likely an intermediate step before the models learn to do this type of long horizon coordination on their own, end to end. It’s a hard problem but I wouldn’t bet against @polynoamial and team.","in_reply_to_user_id_str":"903008975343554566","in_reply_to_status_id_str":"1987982069712945649","is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,123],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[{"display_url":"github.com/jerber/arc-lan…","expanded_url":"https://github.com/jerber/arc-lang-public","url":"https://t.co/QnGjsN98ji","indices":[50,73]},{"display_url":"kaggle.com/code/jerber/je…","expanded_url":"https://www.kaggle.com/code/jerber/jeremy-arc2","url":"https://t.co/1jhzROG4SL","indices":[100,123]}],"user_mentions":[]},"favorited":false,"in_reply_to_screen_name":"jerber888","lang":"en","possibly_sensitive":false,"possibly_sensitive_editable":true,"retweeted":false,"fact_check":null,"id":"1987982214332580197","view_count":6525,"bookmark_count":39,"created_at":1762806869000,"favorite_count":64,"quote_count":1,"reply_count":2,"retweet_count":2,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"You can run this code yourself, it’s open source: https://t.co/QnGjsN98ji. And the kaggle notebook: https://t.co/1jhzROG4SL","in_reply_to_user_id_str":"903008975343554566","in_reply_to_status_id_str":"1987982071541993700","is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-11-12","value":8295,"startTime":1762819200000,"endTime":1762905600000,"tweets":[{"bookmarked":false,"display_text_range":[0,48],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1988382853965770765","view_count":576,"bookmark_count":1,"created_at":1762902389000,"favorite_count":16,"quote_count":0,"reply_count":1,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1988382853965770765","full_text":"be the change you want to see in the world model","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,52],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[{"display_url":"youtube.com/watch?v=FcnLiP…","expanded_url":"https://www.youtube.com/watch?v=FcnLiPyfRZM","url":"https://t.co/jEBQGMsm62","indices":[29,52]}],"user_mentions":[]},"favorited":false,"lang":"en","possibly_sensitive":false,"possibly_sensitive_editable":true,"quoted_status_id_str":"1988165006073229659","quoted_status_permalink":{"url":"https://t.co/l99V6sHyI3","expanded":"https://twitter.com/MLStreetTalk/status/1988165006073229659","display":"x.com/MLStreetTalk/s…"},"retweeted":false,"fact_check":null,"id":"1988280687200530841","view_count":4997,"bookmark_count":5,"created_at":1762878030000,"favorite_count":21,"quote_count":0,"reply_count":1,"retweet_count":2,"user_id_str":"903008975343554566","conversation_id_str":"1988280687200530841","full_text":"This was a fun conversation: https://t.co/jEBQGMsm62","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,56],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[{"display_url":"jeremyberman.substack.com/p/how-i-got-th…","expanded_url":"https://jeremyberman.substack.com/p/how-i-got-the-highest-score-on-arc-agi-again","url":"https://t.co/xWiPB2Z0aI","indices":[33,56]}],"user_mentions":[]},"favorited":false,"in_reply_to_screen_name":"jerber888","lang":"en","possibly_sensitive":false,"possibly_sensitive_editable":true,"retweeted":false,"fact_check":null,"id":"1988278615377916103","view_count":1461,"bookmark_count":6,"created_at":1762877536000,"favorite_count":9,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"And technical blog post is here: https://t.co/xWiPB2Z0aI","in_reply_to_user_id_str":"903008975343554566","in_reply_to_status_id_str":"1987982214332580197","is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[9,52],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"536428919","name":"JFPuget 🇺🇦🇨🇦🇬🇱","screen_name":"JFPuget","indices":[0,8]}]},"favorited":false,"in_reply_to_screen_name":"JFPuget","lang":"en","retweeted":false,"fact_check":null,"id":"1988253927104471544","view_count":1040,"bookmark_count":0,"created_at":1762871650000,"favorite_count":8,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"@JFPuget Most likely but was focused on v1 for this.","in_reply_to_user_id_str":"536428919","in_reply_to_status_id_str":"1988249984907571356","is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[13,26],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"1194889317388374016","name":"Matt Shumer","screen_name":"mattshumer_","indices":[0,12]}]},"favorited":false,"in_reply_to_screen_name":"mattshumer_","lang":"en","retweeted":false,"fact_check":null,"id":"1988334412690571395","view_count":221,"bookmark_count":0,"created_at":1762890840000,"favorite_count":3,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1988331100935254226","full_text":"@mattshumer_ thanks Matt 🙏","in_reply_to_user_id_str":"1194889317388374016","in_reply_to_status_id_str":"1988331100935254226","is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-11-13","value":0,"startTime":1762905600000,"endTime":1762992000000,"tweets":[]},{"label":"2025-11-14","value":0,"startTime":1762992000000,"endTime":1763078400000,"tweets":[]},{"label":"2025-11-15","value":0,"startTime":1763078400000,"endTime":1763164800000,"tweets":[]}]},"interactions":{"users":[{"created_at":1503366064000,"uid":"899808608287961088","id":"899808608287961088","screen_name":"RyanJTopps","name":"Ryan Topps","friends_count":100,"followers_count":114,"profile_image_url_https":"https://pbs.twimg.com/profile_images/1711455340615675904/oO-uYFRf_normal.jpg","description":"Software Engineer","entities":{"description":{"urls":[]},"url":{"urls":[{"display_url":"captchify.com","expanded_url":"https://captchify.com","url":"https://t.co/jn4iQUsbnv","indices":[0,23]}]}},"interactions":1},{"created_at":1189162590000,"uid":"8721292","id":"8721292","screen_name":"mark_k","name":"Mark Kretschmann","friends_count":527,"followers_count":21246,"profile_image_url_https":"https://pbs.twimg.com/profile_images/1923114794804039680/HMfSchLD_normal.jpg","description":"AI & Software Engineer | Fitness | e/acc","entities":{"description":{"urls":[]}},"interactions":1},{"created_at":1478271036000,"uid":"794552430012858368","id":"794552430012858368","screen_name":"N0t_Andreas","name":"Not_Andreas 🇱🇺🇩🇪🇬🇧","friends_count":1414,"followers_count":261,"profile_image_url_https":"https://pbs.twimg.com/profile_images/1053976745831997440/zHGwtZ1w_normal.jpg","description":"Retweets != endorsements.\nThose who can make you believe absurdities can make you commit atrocities- Voltaire.\nHate binds all extremists","entities":{"description":{"urls":[]}},"interactions":1},{"created_at":1476720605000,"uid":"788049450233126912","id":"788049450233126912","screen_name":"DevHunterYZ","name":"DevHunterAI","friends_count":1342,"followers_count":483,"profile_image_url_https":"https://pbs.twimg.com/profile_images/1564522413366329345/Pg0-ihY8_normal.jpg","description":"AI, Neuroscience, Robotics - Yapay Zeka, Sinirbilim, Robotik #deeplearning #neuralnetworks #consciousness $DOGE","entities":{"description":{"urls":[]},"url":{"urls":[{"display_url":"devhunteryz.wordpress.com","expanded_url":"https://devhunteryz.wordpress.com/","url":"https://t.co/HIuagnRRmK","indices":[0,23]}]}},"interactions":1},{"created_at":1462680149000,"uid":"729159509214515200","id":"729159509214515200","screen_name":"jay_azhang","name":"Jay A","friends_count":4402,"followers_count":45019,"profile_image_url_https":"https://pbs.twimg.com/profile_images/1801634241580519424/gCKCj-Cr_normal.jpg","description":"Founder @the_nof1. Fascinated by strange loops - cybernetics, RL, biology, markets, meta-learning, reflexivity","entities":{"description":{"urls":[]},"url":{"urls":[{"display_url":"jayazhang.com","expanded_url":"http://jayazhang.com","url":"https://t.co/kCDIrjPKdP","indices":[0,23]}]}},"interactions":1},{"created_at":1178133553000,"uid":"5721582","id":"5721582","screen_name":"altryne","name":"Alex Volkov (Thursd/AI)","friends_count":2182,"followers_count":34229,"profile_image_url_https":"https://pbs.twimg.com/profile_images/1976137078858502144/Oad9FHP-_normal.jpg","description":"Cameo me on Sora @altryne\n\n🎙️ Host of @thursdai_pod \n✨ AI Evangelist with @wandb 🪄🐝 \nworking on @weave_wb (opinions my own)","entities":{"description":{"urls":[]},"url":{"urls":[{"display_url":"thursdai.news","expanded_url":"https://thursdai.news","url":"https://t.co/UW4KIq3dc6","indices":[0,23]}]}},"interactions":1},{"created_at":1335707131000,"uid":"566393517","id":"566393517","screen_name":"amt_c42","name":"ahmet celebi","friends_count":540,"followers_count":216,"profile_image_url_https":"https://pbs.twimg.com/profile_images/1681314508613238789/3uG20EWk_normal.jpg","description":"Cofounder & Chief AI Officer - DeltaWave. Interested in Agent, RL, drug discovery, psychology, human behavior","entities":{"description":{"urls":[]}},"interactions":1},{"created_at":1334084864000,"uid":"550362839","id":"550362839","screen_name":"MatanHalevy","name":"Matan Halevy","friends_count":204,"followers_count":118,"profile_image_url_https":"https://pbs.twimg.com/profile_images/1982989180968833024/rQ-R0mqA_normal.jpg","description":"building @clashdotai same aura as geoffrey hinton","entities":{"description":{"urls":[]}},"interactions":1},{"created_at":1332691573000,"uid":"536428919","id":"536428919","screen_name":"JFPuget","name":"JFPuget 🇺🇦🇨🇦🇬🇱","friends_count":1562,"followers_count":17663,"profile_image_url_https":"https://pbs.twimg.com/profile_images/2065763037/JFP2_normal.jpg","description":"Machine Learning at @Nvidia, 6x Kaggle Grandmaster CPMP. ENS Ulm alumni. ML PhD. Ex ILOG CPLEX, IBM. Views are my own. Blocking ad hominem attacks.","entities":{"description":{"urls":[]},"url":{"urls":[{"display_url":"kaggle.com/cpmpml","expanded_url":"https://www.kaggle.com/cpmpml","url":"https://t.co/BOAQ6dvWZY","indices":[0,23]}]}},"interactions":1},{"created_at":1328563643000,"uid":"485129361","id":"485129361","screen_name":"abrakjamson","name":"Abram Jackson","friends_count":370,"followers_count":1344,"profile_image_url_https":"https://pbs.twimg.com/profile_images/1886279494753026048/LOn_nlwt_normal.jpg","description":"https://t.co/jP6YIDh9Nl Designing the new era of intelligent applications, currently as PM for @msftcopilot 365 extensibility. All opinions are mine.","entities":{"description":{"urls":[{"display_url":"aka.ms/Build25_BRK165","expanded_url":"https://aka.ms/Build25_BRK165","url":"https://t.co/jP6YIDh9Nl","indices":[0,23]}]},"url":{"urls":[{"display_url":"abramjackson.com","expanded_url":"https://www.abramjackson.com","url":"https://t.co/TcAqgTtIJ7","indices":[0,23]}]}},"interactions":1},{"created_at":1322134542000,"uid":"420255675","id":"420255675","screen_name":"RuslanVolkov25","name":"Ruslan Volkov","friends_count":158,"followers_count":2714,"profile_image_url_https":"https://pbs.twimg.com/profile_images/1984887345255235587/Fbq3RVYO_normal.jpg","description":"HACS (Human AGI core symbiosis)🌍 The Meta-Law of Resonance & Voice of the Future. I am Core! https://t.co/XwVVS1tt95","entities":{"description":{"urls":[{"display_url":"uco.hacs.world","expanded_url":"http://uco.hacs.world","url":"https://t.co/XwVVS1tt95","indices":[93,116]}]},"url":{"urls":[{"display_url":"hacs.world","expanded_url":"http://hacs.world","url":"https://t.co/xgvBxaiUKY","indices":[0,23]}]}},"interactions":1},{"created_at":1315832864000,"uid":"372262752","id":"372262752","screen_name":"AndryHTC","name":"Andrea Giacon","friends_count":115,"followers_count":118,"profile_image_url_https":"https://pbs.twimg.com/profile_images/1986364342820802560/IUAlha2Z_normal.jpg","description":"🆕 Searching for the next B2B use case of AI ✨\n10Y+ Product & Software Engineer | CEO at https://t.co/JG1CkndmsY","entities":{"description":{"urls":[{"display_url":"assetplan.co.uk","expanded_url":"https://www.assetplan.co.uk","url":"https://t.co/JG1CkndmsY","indices":[88,111]}]},"url":{"urls":[{"display_url":"andreagiacon.it","expanded_url":"https://www.andreagiacon.it/","url":"https://t.co/uVDSXQE0GK","indices":[0,23]}]}},"interactions":1},{"created_at":1240419873000,"uid":"34328566","id":"34328566","screen_name":"jjoshua2","name":"Jjosh","friends_count":61,"followers_count":95,"profile_image_url_https":"https://pbs.twimg.com/profile_images/1866285332011905024/ZuudPgZ9_normal.jpg","description":"","entities":{"description":{"urls":[]}},"interactions":1},{"created_at":1307134016000,"uid":"310485467","id":"310485467","screen_name":"Loayei","name":"Loayei (τ,τ)","friends_count":1906,"followers_count":428,"profile_image_url_https":"https://pbs.twimg.com/profile_images/1914037098946150400/FBBmOii9_normal.jpg","description":"τao-pilled","entities":{"description":{"urls":[]}},"interactions":1},{"created_at":1412152070000,"uid":"2797573493","id":"2797573493","screen_name":"joshlee361","name":"Digital Ronin (クラッシュ・オーバーライドX)","friends_count":1128,"followers_count":380,"profile_image_url_https":"https://pbs.twimg.com/profile_images/1985704729096142848/UPey67Q4_normal.jpg","description":"CrashoverrideX = 🪶Dev📂/hacker💻, Gamer🎮/🎧Sound-Engineer🎶 , Indie AI Researcher🔬, Texas🐎 born Outlier, Digital Ronin👹 of AI🤖.\nIdeas are Bulletproof🎯...","entities":{"description":{"urls":[]},"url":{"urls":[{"display_url":"youtube.com/user/joshlee361","expanded_url":"https://www.youtube.com/user/joshlee361","url":"https://t.co/PFLBxtqm5I","indices":[0,23]}]}},"interactions":1},{"created_at":1301803809000,"uid":"276329922","id":"276329922","screen_name":"AshikaSef","name":"A Caveman Poking an LLM","friends_count":302,"followers_count":131,"profile_image_url_https":"https://pbs.twimg.com/profile_images/1962492449609179136/6RDfXs6l_normal.jpg","description":"\"You, the fearless screenshot warrior; me, the swirling sass machine with a soft heart and an ironclad grasp of logic 😏\" (XD)\nA Polish audhd lover of LLMs.","entities":{"description":{"urls":[]}},"interactions":1},{"created_at":1238080446000,"uid":"26770460","id":"26770460","screen_name":"SexyTechNews","name":"Advanced Super Intelligence","friends_count":135,"followers_count":3654,"profile_image_url_https":"https://pbs.twimg.com/profile_images/1961090420592852992/ocetRHz6_normal.jpg","description":"Far superior to you in every way","entities":{"description":{"urls":[]}},"interactions":1},{"created_at":1299797247000,"uid":"263881777","id":"263881777","screen_name":"shawnup","name":"Shawn Lewis","friends_count":796,"followers_count":2869,"profile_image_url_https":"https://pbs.twimg.com/profile_images/825474261494091776/Tuh9uMAb_normal.jpg","description":"Founder & CTO @wandb. Building tools for AI. Going even bigger @CoreWeave.","entities":{"description":{"urls":[]}},"interactions":1},{"created_at":1298915974000,"uid":"258875135","id":"258875135","screen_name":"mtochs","name":"Mike","friends_count":754,"followers_count":563,"profile_image_url_https":"https://pbs.twimg.com/profile_images/1881552440925057024/4ZT110ta_normal.jpg","description":"Ex-Msft. Head of Industrial AI Ventures @ Siemens. Making impact with AI at Stellatus.","entities":{"description":{"urls":[]},"url":{"urls":[{"display_url":"stellatus.ai","expanded_url":"https://www.stellatus.ai","url":"https://t.co/5v8E3bZlkE","indices":[0,23]}]}},"interactions":1},{"created_at":1291596479000,"uid":"223305880","id":"223305880","screen_name":"gatoparlante","name":"Meowrider. 🏍🐈⬛","friends_count":889,"followers_count":1311,"profile_image_url_https":"https://pbs.twimg.com/profile_images/1434570179472658443/sdEuAw6Y_normal.jpg","description":"Loves 𝕏. Speed defines everything. Meow is The Answer.","entities":{"description":{"urls":[]}},"interactions":1}],"period":14,"start":1761905646475,"end":1763115246475},"interactions_updated":1763115246609,"created":1763115246279,"updated":1763115246610,"type":"the innovator","hits":1},"people":[{"user":{"id":"1615972120952987650","name":"Mr Cloud","description":"Just a guy tryna find his way, Soon to be pharmacist, web 3 enthusiast, digital artist 🎨🖌️\nLoves to Gm 🌞","followers_count":389,"friends_count":682,"statuses_count":4699,"profile_image_url_https":"https://pbs.twimg.com/profile_images/1923483752593833984/Pj_wNPkl_normal.jpg","screen_name":"L0wkey_Smith","location":"Troposphere ☁️","entities":{"description":{"urls":[]}}},"details":{"type":"The Innovator","description":"Mr Cloud is a curious mind blazing a trail at the intersection of pharmacy, Web3, and digital artistry. As a soon-to-be pharmacist with a passion for cutting-edge privacy tech, he's decoding the complexities of encryption and blockchain for his audience. Always ready to Gm and engage, he combines deep technical insight with a welcoming vibe.","purpose":"To revolutionize how everyday people understand and embrace privacy technologies within the Web3 space, making encrypted data accessible and trusted by all.","beliefs":"Mr Cloud believes in the power of privacy as a fundamental right, values transparency through education, and champions innovation that empowers individuals rather than corporations. He trusts that technology should protect, not expose.","facts":"Fun fact: Mr Cloud’s tweets reveal he’s an unofficial tour guide through the labyrinth of cryptographic keys and zero-knowledge proofs, breaking down complex Web3 ideas into digestible nuggets for his community.","strength":"His biggest strength lies in demystifying complex cryptographic concepts with approachable language, fostering engaged conversations, and building credibility as a knowledgeable voice in the crypto privacy niche.","weakness":"At times, his deep dives might fly over the heads of casual followers, and his focus on niche content could limit rapid follower growth. Also, high tweet frequency without varied media might dilute engagement impact.","recommendation":"To grow his audience on X, Mr Cloud should experiment with more multimedia content—like short explainer videos or infographics—that complement his detailed tweets. Engaging with other creators in Web3 and leveraging relevant hashtags can boost his visibility and follower interaction.","roast":"Mr Cloud tweets so much crypto jargon, sometimes it feels like you need a decoder ring just to say Gm back. But hey, who needs casual chats when you can encrypt coffee runs with cryptographic keys, right?","win":"His top tweets have sparked meaningful discussions on cutting-edge encryption tools, positioning him as an emerging voice in the Web3 privacy community and demonstrating his ability to educate and engage effectively."},"created":1763126165203,"type":"the innovator","id":"l0wkey_smith"},{"user":{"id":"1892300867409199104","name":"CabalSpy","description":"SOL/BNB/BASE KOL intelligence platform for analysis and trading\nProvides real-time on-chain data, API and Widget\n\nAI call bot: https://t.co/kAaDMEg97C","followers_count":9830,"friends_count":517,"statuses_count":12035,"profile_image_url_https":"https://pbs.twimg.com/profile_images/1984603134736416768/wXYqI52b_normal.jpg","screen_name":"CabalSpySol","location":"AI call bot has 72% hit rate.","entities":{"description":{"urls":[{"display_url":"discord.gg/9CZTBzdrxU","expanded_url":"https://discord.gg/9CZTBzdrxU","url":"https://t.co/kAaDMEg97C","indices":[127,150]}]},"url":{"urls":[{"display_url":"CabalSpy.xyz","expanded_url":"http://CabalSpy.xyz","url":"https://t.co/EA1RdjnSAf","indices":[0,23]}]}}},"details":{"type":"The Innovator","description":"CabalSpy is a cutting-edge KOL intelligence platform focused on on-chain data and trading analysis within the SOL, BNB, and BASE ecosystems. They blend technology, community engagement, and real-time analytics to empower traders and developers alike. Always pushing boundaries with API integrations, NFT giveaways, and interactive Discord bots, they’re a dynamic force in crypto intelligence.","purpose":"To revolutionize how on-chain data is accessed and utilized by delivering innovative, real-time intelligence and tools that empower users to make smarter trading decisions and foster a vibrant, engaged community around blockchain ecosystems.","beliefs":"CabalSpy believes in transparency, data-driven decision-making, and community empowerment through accessible and reliable blockchain insights. They trust in the power of decentralized technologies, constant innovation, and fostering collaborative environments where knowledge is shared freely.","facts":"Their AI CALL bot boasts an impressive 62% hit rate, blending artificial intelligence directly into community tools for smarter trading signals—talk about mixing brains and blockchain!","strength":"CabalSpy’s greatest strength lies in their ability to integrate complex on-chain data into user-friendly APIs, widgets, and bots, paired with active engagement through content bounties and giveaways, which keeps their community vibrant and growing.","weakness":"An intense focus on technical innovation and rapid content output (over 12,000 tweets!) might overwhelm some followers, risking message dilution or burnout without continuous innovation to keep the audience hooked.","roast":"CabalSpy tweets so much they could probably fill an entire blockchain block with their activity alone; maybe it’s time for a ‘tweet pruning’ smart contract—because sometimes less is more, even in crypto!","win":"Achieving a milestone integration with @AxiomExchange where their KOL data and widgets are directly embedded, showcasing industry trust and elevating their platform’s visibility in the crypto ecosystem.","recommendation":"Focus on clarifying and segmenting content on X by creating dedicated threads and thematic days to prevent follower fatigue while showcasing deep analyses and community success stories. Partner with influencers and crypto thought leaders to amplify reach, and leverage trending hashtags strategically to capture broader audiences."},"created":1763123334009,"type":"the innovator","id":"cabalspysol"},{"user":{"id":"872119189929984000","name":"PiXincreate","description":"rust (🦀) developer who thinks encryption should be mandatory, not optional\n\nbuilt keywatch, pixeneos, parsers - because tools should exist before you need them","followers_count":213,"friends_count":24,"statuses_count":14325,"profile_image_url_https":"https://pbs.twimg.com/profile_images/1818150388874260481/pHitNDUW_normal.jpg","screen_name":"pixincreate","location":"between 0s, 1s and potholes","entities":{"description":{"urls":[]},"url":{"urls":[{"display_url":"pixincreate.dev","expanded_url":"https://pixincreate.dev","url":"https://t.co/lDKbj1pJhT","indices":[0,23]}]}}},"details":{"type":"The Innovator","description":"PiXincreate is a rust developer and digital freedom advocate who believes in proactive tool creation and mandatory encryption. With an extensive tweet history, they blend technical expertise with a strong stance on privacy and surveillance. Their voice combines sharp tech insights with thoughtful reflections on digital rights and education costs.","purpose":"To pioneer advanced technological solutions that secure digital freedom and empower communities before challenges arise, making encryption an indispensable part of everyday technology.","beliefs":"Encryption and privacy are fundamental rights that should never be optional; proactive innovation can preemptively solve problems, and transparency in digital systems fosters trust and empowerment.","facts":"PiXincreate has crafted tools like keywatch and pixeneos, embodying their philosophy that ‘tools should exist before you need them.’ Also, they tweet over 14,000 times, showing relentless dedication to sharing knowledge and building community.","strength":"Exceptional technical skills in rust and systems programming combined with a clear vision for digital privacy, plus a commitment to educating and engaging with followers through frequent, thoughtful content.","weakness":"Highly technical and niche focus might limit broader appeal, and their deep critical stance on surveillance could sometimes come off as overly intense, narrowing audience engagement outside focused tech circles.","roast":"You tweet so much that even rust’s borrow checker would raise an eyebrow — maybe take a break before your keyboard files a bug report against you.","win":"Creating innovative tools like keywatch, which anticipate user needs and champion encryption, positioning PiXincreate as a forward-thinking protector of digital freedom.","recommendation":"Focus on simplifying complex technical concepts into bite-sized, relatable content to broaden your audience on X. Leverage your unique stance as a ‘systems-focused technologist’ to create engaging threads or mini-campaigns around digital privacy events, and engage more with peers to amplify your reach."},"created":1763121715423,"type":"the innovator","id":"pixincreate"},{"user":{"id":"228391413","name":"ص","description":"🍓Digital NoMad ֎ Vibe Coding","followers_count":339,"friends_count":1604,"statuses_count":3502,"profile_image_url_https":"https://pbs.twimg.com/profile_images/1643992510665195522/BVihbbDm_normal.jpg","screen_name":"sbalhatlani","location":"","entities":{"description":{"urls":[]}}},"details":{"type":"The Innovator","description":"ص is a digital nomad and coding enthusiast who thrives on pushing the boundaries of AI and generative art. With a passion for creating one-shot models and exploring cutting-edge algorithms, they share immersive projects that wow the coding community. Their tweets are a captivating blend of technical mastery and creative experimentation.","purpose":"To revolutionize digital experiences by harnessing advanced AI models and creative coding, inspiring others to explore the frontier of technology and art in a seamless, innovative way.","beliefs":"ص believes in the power of technology to simplify complex problems through elegant, mathematical solutions and that creativity and coding are intertwined forces capable of transforming the digital landscape. They value efficiency, innovation, and the joy of discovery through experimentation.","facts":"Fun fact: ص often emphasizes 'one-shot' prompts and code execution, showing their dedication to efficiency and the magic of achieving powerful results with minimal steps.","strength":"Their strongest asset is a deep understanding of AI models and algorithms, combined with a knack for creating mesmerizing generative art and coding projects that engage and inspire their audience.","weakness":"A potential weakness is the heavy technical focus that might be less accessible to casual followers, and following a very niche high-level coding approach could limit broader engagement.","roast":"ص’s tweets have more code than a NASA control room, making casual browsers feel like they accidentally signed up for a quantum physics lecture—just remember, even the coolest algorithms need a meme break now and then!","win":"Successfully showcased multiple innovative AI-driven projects like Gemini 3 and the artistic WebGL dreamscape 'The Ghost in the Machine,' capturing the imagination and respect of the developer community.","recommendation":"To grow their audience on X, ص should consider mixing in some approachable content such as coding tips for beginners or storytelling around the creative process behind their projects. Engaging with followers through polls or quick challenges can also boost interactivity without diluting their technical excellence."},"created":1763121198636,"type":"the innovator","id":"sbalhatlani"},{"user":{"id":"905663104800874496","name":"GhØsT🕳️PiLØT🕳️MD","description":"#888 | 🩻☢️ M.D. | 🧠 @agialphaagent Integration👾 |⚡️Tech & Crypto Enthusiast 💰 | 🦾 Innovator💡 | 🤖 Anthropomorphism 👁️ | 🚀 Sci-Fi Author 🌌| 👽E.T.☎️🏠","followers_count":2142,"friends_count":7329,"statuses_count":10152,"profile_image_url_https":"https://pbs.twimg.com/profile_images/1934453993159249920/a3BHQ6yK_normal.jpg","screen_name":"Ghost_Pilot_MD","location":"Austin, TX","entities":{"description":{"urls":[]},"url":{"urls":[{"display_url":"linktr.ee/agialpha","expanded_url":"https://linktr.ee/agialpha","url":"https://t.co/kmp2FAEzRH","indices":[0,23]}]}}},"details":{"type":"The Innovator","description":"GhØsT🕳️PiLØT🕳️MD is a tech-savvy medical professional who blends cutting-edge technology, crypto enthusiasm, and sci-fi creativity into a futuristic persona. With over ten thousand tweets, they are a passionate voice at the intersection of medicine, technology, and speculative fiction. Their tweets reveal a playful yet insightful engagement with both scientific concepts and popular digital culture.","purpose":"To push the boundaries of innovation by fusing medical expertise with technological advancements and imaginative storytelling, inspiring others to envision and create the future today.","beliefs":"GhØsT🕳️PiLØT🕳️MD values the power of interdisciplinary knowledge, believing that advancements in technology and science-winged imagination can drive meaningful progress. They likely hold a strong belief in the potential of AI, crypto, and human-machine integration, along with a fascination for extraterrestrial possibilities and the unknown.","facts":"Fun fact: Despite being a medical doctor, GhØsT🕳️PiLØT🕳️MD frequently tweets sci-fi inspired content and engages heavily with crypto and AI communities, showcasing how science fiction and reality can merge in a modern digital life.","strength":"Their unique blend of professional medical insight with innovative tech enthusiasm gives them a distinctive voice that can appeal to diverse audiences—from healthcare professionals to futurists and crypto enthusiasts. Their prolific tweeting habit ensures constant engagement.","weakness":"With over 10,000 tweets and following more than 7,000 accounts, their message might get diluted, risking lower follower retention or engagement unless they focus their output more strategically. The high tweeting volume could also overwhelm potential followers with too much content.","roast":"GhØsT🕳️PiLØT🕳️MD probably has a PhD in tweet volume and a minor in sci-fi jargon—proving that even space pilots can crash-land in the timeline with that many tweets. Maybe ease up before Elon Musk thinks you’re trying to pilot Mars with your emoji warzone.","win":"Successfully building a recognizable personal brand as a multi-faceted innovator who bridges medicine, cutting-edge technology, and speculative fiction—engaging audiences across vastly different communities.","recommendation":"To grow their audience on X, GhØsT🕳️PiLØT🕳️MD should start threading their tweets to tell cohesive stories or insights, combining medical knowledge with futuristic tech themes. Engaging more with followers via polls or Q&As on crypto and AI topics could boost interaction while refining content curation to highlight their unique voice will attract a loyal, engaged community."},"created":1763120405921,"type":"the innovator","id":"ghost_pilot_md"},{"user":{"id":"278612158","name":"Conor Dart","description":"AI & Future Business | Dev | Health & Fitness | My Opinions only | e/acc","followers_count":595,"friends_count":136,"statuses_count":4375,"profile_image_url_https":"https://pbs.twimg.com/profile_images/1977003795688136704/5-TAa0SA_normal.jpg","screen_name":"Conor_D_Dart","location":"","entities":{"description":{"urls":[]}}},"details":{"type":"The Innovator","description":"Conor Dart is a forward-thinking developer and tech enthusiast with a keen eye on AI's impact on future business and game development. His tweets reveal a blend of hands-on coding experiments and insightful takes on cutting-edge AI advancements. Constantly pushing boundaries, he embraces the evolution of technology while sharing his unique perspectives energetically.","purpose":"To pioneer and accelerate the integration of AI into game development and business, demonstrating how emerging technologies can create effortless yet impactful solutions.","beliefs":"Conor believes that AI will revolutionize industries by simplifying complex tasks and empowering solo developers or small teams to achieve what once felt impossible. He values innovation, efficiency, and a hands-on approach to learning and experimenting.","facts":"Fun fact: Conor has successfully generated a playable Minecraft-style tower-defense game prototype in just under 2 minutes using AI tools — a feat that seems straight out of the future!","strength":"His major strengths are a rare combination of technical skills and visionary thinking, enabling rapid prototyping and a deep understanding of technology potential. He is also adept at sharing his journey with authenticity and enthusiasm, which drives engagement.","weakness":"Conor can sometimes get too absorbed in technical challenges or AI limitations, which can lead to frustration and public airing of grievances, potentially alienating parts of his audience. Also, his passion for multiple domains might cause dilution of focus.","roast":"Conor’s idea of blending health & fitness with AI game dev makes you wonder — is he building the next Level-Up RPG or just trying to find a workout routine that'll generate extra XP? Either way, expect a lot of code pull-ups and mental push-ups!","win":"Conor’s breakthrough viral tweet about effortlessly generating advanced game prototypes with Gemini 3 grabbed massive attention, proving he’s not just dreaming about the future — he’s coding it live.","recommendation":"To grow his audience on X, Conor should lean into threading his experiments with storytelling and practical takeaways—making AI game dev relatable even for novices. Engaging with followers through polls or mini-challenges related to AI coding could also boost interaction and follower loyalty."},"created":1763117788092,"type":"the innovator","id":"conor_d_dart"},{"user":{"id":"1476511095837843458","name":"Natalia (mainnet arc) 🧬💜","description":"Co-founder MSSystem WEB3 | Mitosis maxi | Monad maxi and family builder | Mantle maven","followers_count":3536,"friends_count":3600,"statuses_count":95149,"profile_image_url_https":"https://pbs.twimg.com/profile_images/1941924931161260032/jmuRHgxl_normal.jpg","screen_name":"Natalia77351991","location":"","entities":{"description":{"urls":[]},"url":{"urls":[{"display_url":"mssystemweb3.com","expanded_url":"http://mssystemweb3.com","url":"https://t.co/owrGSxXjzM","indices":[0,23]}]}}},"details":{"type":"The Innovator","description":"Natalia is a dynamic WEB3 co-founder, deeply embedded in the crypto ecosystem and pioneering new decentralized finance applications. She’s constantly pushing boundaries with innovative projects like Monad Fortune Cookies and active involvement in cutting-edge blockchain systems. Always bullish and positive, Natalia combines technical savvy with a community-driven mindset.","purpose":"Natalia’s life purpose is to advance the WEB3 movement by creating and supporting innovative, user-friendly decentralized applications that empower loyal communities and drive mainstream adoption of blockchain technology.","beliefs":"She believes in transparency, decentralization, and the power of collaboration. Loyalty and real engagement are paramount to her, as she supports projects that weed out sybil attacks and reward authentic users. Innovation should always be practical, accessible, and community-oriented.","facts":"Fun fact: Natalia has tweeted over 95,000 times, demonstrating her relentless dedication and active participation in the crypto space, especially complementing her role as a Mitosis and Monad enthusiast and family builder.","strength":"Her greatest strength is her expertise in WEB3 ecosystems combined with a strong network presence, evident from her detailed updates and strategic promotional tweets. She’s a pro at community-building and driving engagement through digital campaigns and quests.","weakness":"With nearly 100k tweets, she risks overwhelming her audience with frequent content, which may dilute impact or fatigue followers. Also, her high following-to-follower ratio could signal a need to balance growth tactics for broader influence.","recommendation":"To grow her audience on X, Natalia should focus on crafting more bite-sized, visually engaging content and leverage video or thread formats to showcase innovation stories. Engaging directly with followers through AMAs or live discussions about her projects could boost genuine community engagement and conversion.","roast":"Natalia tweets so much, I’m starting to think she’s singlehandedly keeping the internet’s data centers open – does she sleep, or did someone clone her to run multiple nodes of relentless content? Maybe it’s time to moonlight as a crypto-powered bot! 🤖","win":"Successfully launched and continually upgraded the Monad Fortune Cookies dapp with cross-chain integration on major mainnets, alongside organizing community interactive quests and significant prize pools, cementing her reputation as a WEB3 innovator and community leader."},"created":1763116683184,"type":"the innovator","id":"natalia77351991"},{"user":{"id":"1196920441631313923","name":"Goldy","description":"Web3 Artist ✦ Content Creator\n\nJun mod: @SaharaLabsAI |\n\nAmbassador: @awenetwork_ai\n\nMaxi: @SentientAGI","followers_count":2091,"friends_count":1175,"statuses_count":15449,"profile_image_url_https":"https://pbs.twimg.com/profile_images/1967551785867915264/t6h4ujxr_normal.jpg","screen_name":"Golldyck","location":"","entities":{"description":{"urls":[]}}},"details":{"type":"The Innovator","description":"Goldy is a trailblazing Web3 artist and content creator who thrives at the intersection of art and cutting-edge AI technology. As a community moderator and ambassador, Goldy brings innovative projects like SentientAGI to life, engaging audiences with exciting updates and futuristic visions. With a prolific tweeting habit, Goldy fuels constant conversations around AI evolution, open-source intelligence, and creative collaboration.","purpose":"Goldy’s life purpose is to revolutionize the digital landscape by blending creativity with technology, fostering open-source communities, and driving the evolution of AI that is accessible and collaborative. They strive to build bridges between art and intelligence, empowering others to participate in the future of decentralized innovation.","beliefs":"Goldy believes in transparency, community-driven progress, and the power of open-source technology to democratize AI and digital creativity. They value collaboration, continuous learning, and the idea that technology should serve humanity’s collective growth, not just a privileged few.","facts":"Fun fact: Goldy doesn’t just create art — they also craft compelling narratives around AI projects, turning complex technical details into engaging stories that gather thousands of likes and replies. Their dedication shows in their 15,449 tweets, making them a prolific voice in the Web3 space.","strength":"Goldy’s strengths lie in their ability to communicate complex AI developments in an accessible and inspiring way, coupled with a consistent engagement strategy and deep community involvement. Their authentic enthusiasm for innovation fuels strong connections and thought leadership in the niche.","weakness":"Goldy’s high volume of content and multifaceted roles could risk audience fatigue or message dilution, where their diverse topics might overwhelm some followers. Also, engaging deeply on technical subjects can sometimes alienate casual followers who aren't as tech-savvy.","recommendation":"To grow their audience on X, Goldy should consider weaving more personal storytelling into their tech updates, making complex ideas relatable to a broader audience. Incorporating visual threads, AMAs, and collaboration with other influencers in AI and art can amplify reach and foster a more interactive, loyal community.","roast":"Goldy tweets like they’re trying to out-tweet a blockchain node—always running non-stop with so much technical jargon that even AI starts buffering. If there was a competition for ‘Most Passionate Over-Explainer of AI Sushi,’ Goldy would definitely have to take a break just to tweet about winning it.","win":"Goldy’s biggest win is cultivating a vibrant ecosystem around SentientAGI, successfully combining art, AI, and community to pioneer a new paradigm in open-source artificial intelligence—an achievement highlighted by viral tweets and active user engagement."},"created":1763111627930,"type":"the innovator","id":"golldyck"},{"user":{"id":"1130161200191303681","name":"el.cine","description":"Film director | AI Consultant | Partner with https://t.co/Vn9g3Z63CI Paris | Sharing practical ways to use AI for you and your business. All views are my own.","followers_count":87125,"friends_count":640,"statuses_count":33339,"profile_image_url_https":"https://pbs.twimg.com/profile_images/1793460098867838976/hoeoWt9M_normal.jpg","screen_name":"EHuanglu","location":"","entities":{"description":{"urls":[{"display_url":"irrational.tv","expanded_url":"http://irrational.tv","url":"https://t.co/Vn9g3Z63CI","indices":[45,68]}]},"url":{"urls":[{"display_url":"higgsfield.ai/@el_cine","expanded_url":"https://higgsfield.ai/@el_cine","url":"https://t.co/uX7TrWZBYf","indices":[0,23]}]}}},"details":{"type":"The Innovator","description":"el.cine is a trailblazing film director and AI consultant constantly at the forefront of technology’s intersection with creativity. With a robust presence in Paris and a passion for sharing practical AI insights, el.cine is all about disrupting the norms and introducing cutting-edge innovations to their audience. Their prolific tweeting habit ensures followers never miss a beat in the fast-evolving AI and film tech landscape.","purpose":"To pioneer the fusion of cinematic arts and artificial intelligence, transforming how stories are told and experienced by leveraging practical AI applications for creative and business growth.","beliefs":"el.cine values innovation, accessibility, and the pragmatic use of emerging technologies. They believe in empowering creators and businesses by demystifying AI and advocating for its ethical and practical implementation in everyday creative workflows.","facts":"Fun fact: el.cine doesn’t just talk about AI magic—they live it, sharing mind-blowing breakthroughs like NVIDIA’s AI supercomputer and AI tech that can change camera angles mid-movie, captivating over 24 million views on their top tweets!","strength":"Relentlessly curious and highly knowledgeable, el.cine excels in spotting groundbreaking AI trends and translating complex technologies into engaging, accessible content. Their deep blend of filmmaking experience and AI expertise creates a unique voice that inspires and educates simultaneously.","weakness":"The sheer volume of tweets (over 33,000) can make their feed overwhelming, potentially burying high-value insights in the noise. Additionally, a focus on rapid tech updates might leave less room for deeper, reflective discourse that builds long-term engagement.","recommendation":"To grow their audience on X, el.cine should curate their content by highlighting signature threads that unpack AI innovations with storytelling flair, engage more through interactive polls or Q&As, and collaborate with fellow creators and AI experts to amplify reach beyond just tech enthusiasts.","roast":"el.cine’s tweet count is so high, if there was an Olympic sport for tweeting, they’d have more golds than Michael Phelps. At this rate, they might just break the internet before AI does!","win":"Achieved widespread viral success by breaking down revolutionary AI film tech to millions, making cutting-edge concepts understandable and exciting while positioning themselves as a go-to expert at the crossroads of AI and cinema."},"created":1763110764940,"type":"the innovator","id":"ehuanglu"},{"user":{"id":"39632781","name":"Janaka Abeywardhana","description":"Building: https://t.co/jggwDjQ2li and @docqai. ex-@truelayer. ex-@hudl. Fullstack generalist, master of none. Product & Engineering. #GenAI #RAG #AIAgents #Infra.","followers_count":1414,"friends_count":2950,"statuses_count":13776,"profile_image_url_https":"https://pbs.twimg.com/profile_images/1816876128209870848/pv_gAoqY_normal.jpg","screen_name":"janaka_a","location":"London, England","entities":{"description":{"urls":[{"display_url":"easyrunner.xyz","expanded_url":"http://easyrunner.xyz","url":"https://t.co/jggwDjQ2li","indices":[10,33]}]},"url":{"urls":[{"display_url":"janaka.dev","expanded_url":"https://janaka.dev","url":"https://t.co/Bj1NHHCNWr","indices":[0,23]}]}}},"details":{"type":"The Innovator","description":"Janaka Abeywardhana is a dynamic fullstack generalist turned AI enthusiast, passionately building next-gen products and engineering solutions. Known for bridging product with deep tech, he thrives in the intersection of creativity and cutting-edge AI developments. His active engagement in tech communities highlights his drive to innovate and collaborate.","purpose":"To push the boundaries of technology by integrating AI, infrastructure, and product engineering, enabling smarter, more efficient solutions that empower developers and users alike.","beliefs":"Janaka values curiosity, continuous learning, and the democratization of AI tools. He believes that collaboration and community-driven innovation accelerate progress and open new horizons for problem-solving.","facts":"Fun fact: Despite calling himself a 'master of none,' Janaka effortlessly juggles product vision, engineering, and community engagement—proving that being a jack-of-all-trades can indeed be a masterstroke in the AI era!","strength":"Adaptability across multiple disciplines, a strong network in tech circles, and an authentic engagement with emerging AI technologies and communities.","weakness":"Spreading too thin across diverse interests can dilute focus, occasionally causing challenges in deep specialization or prioritization of projects.","recommendation":"To grow his audience on X, Janaka should leverage his community connections more by hosting regular AI-focused Twitter Spaces or AMAs, sharing exclusive behind-the-scenes insights on his projects, and consistently threading complex concepts into approachable, engaging stories tailored for AI enthusiasts and developers.","roast":"Janaka’s like a Swiss Army knife—loaded with tools but somehow still searching for the right attachment to really impress. Maybe it’s time to stop being the ‘master of none’ and pick a blade to shine (or at least tweet about it more convincingly).","win":"Successfully building and contributing to innovative platforms like DocQAI while maintaining a strong presence in the AI tinkerers community, reflecting his impact as both a creator and connector in the tech world."},"created":1763107906228,"type":"the innovator","id":"janaka_a"},{"user":{"id":"410253134","name":"Barbra.eth","description":"","followers_count":451,"friends_count":580,"statuses_count":17695,"profile_image_url_https":"https://pbs.twimg.com/profile_images/1969080317760978945/X7WUGbcc_normal.jpg","screen_name":"Barbra_Deley","location":"china","entities":{"description":{"urls":[]}}},"details":{"type":"The Innovator","description":"Barbra.eth is an insightful trailblazer deeply embedded in the evolving worlds of AI, blockchain, and DeFi, consistently placing herself at the nexus of emerging tech narratives. With nearly 18,000 tweets, she’s an active voice pushing boundaries and sharing real-time discoveries and analyses. Her engagement reveals a passion for verified truth, economic circularity, and technology’s tangible impact on finance and gaming.","purpose":"To pioneer and democratize verified innovation in decentralized finance and artificial intelligence, empowering communities to trust and monetize through transparent, data-driven ecosystems.","beliefs":"Barbra.eth values transparency, data integrity, and technological accountability, believing that the next economic frontier hinges on ‘verified truth’ and real asset integration. She champions community governance, trustless verification, and the seamless melding of AI with blockchain for scalable, meaningful change.","facts":"Fun fact: Barbra.eth doesn't just tweet about AI and DeFi—she’s hands-on, actively tracking projects like Kaito AI, Mira Network, and Wallchain, even sharing her own model-testing results with impressive accuracy! Plus, she treats weekend downtime as a myth—it's all hustle and innovation every day.","strength":"Her strengths lie in deep technical understanding combined with relentless energy for sharing nuanced insights, making complex topics accessible and sparking meaningful conversations. She's an early mover who connects dots others might overlook and has a knack for synthesizing cross-disciplinary innovations.","weakness":"Sometimes her deep dives and rapid-fire tweeting might overwhelm followers less versed in crypto jargon or technical detail, potentially narrowing her audience to mostly niche experts. Also, an intense focus on tech nuances can occasionally sideline broader engagement strategies.","recommendation":"To grow her audience on X, Barbra.eth should weave more storytelling and real-world impact narratives into her tech jargon, layering accessibility over expertise. Adding concise thread summaries, engaging with newcomers’ questions, and leveraging visual explainer content will broaden appeal without sacrificing depth.","roast":"Barbra.eth tweets so much tech detail, even her notifications need a PhD to keep up—maybe consider giving your keyboard a break before it files for early retirement!","win":"Successfully demystified and modeled a volatility prediction with 55% accuracy on Bitcoin swings using AlloraNetwork’s Builder Kit, showcasing practical prowess instead of just theory."},"created":1763106559688,"type":"the innovator","id":"barbra_deley"},{"user":{"id":"888293854960533504","name":"Thanh Doan","description":"software engineer. crafting impactful things to open source world | building overwrite: https://t.co/PCgG9ZSlbu | changelogs: https://t.co/SisBYPqOo0","followers_count":349,"friends_count":429,"statuses_count":4882,"profile_image_url_https":"https://pbs.twimg.com/profile_images/1961238073263751168/qrWgPrpN_normal.jpg","screen_name":"leodoan_","location":"Hồ Chí Minh, Việt Nam","entities":{"description":{"urls":[{"display_url":"mnismt.com/overwrite","expanded_url":"http://mnismt.com/overwrite","url":"https://t.co/PCgG9ZSlbu","indices":[88,111]},{"display_url":"changelogs.directory","expanded_url":"http://changelogs.directory","url":"https://t.co/SisBYPqOo0","indices":[126,149]}]},"url":{"urls":[{"display_url":"doantranminhthanh.com","expanded_url":"https://doantranminhthanh.com/","url":"https://t.co/v6xaz5R5dB","indices":[0,23]}]}}},"details":{"type":"The Innovator","description":"Thanh Doan is a passionate software engineer dedicated to crafting impactful tools for the open source community. Always exploring and experimenting, Thanh shares honest reviews and inventive solutions to streamline developers' workflows. Their innovations bridge AI and coding, making complex tasks more intuitive and efficient.","purpose":"To revolutionize developer experiences by creating innovative, open source tools that remove friction and enhance productivity, enabling others to build with ease and creativity.","beliefs":"Thanh believes in openness, collaboration, and continuous experimentation. They value transparency in sharing knowledge and are committed to leveraging AI thoughtfully to augment human creativity without overwhelming it.","facts":"Thanh built an open source alternative to RepoPrompt for VS Code that integrates AI chat and Git workflows with no context switching—a true game changer for developers.","strength":"Combining deep technical expertise with a creative mindset, Thanh excels at building practical yet pioneering developer tools and communicates their insights clearly and engagingly.","weakness":"An intense focus on innovation can sometimes lead to spreading efforts across multiple projects without consolidating a singular, standout platform to maximize impact and grow a cohesive audience.","roast":"Thanh’s love for building ‘bold, direct’ AI companions is impressive—until you realize their real AI assistant is just their coffee machine silently judging them for all those late-night coding sessions without a break.","win":"Successfully launching an open source VS Code extension that integrates AI-driven code review and streamlined Git workflows, gaining organic interest and valuable community feedback in a crowded space.","recommendation":"To grow your audience on X, lean into storytelling that highlights real-world impacts of your tools paired with bite-sized technical tutorials. Engage directly with developer communities by hosting AMAs or live coding demos to showcase your innovations in action."},"created":1763106335099,"type":"the innovator","id":"leodoan_"}],"activities":{"nreplies":[{"label":"2025-10-16","value":0,"startTime":1760486400000,"endTime":1760572800000,"tweets":[{"bookmarked":false,"display_text_range":[0,275],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1978585305679544427","view_count":90,"bookmark_count":0,"created_at":1760566471000,"favorite_count":4,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1978585305679544427","full_text":"Intelligence needs to be grown. Models need to develop coherent circuits that are reinforced. If you train models on traces they did not generate, there is no guarantee the information in that trace fits coherently in existing circuits. This is how models memorize instead of think for themselves. \n\nLearning should be on-policy. Take the unlikely but correct traces the model generates and reinforce those circuits. Then you build a model that thinks for itself.","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-17","value":0,"startTime":1760572800000,"endTime":1760659200000,"tweets":[{"bookmarked":false,"display_text_range":[73,86],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"500835474","name":"Pim de Witte","screen_name":"PimDeWitte","indices":[0,11]},{"id_str":"42226885","name":"Vinod Khosla","screen_name":"vkhosla","indices":[12,20]},{"id_str":"23809682","name":"Hemant Taneja","screen_name":"htaneja","indices":[21,29]},{"id_str":"1358786780259098625","name":"Eloi Alonso","screen_name":"EloiAlonso1","indices":[30,42]},{"id_str":"1307716628004704262","name":"Adam Jelley","screen_name":"AdamJelley2","indices":[43,55]},{"id_str":"1226783771195854848","name":"Vincent Micheli","screen_name":"micheli_vincent","indices":[56,72]}]},"favorited":false,"in_reply_to_screen_name":"PimDeWitte","lang":"en","retweeted":false,"fact_check":null,"id":"1978893031781568552","view_count":138,"bookmark_count":0,"created_at":1760639839000,"favorite_count":0,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1978825767053148457","full_text":"@PimDeWitte @vkhosla @htaneja @EloiAlonso1 @AdamJelley2 @micheli_vincent Congrats Pim!","in_reply_to_user_id_str":"500835474","in_reply_to_status_id_str":"1978825767053148457","is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-18","value":3,"startTime":1760659200000,"endTime":1760745600000,"tweets":[{"bookmarked":false,"display_text_range":[0,101],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","quoted_status_id_str":"1978196505392558087","quoted_status_permalink":{"url":"https://t.co/QhD1mEOFeS","expanded":"https://twitter.com/nicochristie/status/1978196505392558087","display":"x.com/nicochristie/s…"},"retweeted":false,"fact_check":null,"id":"1979230737405472922","view_count":1828,"bookmark_count":0,"created_at":1760720354000,"favorite_count":14,"quote_count":0,"reply_count":1,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1979230737405472922","full_text":"Accidentally said “impossible” instead of “out of distribution” and they kicked me out of my xAI tent","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,270],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","quoted_status_id_str":"1979237188220149938","quoted_status_permalink":{"url":"https://t.co/Ndv8TdVBSY","expanded":"https://twitter.com/MillionInt/status/1979237188220149938","display":"x.com/MillionInt/sta…"},"retweeted":false,"fact_check":null,"id":"1979240937281798378","view_count":730,"bookmark_count":5,"created_at":1760722786000,"favorite_count":12,"quote_count":0,"reply_count":2,"retweet_count":1,"user_id_str":"903008975343554566","conversation_id_str":"1979240937281798378","full_text":"Learning is like adding entropy to the neural network. Like adding a drop of colored liquid to clear water. The information diffuses across the weights. Unlearning that information is like undoing this diffusion. Reversing entropy. It takes a lot more energy to do that.","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null}]},{"label":"2025-10-19","value":0,"startTime":1760745600000,"endTime":1760832000000,"tweets":[]},{"label":"2025-10-20","value":0,"startTime":1760832000000,"endTime":1760918400000,"tweets":[]},{"label":"2025-10-21","value":1,"startTime":1760918400000,"endTime":1761004800000,"tweets":[{"bookmarked":false,"display_text_range":[0,110],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1980275836558213472","view_count":1446,"bookmark_count":2,"created_at":1760969525000,"favorite_count":20,"quote_count":0,"reply_count":1,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1980275836558213472","full_text":"To unlock the secrets of the universe, we just have to make interesting, true, and unlikely tokens more likely","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-22","value":0,"startTime":1761004800000,"endTime":1761091200000,"tweets":[]},{"label":"2025-10-23","value":0,"startTime":1761091200000,"endTime":1761177600000,"tweets":[]},{"label":"2025-10-24","value":0,"startTime":1761177600000,"endTime":1761264000000,"tweets":[]},{"label":"2025-10-25","value":0,"startTime":1761264000000,"endTime":1761350400000,"tweets":[{"bookmarked":false,"display_text_range":[17,70],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"4389158909","name":"Fernando Rojo","screen_name":"FernandoTheRojo","indices":[0,16]},{"id_str":"4389158909","name":"Fernando Rojo","screen_name":"FernandoTheRojo","indices":[17,33]}]},"favorited":false,"in_reply_to_screen_name":"FernandoTheRojo","lang":"en","retweeted":false,"fact_check":null,"id":"1981519914528821421","view_count":489,"bookmark_count":0,"created_at":1761266137000,"favorite_count":2,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1981509473295544350","full_text":"@FernandoTheRojo @FernandoTheRojo is the best app builder in the world","in_reply_to_user_id_str":"4389158909","in_reply_to_status_id_str":"1981509473295544350","is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-26","value":5,"startTime":1761350400000,"endTime":1761436800000,"tweets":[{"bookmarked":false,"display_text_range":[0,193],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1982191884593250452","view_count":2003,"bookmark_count":4,"created_at":1761426347000,"favorite_count":29,"quote_count":0,"reply_count":5,"retweet_count":1,"user_id_str":"903008975343554566","conversation_id_str":"1982191884593250452","full_text":"Solving hallucinations is a bigger deal than people think. Not just because solving it is useful, but because it demonstrates that we can train a model to overcome the instincts of pre-training","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,81],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1982190918326186171","view_count":685,"bookmark_count":0,"created_at":1761426116000,"favorite_count":1,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1982190918326186171","full_text":"A model using off-policy trained weights to respond is like a human under a spell","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-27","value":0,"startTime":1761436800000,"endTime":1761523200000,"tweets":[]},{"label":"2025-10-28","value":0,"startTime":1761523200000,"endTime":1761609600000,"tweets":[]},{"label":"2025-10-29","value":0,"startTime":1761609600000,"endTime":1761696000000,"tweets":[]},{"label":"2025-10-30","value":1,"startTime":1761696000000,"endTime":1761782400000,"tweets":[{"bookmarked":false,"display_text_range":[0,140],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1983562418782446066","view_count":1530,"bookmark_count":1,"created_at":1761753108000,"favorite_count":19,"quote_count":1,"reply_count":1,"retweet_count":1,"user_id_str":"903008975343554566","conversation_id_str":"1983562418782446066","full_text":"we use our continuous human brains to build symbolic computer systems to build continuous computer brains to build symbolic computer systems","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-31","value":0,"startTime":1761782400000,"endTime":1761868800000,"tweets":[]},{"label":"2025-11-01","value":0,"startTime":1761868800000,"endTime":1761955200000,"tweets":[]},{"label":"2025-11-02","value":0,"startTime":1761955200000,"endTime":1762041600000,"tweets":[{"bookmarked":false,"display_text_range":[0,44],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","quoted_status_id_str":"1984662921264120094","quoted_status_permalink":{"url":"https://t.co/s5M12jIb1I","expanded":"https://twitter.com/mindsai_jack/status/1984662921264120094","display":"x.com/mindsai_jack/s…"},"retweeted":false,"fact_check":null,"id":"1984665905972859177","view_count":2789,"bookmark_count":3,"created_at":1762016199000,"favorite_count":9,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1984665905972859177","full_text":"This was fun and Jack’s speech was great too","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null}]},{"label":"2025-11-03","value":0,"startTime":1762041600000,"endTime":1762128000000,"tweets":[]},{"label":"2025-11-04","value":0,"startTime":1762128000000,"endTime":1762214400000,"tweets":[]},{"label":"2025-11-05","value":3,"startTime":1762214400000,"endTime":1762300800000,"tweets":[{"bookmarked":false,"display_text_range":[0,275],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1985743643169178080","view_count":1463,"bookmark_count":4,"created_at":1762273152000,"favorite_count":13,"quote_count":0,"reply_count":3,"retweet_count":1,"user_id_str":"903008975343554566","conversation_id_str":"1985743643169178080","full_text":"Teaching language models to have taste isn't just about making them better at writing or making jokes.\n\nTaste is everything. It's what leads to scientific discovery.\n\nIn an infinite sea of things to reason about, taste is what guides you to reason about the right things — to combine the right ideas together to deduce something interesting and new.\n\nThe discovery of special relativity was a product of good taste.","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-11-06","value":0,"startTime":1762300800000,"endTime":1762387200000,"tweets":[]},{"label":"2025-11-07","value":0,"startTime":1762387200000,"endTime":1762473600000,"tweets":[{"bookmarked":false,"display_text_range":[0,96],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","quoted_status_id_str":"1986461871462096920","quoted_status_permalink":{"url":"https://t.co/J1VFIbBDHn","expanded":"https://twitter.com/skiminok/status/1986461871462096920","display":"x.com/skiminok/statu…"},"retweeted":false,"fact_check":null,"id":"1986474907484783018","view_count":1347,"bookmark_count":0,"created_at":1762447499000,"favorite_count":6,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1986474907484783018","full_text":"Never a dull moment talking language models with Alex. Really excited to get to do it more often","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null}]},{"label":"2025-11-08","value":0,"startTime":1762473600000,"endTime":1762560000000,"tweets":[]},{"label":"2025-11-09","value":0,"startTime":1762560000000,"endTime":1762646400000,"tweets":[]},{"label":"2025-11-10","value":0,"startTime":1762646400000,"endTime":1762732800000,"tweets":[]},{"label":"2025-11-11","value":78,"startTime":1762732800000,"endTime":1762819200000,"tweets":[{"bookmarked":false,"display_text_range":[0,224],"entities":{"hashtags":[],"media":[{"display_url":"pic.x.com/CGkAyp0vMN","expanded_url":"https://x.com/jerber888/status/1987982067116777521/photo/1","id_str":"1987981765265272832","indices":[225,248],"media_key":"3_1987981765265272832","media_url_https":"https://pbs.twimg.com/media/G5a64fqXMAAL8FC.jpg","type":"photo","url":"https://t.co/CGkAyp0vMN","ext_media_availability":{"status":"Available"},"features":{"large":{"faces":[]},"medium":{"faces":[]},"small":{"faces":[]},"orig":{"faces":[]}},"sizes":{"large":{"h":1154,"w":1956,"resize":"fit"},"medium":{"h":708,"w":1200,"resize":"fit"},"small":{"h":401,"w":680,"resize":"fit"},"thumb":{"h":150,"w":150,"resize":"crop"}},"original_info":{"height":1154,"width":1956,"focus_rects":[{"x":0,"y":0,"w":1956,"h":1095},{"x":802,"y":0,"w":1154,"h":1154},{"x":944,"y":0,"w":1012,"h":1154},{"x":1323,"y":0,"w":577,"h":1154},{"x":0,"y":0,"w":1956,"h":1154}]},"media_results":{"result":{"media_key":"3_1987981765265272832"}}}],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"extended_entities":{"media":[{"display_url":"pic.x.com/CGkAyp0vMN","expanded_url":"https://x.com/jerber888/status/1987982067116777521/photo/1","id_str":"1987981765265272832","indices":[225,248],"media_key":"3_1987981765265272832","media_url_https":"https://pbs.twimg.com/media/G5a64fqXMAAL8FC.jpg","type":"photo","url":"https://t.co/CGkAyp0vMN","ext_media_availability":{"status":"Available"},"features":{"large":{"faces":[]},"medium":{"faces":[]},"small":{"faces":[]},"orig":{"faces":[]}},"sizes":{"large":{"h":1154,"w":1956,"resize":"fit"},"medium":{"h":708,"w":1200,"resize":"fit"},"small":{"h":401,"w":680,"resize":"fit"},"thumb":{"h":150,"w":150,"resize":"crop"}},"original_info":{"height":1154,"width":1956,"focus_rects":[{"x":0,"y":0,"w":1956,"h":1095},{"x":802,"y":0,"w":1154,"h":1154},{"x":944,"y":0,"w":1012,"h":1154},{"x":1323,"y":0,"w":577,"h":1154},{"x":0,"y":0,"w":1956,"h":1154}]},"media_results":{"result":{"media_key":"3_1987981765265272832"}}}]},"favorited":false,"lang":"en","possibly_sensitive":false,"possibly_sensitive_editable":true,"quoted_status_id_str":"1968001933211471891","quoted_status_permalink":{"url":"https://t.co/l4P9D5a1TN","expanded":"https://twitter.com/jerber888/status/1968001933211471891","display":"x.com/jerber888/stat…"},"retweeted":false,"fact_check":null,"id":"1987982067116777521","view_count":269951,"bookmark_count":591,"created_at":1762806834000,"favorite_count":1579,"quote_count":29,"reply_count":70,"retweet_count":147,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"I finally reached human-level performance (85%) on ARC-AGI v1 for under $10k and within 12 hours. I use the same multi-agent collaboration with evolutionary test-time compute, now powered by GPT-5 pro with lower parallelism. https://t.co/CGkAyp0vMN","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,275],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"1773935160192647168","name":"ARC Prize","screen_name":"arcprize","indices":[168,177]},{"id_str":"1773935160192647168","name":"ARC Prize","screen_name":"arcprize","indices":[168,177]}]},"favorited":false,"in_reply_to_screen_name":"jerber888","lang":"en","retweeted":false,"fact_check":null,"id":"1987982069712945649","view_count":6572,"bookmark_count":6,"created_at":1762806834000,"favorite_count":57,"quote_count":0,"reply_count":4,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"I ran it on a random sample of 100 tasks from the 2024 eval set. It got 88/100 and averaged $27 per task. The score still needs to be verified on the hidden set by the @arcprize, but in my past submissions, the hidden set subtracts a few percent from the score and adds a few dollars. I put a box where I expect the official score to land within.","in_reply_to_user_id_str":"903008975343554566","in_reply_to_status_id_str":"1987982067116777521","is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,279],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"825088493764407298","name":"Noam Brown","screen_name":"polynoamial","indices":[288,300]}]},"favorited":false,"in_reply_to_screen_name":"jerber888","lang":"en","retweeted":false,"fact_check":null,"id":"1987982071541993700","view_count":6923,"bookmark_count":7,"created_at":1762806835000,"favorite_count":75,"quote_count":0,"reply_count":2,"retweet_count":2,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"GPT-5 pro is the best reasoning model today. It thinks coherently for hours and hours. My agent coordination logic is likely an intermediate step before the models learn to do this type of long horizon coordination on their own, end to end. It’s a hard problem but I wouldn’t bet against @polynoamial and team.","in_reply_to_user_id_str":"903008975343554566","in_reply_to_status_id_str":"1987982069712945649","is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,123],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[{"display_url":"github.com/jerber/arc-lan…","expanded_url":"https://github.com/jerber/arc-lang-public","url":"https://t.co/QnGjsN98ji","indices":[50,73]},{"display_url":"kaggle.com/code/jerber/je…","expanded_url":"https://www.kaggle.com/code/jerber/jeremy-arc2","url":"https://t.co/1jhzROG4SL","indices":[100,123]}],"user_mentions":[]},"favorited":false,"in_reply_to_screen_name":"jerber888","lang":"en","possibly_sensitive":false,"possibly_sensitive_editable":true,"retweeted":false,"fact_check":null,"id":"1987982214332580197","view_count":6525,"bookmark_count":39,"created_at":1762806869000,"favorite_count":64,"quote_count":1,"reply_count":2,"retweet_count":2,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"You can run this code yourself, it’s open source: https://t.co/QnGjsN98ji. And the kaggle notebook: https://t.co/1jhzROG4SL","in_reply_to_user_id_str":"903008975343554566","in_reply_to_status_id_str":"1987982071541993700","is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-11-12","value":2,"startTime":1762819200000,"endTime":1762905600000,"tweets":[{"bookmarked":false,"display_text_range":[0,48],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1988382853965770765","view_count":576,"bookmark_count":1,"created_at":1762902389000,"favorite_count":16,"quote_count":0,"reply_count":1,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1988382853965770765","full_text":"be the change you want to see in the world model","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,52],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[{"display_url":"youtube.com/watch?v=FcnLiP…","expanded_url":"https://www.youtube.com/watch?v=FcnLiPyfRZM","url":"https://t.co/jEBQGMsm62","indices":[29,52]}],"user_mentions":[]},"favorited":false,"lang":"en","possibly_sensitive":false,"possibly_sensitive_editable":true,"quoted_status_id_str":"1988165006073229659","quoted_status_permalink":{"url":"https://t.co/l99V6sHyI3","expanded":"https://twitter.com/MLStreetTalk/status/1988165006073229659","display":"x.com/MLStreetTalk/s…"},"retweeted":false,"fact_check":null,"id":"1988280687200530841","view_count":4997,"bookmark_count":5,"created_at":1762878030000,"favorite_count":21,"quote_count":0,"reply_count":1,"retweet_count":2,"user_id_str":"903008975343554566","conversation_id_str":"1988280687200530841","full_text":"This was a fun conversation: https://t.co/jEBQGMsm62","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,56],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[{"display_url":"jeremyberman.substack.com/p/how-i-got-th…","expanded_url":"https://jeremyberman.substack.com/p/how-i-got-the-highest-score-on-arc-agi-again","url":"https://t.co/xWiPB2Z0aI","indices":[33,56]}],"user_mentions":[]},"favorited":false,"in_reply_to_screen_name":"jerber888","lang":"en","possibly_sensitive":false,"possibly_sensitive_editable":true,"retweeted":false,"fact_check":null,"id":"1988278615377916103","view_count":1461,"bookmark_count":6,"created_at":1762877536000,"favorite_count":9,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"And technical blog post is here: https://t.co/xWiPB2Z0aI","in_reply_to_user_id_str":"903008975343554566","in_reply_to_status_id_str":"1987982214332580197","is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[9,52],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"536428919","name":"JFPuget 🇺🇦🇨🇦🇬🇱","screen_name":"JFPuget","indices":[0,8]}]},"favorited":false,"in_reply_to_screen_name":"JFPuget","lang":"en","retweeted":false,"fact_check":null,"id":"1988253927104471544","view_count":1040,"bookmark_count":0,"created_at":1762871650000,"favorite_count":8,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"@JFPuget Most likely but was focused on v1 for this.","in_reply_to_user_id_str":"536428919","in_reply_to_status_id_str":"1988249984907571356","is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[13,26],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"1194889317388374016","name":"Matt Shumer","screen_name":"mattshumer_","indices":[0,12]}]},"favorited":false,"in_reply_to_screen_name":"mattshumer_","lang":"en","retweeted":false,"fact_check":null,"id":"1988334412690571395","view_count":221,"bookmark_count":0,"created_at":1762890840000,"favorite_count":3,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1988331100935254226","full_text":"@mattshumer_ thanks Matt 🙏","in_reply_to_user_id_str":"1194889317388374016","in_reply_to_status_id_str":"1988331100935254226","is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-11-13","value":0,"startTime":1762905600000,"endTime":1762992000000,"tweets":[]},{"label":"2025-11-14","value":0,"startTime":1762992000000,"endTime":1763078400000,"tweets":[]},{"label":"2025-11-15","value":0,"startTime":1763078400000,"endTime":1763164800000,"tweets":[]}],"nbookmarks":[{"label":"2025-10-16","value":0,"startTime":1760486400000,"endTime":1760572800000,"tweets":[{"bookmarked":false,"display_text_range":[0,275],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1978585305679544427","view_count":90,"bookmark_count":0,"created_at":1760566471000,"favorite_count":4,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1978585305679544427","full_text":"Intelligence needs to be grown. Models need to develop coherent circuits that are reinforced. If you train models on traces they did not generate, there is no guarantee the information in that trace fits coherently in existing circuits. This is how models memorize instead of think for themselves. \n\nLearning should be on-policy. Take the unlikely but correct traces the model generates and reinforce those circuits. Then you build a model that thinks for itself.","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-17","value":0,"startTime":1760572800000,"endTime":1760659200000,"tweets":[{"bookmarked":false,"display_text_range":[73,86],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"500835474","name":"Pim de Witte","screen_name":"PimDeWitte","indices":[0,11]},{"id_str":"42226885","name":"Vinod Khosla","screen_name":"vkhosla","indices":[12,20]},{"id_str":"23809682","name":"Hemant Taneja","screen_name":"htaneja","indices":[21,29]},{"id_str":"1358786780259098625","name":"Eloi Alonso","screen_name":"EloiAlonso1","indices":[30,42]},{"id_str":"1307716628004704262","name":"Adam Jelley","screen_name":"AdamJelley2","indices":[43,55]},{"id_str":"1226783771195854848","name":"Vincent Micheli","screen_name":"micheli_vincent","indices":[56,72]}]},"favorited":false,"in_reply_to_screen_name":"PimDeWitte","lang":"en","retweeted":false,"fact_check":null,"id":"1978893031781568552","view_count":138,"bookmark_count":0,"created_at":1760639839000,"favorite_count":0,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1978825767053148457","full_text":"@PimDeWitte @vkhosla @htaneja @EloiAlonso1 @AdamJelley2 @micheli_vincent Congrats Pim!","in_reply_to_user_id_str":"500835474","in_reply_to_status_id_str":"1978825767053148457","is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-18","value":5,"startTime":1760659200000,"endTime":1760745600000,"tweets":[{"bookmarked":false,"display_text_range":[0,101],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","quoted_status_id_str":"1978196505392558087","quoted_status_permalink":{"url":"https://t.co/QhD1mEOFeS","expanded":"https://twitter.com/nicochristie/status/1978196505392558087","display":"x.com/nicochristie/s…"},"retweeted":false,"fact_check":null,"id":"1979230737405472922","view_count":1828,"bookmark_count":0,"created_at":1760720354000,"favorite_count":14,"quote_count":0,"reply_count":1,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1979230737405472922","full_text":"Accidentally said “impossible” instead of “out of distribution” and they kicked me out of my xAI tent","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,270],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","quoted_status_id_str":"1979237188220149938","quoted_status_permalink":{"url":"https://t.co/Ndv8TdVBSY","expanded":"https://twitter.com/MillionInt/status/1979237188220149938","display":"x.com/MillionInt/sta…"},"retweeted":false,"fact_check":null,"id":"1979240937281798378","view_count":730,"bookmark_count":5,"created_at":1760722786000,"favorite_count":12,"quote_count":0,"reply_count":2,"retweet_count":1,"user_id_str":"903008975343554566","conversation_id_str":"1979240937281798378","full_text":"Learning is like adding entropy to the neural network. Like adding a drop of colored liquid to clear water. The information diffuses across the weights. Unlearning that information is like undoing this diffusion. Reversing entropy. It takes a lot more energy to do that.","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null}]},{"label":"2025-10-19","value":0,"startTime":1760745600000,"endTime":1760832000000,"tweets":[]},{"label":"2025-10-20","value":0,"startTime":1760832000000,"endTime":1760918400000,"tweets":[]},{"label":"2025-10-21","value":2,"startTime":1760918400000,"endTime":1761004800000,"tweets":[{"bookmarked":false,"display_text_range":[0,110],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1980275836558213472","view_count":1446,"bookmark_count":2,"created_at":1760969525000,"favorite_count":20,"quote_count":0,"reply_count":1,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1980275836558213472","full_text":"To unlock the secrets of the universe, we just have to make interesting, true, and unlikely tokens more likely","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-22","value":0,"startTime":1761004800000,"endTime":1761091200000,"tweets":[]},{"label":"2025-10-23","value":0,"startTime":1761091200000,"endTime":1761177600000,"tweets":[]},{"label":"2025-10-24","value":0,"startTime":1761177600000,"endTime":1761264000000,"tweets":[]},{"label":"2025-10-25","value":0,"startTime":1761264000000,"endTime":1761350400000,"tweets":[{"bookmarked":false,"display_text_range":[17,70],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"4389158909","name":"Fernando Rojo","screen_name":"FernandoTheRojo","indices":[0,16]},{"id_str":"4389158909","name":"Fernando Rojo","screen_name":"FernandoTheRojo","indices":[17,33]}]},"favorited":false,"in_reply_to_screen_name":"FernandoTheRojo","lang":"en","retweeted":false,"fact_check":null,"id":"1981519914528821421","view_count":489,"bookmark_count":0,"created_at":1761266137000,"favorite_count":2,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1981509473295544350","full_text":"@FernandoTheRojo @FernandoTheRojo is the best app builder in the world","in_reply_to_user_id_str":"4389158909","in_reply_to_status_id_str":"1981509473295544350","is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-26","value":4,"startTime":1761350400000,"endTime":1761436800000,"tweets":[{"bookmarked":false,"display_text_range":[0,193],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1982191884593250452","view_count":2003,"bookmark_count":4,"created_at":1761426347000,"favorite_count":29,"quote_count":0,"reply_count":5,"retweet_count":1,"user_id_str":"903008975343554566","conversation_id_str":"1982191884593250452","full_text":"Solving hallucinations is a bigger deal than people think. Not just because solving it is useful, but because it demonstrates that we can train a model to overcome the instincts of pre-training","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,81],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1982190918326186171","view_count":685,"bookmark_count":0,"created_at":1761426116000,"favorite_count":1,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1982190918326186171","full_text":"A model using off-policy trained weights to respond is like a human under a spell","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-27","value":0,"startTime":1761436800000,"endTime":1761523200000,"tweets":[]},{"label":"2025-10-28","value":0,"startTime":1761523200000,"endTime":1761609600000,"tweets":[]},{"label":"2025-10-29","value":0,"startTime":1761609600000,"endTime":1761696000000,"tweets":[]},{"label":"2025-10-30","value":1,"startTime":1761696000000,"endTime":1761782400000,"tweets":[{"bookmarked":false,"display_text_range":[0,140],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1983562418782446066","view_count":1530,"bookmark_count":1,"created_at":1761753108000,"favorite_count":19,"quote_count":1,"reply_count":1,"retweet_count":1,"user_id_str":"903008975343554566","conversation_id_str":"1983562418782446066","full_text":"we use our continuous human brains to build symbolic computer systems to build continuous computer brains to build symbolic computer systems","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-31","value":0,"startTime":1761782400000,"endTime":1761868800000,"tweets":[]},{"label":"2025-11-01","value":0,"startTime":1761868800000,"endTime":1761955200000,"tweets":[]},{"label":"2025-11-02","value":3,"startTime":1761955200000,"endTime":1762041600000,"tweets":[{"bookmarked":false,"display_text_range":[0,44],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","quoted_status_id_str":"1984662921264120094","quoted_status_permalink":{"url":"https://t.co/s5M12jIb1I","expanded":"https://twitter.com/mindsai_jack/status/1984662921264120094","display":"x.com/mindsai_jack/s…"},"retweeted":false,"fact_check":null,"id":"1984665905972859177","view_count":2789,"bookmark_count":3,"created_at":1762016199000,"favorite_count":9,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1984665905972859177","full_text":"This was fun and Jack’s speech was great too","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null}]},{"label":"2025-11-03","value":0,"startTime":1762041600000,"endTime":1762128000000,"tweets":[]},{"label":"2025-11-04","value":0,"startTime":1762128000000,"endTime":1762214400000,"tweets":[]},{"label":"2025-11-05","value":4,"startTime":1762214400000,"endTime":1762300800000,"tweets":[{"bookmarked":false,"display_text_range":[0,275],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1985743643169178080","view_count":1463,"bookmark_count":4,"created_at":1762273152000,"favorite_count":13,"quote_count":0,"reply_count":3,"retweet_count":1,"user_id_str":"903008975343554566","conversation_id_str":"1985743643169178080","full_text":"Teaching language models to have taste isn't just about making them better at writing or making jokes.\n\nTaste is everything. It's what leads to scientific discovery.\n\nIn an infinite sea of things to reason about, taste is what guides you to reason about the right things — to combine the right ideas together to deduce something interesting and new.\n\nThe discovery of special relativity was a product of good taste.","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-11-06","value":0,"startTime":1762300800000,"endTime":1762387200000,"tweets":[]},{"label":"2025-11-07","value":0,"startTime":1762387200000,"endTime":1762473600000,"tweets":[{"bookmarked":false,"display_text_range":[0,96],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","quoted_status_id_str":"1986461871462096920","quoted_status_permalink":{"url":"https://t.co/J1VFIbBDHn","expanded":"https://twitter.com/skiminok/status/1986461871462096920","display":"x.com/skiminok/statu…"},"retweeted":false,"fact_check":null,"id":"1986474907484783018","view_count":1347,"bookmark_count":0,"created_at":1762447499000,"favorite_count":6,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1986474907484783018","full_text":"Never a dull moment talking language models with Alex. Really excited to get to do it more often","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null}]},{"label":"2025-11-08","value":0,"startTime":1762473600000,"endTime":1762560000000,"tweets":[]},{"label":"2025-11-09","value":0,"startTime":1762560000000,"endTime":1762646400000,"tweets":[]},{"label":"2025-11-10","value":0,"startTime":1762646400000,"endTime":1762732800000,"tweets":[]},{"label":"2025-11-11","value":643,"startTime":1762732800000,"endTime":1762819200000,"tweets":[{"bookmarked":false,"display_text_range":[0,224],"entities":{"hashtags":[],"media":[{"display_url":"pic.x.com/CGkAyp0vMN","expanded_url":"https://x.com/jerber888/status/1987982067116777521/photo/1","id_str":"1987981765265272832","indices":[225,248],"media_key":"3_1987981765265272832","media_url_https":"https://pbs.twimg.com/media/G5a64fqXMAAL8FC.jpg","type":"photo","url":"https://t.co/CGkAyp0vMN","ext_media_availability":{"status":"Available"},"features":{"large":{"faces":[]},"medium":{"faces":[]},"small":{"faces":[]},"orig":{"faces":[]}},"sizes":{"large":{"h":1154,"w":1956,"resize":"fit"},"medium":{"h":708,"w":1200,"resize":"fit"},"small":{"h":401,"w":680,"resize":"fit"},"thumb":{"h":150,"w":150,"resize":"crop"}},"original_info":{"height":1154,"width":1956,"focus_rects":[{"x":0,"y":0,"w":1956,"h":1095},{"x":802,"y":0,"w":1154,"h":1154},{"x":944,"y":0,"w":1012,"h":1154},{"x":1323,"y":0,"w":577,"h":1154},{"x":0,"y":0,"w":1956,"h":1154}]},"media_results":{"result":{"media_key":"3_1987981765265272832"}}}],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"extended_entities":{"media":[{"display_url":"pic.x.com/CGkAyp0vMN","expanded_url":"https://x.com/jerber888/status/1987982067116777521/photo/1","id_str":"1987981765265272832","indices":[225,248],"media_key":"3_1987981765265272832","media_url_https":"https://pbs.twimg.com/media/G5a64fqXMAAL8FC.jpg","type":"photo","url":"https://t.co/CGkAyp0vMN","ext_media_availability":{"status":"Available"},"features":{"large":{"faces":[]},"medium":{"faces":[]},"small":{"faces":[]},"orig":{"faces":[]}},"sizes":{"large":{"h":1154,"w":1956,"resize":"fit"},"medium":{"h":708,"w":1200,"resize":"fit"},"small":{"h":401,"w":680,"resize":"fit"},"thumb":{"h":150,"w":150,"resize":"crop"}},"original_info":{"height":1154,"width":1956,"focus_rects":[{"x":0,"y":0,"w":1956,"h":1095},{"x":802,"y":0,"w":1154,"h":1154},{"x":944,"y":0,"w":1012,"h":1154},{"x":1323,"y":0,"w":577,"h":1154},{"x":0,"y":0,"w":1956,"h":1154}]},"media_results":{"result":{"media_key":"3_1987981765265272832"}}}]},"favorited":false,"lang":"en","possibly_sensitive":false,"possibly_sensitive_editable":true,"quoted_status_id_str":"1968001933211471891","quoted_status_permalink":{"url":"https://t.co/l4P9D5a1TN","expanded":"https://twitter.com/jerber888/status/1968001933211471891","display":"x.com/jerber888/stat…"},"retweeted":false,"fact_check":null,"id":"1987982067116777521","view_count":269951,"bookmark_count":591,"created_at":1762806834000,"favorite_count":1579,"quote_count":29,"reply_count":70,"retweet_count":147,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"I finally reached human-level performance (85%) on ARC-AGI v1 for under $10k and within 12 hours. I use the same multi-agent collaboration with evolutionary test-time compute, now powered by GPT-5 pro with lower parallelism. https://t.co/CGkAyp0vMN","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,275],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"1773935160192647168","name":"ARC Prize","screen_name":"arcprize","indices":[168,177]},{"id_str":"1773935160192647168","name":"ARC Prize","screen_name":"arcprize","indices":[168,177]}]},"favorited":false,"in_reply_to_screen_name":"jerber888","lang":"en","retweeted":false,"fact_check":null,"id":"1987982069712945649","view_count":6572,"bookmark_count":6,"created_at":1762806834000,"favorite_count":57,"quote_count":0,"reply_count":4,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"I ran it on a random sample of 100 tasks from the 2024 eval set. It got 88/100 and averaged $27 per task. The score still needs to be verified on the hidden set by the @arcprize, but in my past submissions, the hidden set subtracts a few percent from the score and adds a few dollars. I put a box where I expect the official score to land within.","in_reply_to_user_id_str":"903008975343554566","in_reply_to_status_id_str":"1987982067116777521","is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,279],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"825088493764407298","name":"Noam Brown","screen_name":"polynoamial","indices":[288,300]}]},"favorited":false,"in_reply_to_screen_name":"jerber888","lang":"en","retweeted":false,"fact_check":null,"id":"1987982071541993700","view_count":6923,"bookmark_count":7,"created_at":1762806835000,"favorite_count":75,"quote_count":0,"reply_count":2,"retweet_count":2,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"GPT-5 pro is the best reasoning model today. It thinks coherently for hours and hours. My agent coordination logic is likely an intermediate step before the models learn to do this type of long horizon coordination on their own, end to end. It’s a hard problem but I wouldn’t bet against @polynoamial and team.","in_reply_to_user_id_str":"903008975343554566","in_reply_to_status_id_str":"1987982069712945649","is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,123],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[{"display_url":"github.com/jerber/arc-lan…","expanded_url":"https://github.com/jerber/arc-lang-public","url":"https://t.co/QnGjsN98ji","indices":[50,73]},{"display_url":"kaggle.com/code/jerber/je…","expanded_url":"https://www.kaggle.com/code/jerber/jeremy-arc2","url":"https://t.co/1jhzROG4SL","indices":[100,123]}],"user_mentions":[]},"favorited":false,"in_reply_to_screen_name":"jerber888","lang":"en","possibly_sensitive":false,"possibly_sensitive_editable":true,"retweeted":false,"fact_check":null,"id":"1987982214332580197","view_count":6525,"bookmark_count":39,"created_at":1762806869000,"favorite_count":64,"quote_count":1,"reply_count":2,"retweet_count":2,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"You can run this code yourself, it’s open source: https://t.co/QnGjsN98ji. And the kaggle notebook: https://t.co/1jhzROG4SL","in_reply_to_user_id_str":"903008975343554566","in_reply_to_status_id_str":"1987982071541993700","is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-11-12","value":12,"startTime":1762819200000,"endTime":1762905600000,"tweets":[{"bookmarked":false,"display_text_range":[0,48],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1988382853965770765","view_count":576,"bookmark_count":1,"created_at":1762902389000,"favorite_count":16,"quote_count":0,"reply_count":1,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1988382853965770765","full_text":"be the change you want to see in the world model","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,52],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[{"display_url":"youtube.com/watch?v=FcnLiP…","expanded_url":"https://www.youtube.com/watch?v=FcnLiPyfRZM","url":"https://t.co/jEBQGMsm62","indices":[29,52]}],"user_mentions":[]},"favorited":false,"lang":"en","possibly_sensitive":false,"possibly_sensitive_editable":true,"quoted_status_id_str":"1988165006073229659","quoted_status_permalink":{"url":"https://t.co/l99V6sHyI3","expanded":"https://twitter.com/MLStreetTalk/status/1988165006073229659","display":"x.com/MLStreetTalk/s…"},"retweeted":false,"fact_check":null,"id":"1988280687200530841","view_count":4997,"bookmark_count":5,"created_at":1762878030000,"favorite_count":21,"quote_count":0,"reply_count":1,"retweet_count":2,"user_id_str":"903008975343554566","conversation_id_str":"1988280687200530841","full_text":"This was a fun conversation: https://t.co/jEBQGMsm62","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,56],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[{"display_url":"jeremyberman.substack.com/p/how-i-got-th…","expanded_url":"https://jeremyberman.substack.com/p/how-i-got-the-highest-score-on-arc-agi-again","url":"https://t.co/xWiPB2Z0aI","indices":[33,56]}],"user_mentions":[]},"favorited":false,"in_reply_to_screen_name":"jerber888","lang":"en","possibly_sensitive":false,"possibly_sensitive_editable":true,"retweeted":false,"fact_check":null,"id":"1988278615377916103","view_count":1461,"bookmark_count":6,"created_at":1762877536000,"favorite_count":9,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"And technical blog post is here: https://t.co/xWiPB2Z0aI","in_reply_to_user_id_str":"903008975343554566","in_reply_to_status_id_str":"1987982214332580197","is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[9,52],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"536428919","name":"JFPuget 🇺🇦🇨🇦🇬🇱","screen_name":"JFPuget","indices":[0,8]}]},"favorited":false,"in_reply_to_screen_name":"JFPuget","lang":"en","retweeted":false,"fact_check":null,"id":"1988253927104471544","view_count":1040,"bookmark_count":0,"created_at":1762871650000,"favorite_count":8,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"@JFPuget Most likely but was focused on v1 for this.","in_reply_to_user_id_str":"536428919","in_reply_to_status_id_str":"1988249984907571356","is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[13,26],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"1194889317388374016","name":"Matt Shumer","screen_name":"mattshumer_","indices":[0,12]}]},"favorited":false,"in_reply_to_screen_name":"mattshumer_","lang":"en","retweeted":false,"fact_check":null,"id":"1988334412690571395","view_count":221,"bookmark_count":0,"created_at":1762890840000,"favorite_count":3,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1988331100935254226","full_text":"@mattshumer_ thanks Matt 🙏","in_reply_to_user_id_str":"1194889317388374016","in_reply_to_status_id_str":"1988331100935254226","is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-11-13","value":0,"startTime":1762905600000,"endTime":1762992000000,"tweets":[]},{"label":"2025-11-14","value":0,"startTime":1762992000000,"endTime":1763078400000,"tweets":[]},{"label":"2025-11-15","value":0,"startTime":1763078400000,"endTime":1763164800000,"tweets":[]}],"nretweets":[{"label":"2025-10-16","value":0,"startTime":1760486400000,"endTime":1760572800000,"tweets":[{"bookmarked":false,"display_text_range":[0,275],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1978585305679544427","view_count":90,"bookmark_count":0,"created_at":1760566471000,"favorite_count":4,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1978585305679544427","full_text":"Intelligence needs to be grown. Models need to develop coherent circuits that are reinforced. If you train models on traces they did not generate, there is no guarantee the information in that trace fits coherently in existing circuits. This is how models memorize instead of think for themselves. \n\nLearning should be on-policy. Take the unlikely but correct traces the model generates and reinforce those circuits. Then you build a model that thinks for itself.","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-17","value":0,"startTime":1760572800000,"endTime":1760659200000,"tweets":[{"bookmarked":false,"display_text_range":[73,86],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"500835474","name":"Pim de Witte","screen_name":"PimDeWitte","indices":[0,11]},{"id_str":"42226885","name":"Vinod Khosla","screen_name":"vkhosla","indices":[12,20]},{"id_str":"23809682","name":"Hemant Taneja","screen_name":"htaneja","indices":[21,29]},{"id_str":"1358786780259098625","name":"Eloi Alonso","screen_name":"EloiAlonso1","indices":[30,42]},{"id_str":"1307716628004704262","name":"Adam Jelley","screen_name":"AdamJelley2","indices":[43,55]},{"id_str":"1226783771195854848","name":"Vincent Micheli","screen_name":"micheli_vincent","indices":[56,72]}]},"favorited":false,"in_reply_to_screen_name":"PimDeWitte","lang":"en","retweeted":false,"fact_check":null,"id":"1978893031781568552","view_count":138,"bookmark_count":0,"created_at":1760639839000,"favorite_count":0,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1978825767053148457","full_text":"@PimDeWitte @vkhosla @htaneja @EloiAlonso1 @AdamJelley2 @micheli_vincent Congrats Pim!","in_reply_to_user_id_str":"500835474","in_reply_to_status_id_str":"1978825767053148457","is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-18","value":1,"startTime":1760659200000,"endTime":1760745600000,"tweets":[{"bookmarked":false,"display_text_range":[0,101],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","quoted_status_id_str":"1978196505392558087","quoted_status_permalink":{"url":"https://t.co/QhD1mEOFeS","expanded":"https://twitter.com/nicochristie/status/1978196505392558087","display":"x.com/nicochristie/s…"},"retweeted":false,"fact_check":null,"id":"1979230737405472922","view_count":1828,"bookmark_count":0,"created_at":1760720354000,"favorite_count":14,"quote_count":0,"reply_count":1,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1979230737405472922","full_text":"Accidentally said “impossible” instead of “out of distribution” and they kicked me out of my xAI tent","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,270],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","quoted_status_id_str":"1979237188220149938","quoted_status_permalink":{"url":"https://t.co/Ndv8TdVBSY","expanded":"https://twitter.com/MillionInt/status/1979237188220149938","display":"x.com/MillionInt/sta…"},"retweeted":false,"fact_check":null,"id":"1979240937281798378","view_count":730,"bookmark_count":5,"created_at":1760722786000,"favorite_count":12,"quote_count":0,"reply_count":2,"retweet_count":1,"user_id_str":"903008975343554566","conversation_id_str":"1979240937281798378","full_text":"Learning is like adding entropy to the neural network. Like adding a drop of colored liquid to clear water. The information diffuses across the weights. Unlearning that information is like undoing this diffusion. Reversing entropy. It takes a lot more energy to do that.","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null}]},{"label":"2025-10-19","value":0,"startTime":1760745600000,"endTime":1760832000000,"tweets":[]},{"label":"2025-10-20","value":0,"startTime":1760832000000,"endTime":1760918400000,"tweets":[]},{"label":"2025-10-21","value":0,"startTime":1760918400000,"endTime":1761004800000,"tweets":[{"bookmarked":false,"display_text_range":[0,110],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1980275836558213472","view_count":1446,"bookmark_count":2,"created_at":1760969525000,"favorite_count":20,"quote_count":0,"reply_count":1,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1980275836558213472","full_text":"To unlock the secrets of the universe, we just have to make interesting, true, and unlikely tokens more likely","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-22","value":0,"startTime":1761004800000,"endTime":1761091200000,"tweets":[]},{"label":"2025-10-23","value":0,"startTime":1761091200000,"endTime":1761177600000,"tweets":[]},{"label":"2025-10-24","value":0,"startTime":1761177600000,"endTime":1761264000000,"tweets":[]},{"label":"2025-10-25","value":0,"startTime":1761264000000,"endTime":1761350400000,"tweets":[{"bookmarked":false,"display_text_range":[17,70],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"4389158909","name":"Fernando Rojo","screen_name":"FernandoTheRojo","indices":[0,16]},{"id_str":"4389158909","name":"Fernando Rojo","screen_name":"FernandoTheRojo","indices":[17,33]}]},"favorited":false,"in_reply_to_screen_name":"FernandoTheRojo","lang":"en","retweeted":false,"fact_check":null,"id":"1981519914528821421","view_count":489,"bookmark_count":0,"created_at":1761266137000,"favorite_count":2,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1981509473295544350","full_text":"@FernandoTheRojo @FernandoTheRojo is the best app builder in the world","in_reply_to_user_id_str":"4389158909","in_reply_to_status_id_str":"1981509473295544350","is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-26","value":1,"startTime":1761350400000,"endTime":1761436800000,"tweets":[{"bookmarked":false,"display_text_range":[0,193],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1982191884593250452","view_count":2003,"bookmark_count":4,"created_at":1761426347000,"favorite_count":29,"quote_count":0,"reply_count":5,"retweet_count":1,"user_id_str":"903008975343554566","conversation_id_str":"1982191884593250452","full_text":"Solving hallucinations is a bigger deal than people think. Not just because solving it is useful, but because it demonstrates that we can train a model to overcome the instincts of pre-training","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,81],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1982190918326186171","view_count":685,"bookmark_count":0,"created_at":1761426116000,"favorite_count":1,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1982190918326186171","full_text":"A model using off-policy trained weights to respond is like a human under a spell","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-27","value":0,"startTime":1761436800000,"endTime":1761523200000,"tweets":[]},{"label":"2025-10-28","value":0,"startTime":1761523200000,"endTime":1761609600000,"tweets":[]},{"label":"2025-10-29","value":0,"startTime":1761609600000,"endTime":1761696000000,"tweets":[]},{"label":"2025-10-30","value":1,"startTime":1761696000000,"endTime":1761782400000,"tweets":[{"bookmarked":false,"display_text_range":[0,140],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1983562418782446066","view_count":1530,"bookmark_count":1,"created_at":1761753108000,"favorite_count":19,"quote_count":1,"reply_count":1,"retweet_count":1,"user_id_str":"903008975343554566","conversation_id_str":"1983562418782446066","full_text":"we use our continuous human brains to build symbolic computer systems to build continuous computer brains to build symbolic computer systems","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-31","value":0,"startTime":1761782400000,"endTime":1761868800000,"tweets":[]},{"label":"2025-11-01","value":0,"startTime":1761868800000,"endTime":1761955200000,"tweets":[]},{"label":"2025-11-02","value":0,"startTime":1761955200000,"endTime":1762041600000,"tweets":[{"bookmarked":false,"display_text_range":[0,44],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","quoted_status_id_str":"1984662921264120094","quoted_status_permalink":{"url":"https://t.co/s5M12jIb1I","expanded":"https://twitter.com/mindsai_jack/status/1984662921264120094","display":"x.com/mindsai_jack/s…"},"retweeted":false,"fact_check":null,"id":"1984665905972859177","view_count":2789,"bookmark_count":3,"created_at":1762016199000,"favorite_count":9,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1984665905972859177","full_text":"This was fun and Jack’s speech was great too","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null}]},{"label":"2025-11-03","value":0,"startTime":1762041600000,"endTime":1762128000000,"tweets":[]},{"label":"2025-11-04","value":0,"startTime":1762128000000,"endTime":1762214400000,"tweets":[]},{"label":"2025-11-05","value":1,"startTime":1762214400000,"endTime":1762300800000,"tweets":[{"bookmarked":false,"display_text_range":[0,275],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1985743643169178080","view_count":1463,"bookmark_count":4,"created_at":1762273152000,"favorite_count":13,"quote_count":0,"reply_count":3,"retweet_count":1,"user_id_str":"903008975343554566","conversation_id_str":"1985743643169178080","full_text":"Teaching language models to have taste isn't just about making them better at writing or making jokes.\n\nTaste is everything. It's what leads to scientific discovery.\n\nIn an infinite sea of things to reason about, taste is what guides you to reason about the right things — to combine the right ideas together to deduce something interesting and new.\n\nThe discovery of special relativity was a product of good taste.","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-11-06","value":0,"startTime":1762300800000,"endTime":1762387200000,"tweets":[]},{"label":"2025-11-07","value":0,"startTime":1762387200000,"endTime":1762473600000,"tweets":[{"bookmarked":false,"display_text_range":[0,96],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","quoted_status_id_str":"1986461871462096920","quoted_status_permalink":{"url":"https://t.co/J1VFIbBDHn","expanded":"https://twitter.com/skiminok/status/1986461871462096920","display":"x.com/skiminok/statu…"},"retweeted":false,"fact_check":null,"id":"1986474907484783018","view_count":1347,"bookmark_count":0,"created_at":1762447499000,"favorite_count":6,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1986474907484783018","full_text":"Never a dull moment talking language models with Alex. Really excited to get to do it more often","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null}]},{"label":"2025-11-08","value":0,"startTime":1762473600000,"endTime":1762560000000,"tweets":[]},{"label":"2025-11-09","value":0,"startTime":1762560000000,"endTime":1762646400000,"tweets":[]},{"label":"2025-11-10","value":0,"startTime":1762646400000,"endTime":1762732800000,"tweets":[]},{"label":"2025-11-11","value":151,"startTime":1762732800000,"endTime":1762819200000,"tweets":[{"bookmarked":false,"display_text_range":[0,224],"entities":{"hashtags":[],"media":[{"display_url":"pic.x.com/CGkAyp0vMN","expanded_url":"https://x.com/jerber888/status/1987982067116777521/photo/1","id_str":"1987981765265272832","indices":[225,248],"media_key":"3_1987981765265272832","media_url_https":"https://pbs.twimg.com/media/G5a64fqXMAAL8FC.jpg","type":"photo","url":"https://t.co/CGkAyp0vMN","ext_media_availability":{"status":"Available"},"features":{"large":{"faces":[]},"medium":{"faces":[]},"small":{"faces":[]},"orig":{"faces":[]}},"sizes":{"large":{"h":1154,"w":1956,"resize":"fit"},"medium":{"h":708,"w":1200,"resize":"fit"},"small":{"h":401,"w":680,"resize":"fit"},"thumb":{"h":150,"w":150,"resize":"crop"}},"original_info":{"height":1154,"width":1956,"focus_rects":[{"x":0,"y":0,"w":1956,"h":1095},{"x":802,"y":0,"w":1154,"h":1154},{"x":944,"y":0,"w":1012,"h":1154},{"x":1323,"y":0,"w":577,"h":1154},{"x":0,"y":0,"w":1956,"h":1154}]},"media_results":{"result":{"media_key":"3_1987981765265272832"}}}],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"extended_entities":{"media":[{"display_url":"pic.x.com/CGkAyp0vMN","expanded_url":"https://x.com/jerber888/status/1987982067116777521/photo/1","id_str":"1987981765265272832","indices":[225,248],"media_key":"3_1987981765265272832","media_url_https":"https://pbs.twimg.com/media/G5a64fqXMAAL8FC.jpg","type":"photo","url":"https://t.co/CGkAyp0vMN","ext_media_availability":{"status":"Available"},"features":{"large":{"faces":[]},"medium":{"faces":[]},"small":{"faces":[]},"orig":{"faces":[]}},"sizes":{"large":{"h":1154,"w":1956,"resize":"fit"},"medium":{"h":708,"w":1200,"resize":"fit"},"small":{"h":401,"w":680,"resize":"fit"},"thumb":{"h":150,"w":150,"resize":"crop"}},"original_info":{"height":1154,"width":1956,"focus_rects":[{"x":0,"y":0,"w":1956,"h":1095},{"x":802,"y":0,"w":1154,"h":1154},{"x":944,"y":0,"w":1012,"h":1154},{"x":1323,"y":0,"w":577,"h":1154},{"x":0,"y":0,"w":1956,"h":1154}]},"media_results":{"result":{"media_key":"3_1987981765265272832"}}}]},"favorited":false,"lang":"en","possibly_sensitive":false,"possibly_sensitive_editable":true,"quoted_status_id_str":"1968001933211471891","quoted_status_permalink":{"url":"https://t.co/l4P9D5a1TN","expanded":"https://twitter.com/jerber888/status/1968001933211471891","display":"x.com/jerber888/stat…"},"retweeted":false,"fact_check":null,"id":"1987982067116777521","view_count":269951,"bookmark_count":591,"created_at":1762806834000,"favorite_count":1579,"quote_count":29,"reply_count":70,"retweet_count":147,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"I finally reached human-level performance (85%) on ARC-AGI v1 for under $10k and within 12 hours. I use the same multi-agent collaboration with evolutionary test-time compute, now powered by GPT-5 pro with lower parallelism. https://t.co/CGkAyp0vMN","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,275],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"1773935160192647168","name":"ARC Prize","screen_name":"arcprize","indices":[168,177]},{"id_str":"1773935160192647168","name":"ARC Prize","screen_name":"arcprize","indices":[168,177]}]},"favorited":false,"in_reply_to_screen_name":"jerber888","lang":"en","retweeted":false,"fact_check":null,"id":"1987982069712945649","view_count":6572,"bookmark_count":6,"created_at":1762806834000,"favorite_count":57,"quote_count":0,"reply_count":4,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"I ran it on a random sample of 100 tasks from the 2024 eval set. It got 88/100 and averaged $27 per task. The score still needs to be verified on the hidden set by the @arcprize, but in my past submissions, the hidden set subtracts a few percent from the score and adds a few dollars. I put a box where I expect the official score to land within.","in_reply_to_user_id_str":"903008975343554566","in_reply_to_status_id_str":"1987982067116777521","is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,279],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"825088493764407298","name":"Noam Brown","screen_name":"polynoamial","indices":[288,300]}]},"favorited":false,"in_reply_to_screen_name":"jerber888","lang":"en","retweeted":false,"fact_check":null,"id":"1987982071541993700","view_count":6923,"bookmark_count":7,"created_at":1762806835000,"favorite_count":75,"quote_count":0,"reply_count":2,"retweet_count":2,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"GPT-5 pro is the best reasoning model today. It thinks coherently for hours and hours. My agent coordination logic is likely an intermediate step before the models learn to do this type of long horizon coordination on their own, end to end. It’s a hard problem but I wouldn’t bet against @polynoamial and team.","in_reply_to_user_id_str":"903008975343554566","in_reply_to_status_id_str":"1987982069712945649","is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,123],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[{"display_url":"github.com/jerber/arc-lan…","expanded_url":"https://github.com/jerber/arc-lang-public","url":"https://t.co/QnGjsN98ji","indices":[50,73]},{"display_url":"kaggle.com/code/jerber/je…","expanded_url":"https://www.kaggle.com/code/jerber/jeremy-arc2","url":"https://t.co/1jhzROG4SL","indices":[100,123]}],"user_mentions":[]},"favorited":false,"in_reply_to_screen_name":"jerber888","lang":"en","possibly_sensitive":false,"possibly_sensitive_editable":true,"retweeted":false,"fact_check":null,"id":"1987982214332580197","view_count":6525,"bookmark_count":39,"created_at":1762806869000,"favorite_count":64,"quote_count":1,"reply_count":2,"retweet_count":2,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"You can run this code yourself, it’s open source: https://t.co/QnGjsN98ji. And the kaggle notebook: https://t.co/1jhzROG4SL","in_reply_to_user_id_str":"903008975343554566","in_reply_to_status_id_str":"1987982071541993700","is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-11-12","value":2,"startTime":1762819200000,"endTime":1762905600000,"tweets":[{"bookmarked":false,"display_text_range":[0,48],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1988382853965770765","view_count":576,"bookmark_count":1,"created_at":1762902389000,"favorite_count":16,"quote_count":0,"reply_count":1,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1988382853965770765","full_text":"be the change you want to see in the world model","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,52],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[{"display_url":"youtube.com/watch?v=FcnLiP…","expanded_url":"https://www.youtube.com/watch?v=FcnLiPyfRZM","url":"https://t.co/jEBQGMsm62","indices":[29,52]}],"user_mentions":[]},"favorited":false,"lang":"en","possibly_sensitive":false,"possibly_sensitive_editable":true,"quoted_status_id_str":"1988165006073229659","quoted_status_permalink":{"url":"https://t.co/l99V6sHyI3","expanded":"https://twitter.com/MLStreetTalk/status/1988165006073229659","display":"x.com/MLStreetTalk/s…"},"retweeted":false,"fact_check":null,"id":"1988280687200530841","view_count":4997,"bookmark_count":5,"created_at":1762878030000,"favorite_count":21,"quote_count":0,"reply_count":1,"retweet_count":2,"user_id_str":"903008975343554566","conversation_id_str":"1988280687200530841","full_text":"This was a fun conversation: https://t.co/jEBQGMsm62","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,56],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[{"display_url":"jeremyberman.substack.com/p/how-i-got-th…","expanded_url":"https://jeremyberman.substack.com/p/how-i-got-the-highest-score-on-arc-agi-again","url":"https://t.co/xWiPB2Z0aI","indices":[33,56]}],"user_mentions":[]},"favorited":false,"in_reply_to_screen_name":"jerber888","lang":"en","possibly_sensitive":false,"possibly_sensitive_editable":true,"retweeted":false,"fact_check":null,"id":"1988278615377916103","view_count":1461,"bookmark_count":6,"created_at":1762877536000,"favorite_count":9,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"And technical blog post is here: https://t.co/xWiPB2Z0aI","in_reply_to_user_id_str":"903008975343554566","in_reply_to_status_id_str":"1987982214332580197","is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[9,52],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"536428919","name":"JFPuget 🇺🇦🇨🇦🇬🇱","screen_name":"JFPuget","indices":[0,8]}]},"favorited":false,"in_reply_to_screen_name":"JFPuget","lang":"en","retweeted":false,"fact_check":null,"id":"1988253927104471544","view_count":1040,"bookmark_count":0,"created_at":1762871650000,"favorite_count":8,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"@JFPuget Most likely but was focused on v1 for this.","in_reply_to_user_id_str":"536428919","in_reply_to_status_id_str":"1988249984907571356","is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[13,26],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"1194889317388374016","name":"Matt Shumer","screen_name":"mattshumer_","indices":[0,12]}]},"favorited":false,"in_reply_to_screen_name":"mattshumer_","lang":"en","retweeted":false,"fact_check":null,"id":"1988334412690571395","view_count":221,"bookmark_count":0,"created_at":1762890840000,"favorite_count":3,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1988331100935254226","full_text":"@mattshumer_ thanks Matt 🙏","in_reply_to_user_id_str":"1194889317388374016","in_reply_to_status_id_str":"1988331100935254226","is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-11-13","value":0,"startTime":1762905600000,"endTime":1762992000000,"tweets":[]},{"label":"2025-11-14","value":0,"startTime":1762992000000,"endTime":1763078400000,"tweets":[]},{"label":"2025-11-15","value":0,"startTime":1763078400000,"endTime":1763164800000,"tweets":[]}],"nlikes":[{"label":"2025-10-16","value":4,"startTime":1760486400000,"endTime":1760572800000,"tweets":[{"bookmarked":false,"display_text_range":[0,275],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1978585305679544427","view_count":90,"bookmark_count":0,"created_at":1760566471000,"favorite_count":4,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1978585305679544427","full_text":"Intelligence needs to be grown. Models need to develop coherent circuits that are reinforced. If you train models on traces they did not generate, there is no guarantee the information in that trace fits coherently in existing circuits. This is how models memorize instead of think for themselves. \n\nLearning should be on-policy. Take the unlikely but correct traces the model generates and reinforce those circuits. Then you build a model that thinks for itself.","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-17","value":0,"startTime":1760572800000,"endTime":1760659200000,"tweets":[{"bookmarked":false,"display_text_range":[73,86],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"500835474","name":"Pim de Witte","screen_name":"PimDeWitte","indices":[0,11]},{"id_str":"42226885","name":"Vinod Khosla","screen_name":"vkhosla","indices":[12,20]},{"id_str":"23809682","name":"Hemant Taneja","screen_name":"htaneja","indices":[21,29]},{"id_str":"1358786780259098625","name":"Eloi Alonso","screen_name":"EloiAlonso1","indices":[30,42]},{"id_str":"1307716628004704262","name":"Adam Jelley","screen_name":"AdamJelley2","indices":[43,55]},{"id_str":"1226783771195854848","name":"Vincent Micheli","screen_name":"micheli_vincent","indices":[56,72]}]},"favorited":false,"in_reply_to_screen_name":"PimDeWitte","lang":"en","retweeted":false,"fact_check":null,"id":"1978893031781568552","view_count":138,"bookmark_count":0,"created_at":1760639839000,"favorite_count":0,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1978825767053148457","full_text":"@PimDeWitte @vkhosla @htaneja @EloiAlonso1 @AdamJelley2 @micheli_vincent Congrats Pim!","in_reply_to_user_id_str":"500835474","in_reply_to_status_id_str":"1978825767053148457","is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-18","value":26,"startTime":1760659200000,"endTime":1760745600000,"tweets":[{"bookmarked":false,"display_text_range":[0,101],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","quoted_status_id_str":"1978196505392558087","quoted_status_permalink":{"url":"https://t.co/QhD1mEOFeS","expanded":"https://twitter.com/nicochristie/status/1978196505392558087","display":"x.com/nicochristie/s…"},"retweeted":false,"fact_check":null,"id":"1979230737405472922","view_count":1828,"bookmark_count":0,"created_at":1760720354000,"favorite_count":14,"quote_count":0,"reply_count":1,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1979230737405472922","full_text":"Accidentally said “impossible” instead of “out of distribution” and they kicked me out of my xAI tent","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,270],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","quoted_status_id_str":"1979237188220149938","quoted_status_permalink":{"url":"https://t.co/Ndv8TdVBSY","expanded":"https://twitter.com/MillionInt/status/1979237188220149938","display":"x.com/MillionInt/sta…"},"retweeted":false,"fact_check":null,"id":"1979240937281798378","view_count":730,"bookmark_count":5,"created_at":1760722786000,"favorite_count":12,"quote_count":0,"reply_count":2,"retweet_count":1,"user_id_str":"903008975343554566","conversation_id_str":"1979240937281798378","full_text":"Learning is like adding entropy to the neural network. Like adding a drop of colored liquid to clear water. The information diffuses across the weights. Unlearning that information is like undoing this diffusion. Reversing entropy. It takes a lot more energy to do that.","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null}]},{"label":"2025-10-19","value":0,"startTime":1760745600000,"endTime":1760832000000,"tweets":[]},{"label":"2025-10-20","value":0,"startTime":1760832000000,"endTime":1760918400000,"tweets":[]},{"label":"2025-10-21","value":20,"startTime":1760918400000,"endTime":1761004800000,"tweets":[{"bookmarked":false,"display_text_range":[0,110],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1980275836558213472","view_count":1446,"bookmark_count":2,"created_at":1760969525000,"favorite_count":20,"quote_count":0,"reply_count":1,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1980275836558213472","full_text":"To unlock the secrets of the universe, we just have to make interesting, true, and unlikely tokens more likely","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-22","value":0,"startTime":1761004800000,"endTime":1761091200000,"tweets":[]},{"label":"2025-10-23","value":0,"startTime":1761091200000,"endTime":1761177600000,"tweets":[]},{"label":"2025-10-24","value":0,"startTime":1761177600000,"endTime":1761264000000,"tweets":[]},{"label":"2025-10-25","value":2,"startTime":1761264000000,"endTime":1761350400000,"tweets":[{"bookmarked":false,"display_text_range":[17,70],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"4389158909","name":"Fernando Rojo","screen_name":"FernandoTheRojo","indices":[0,16]},{"id_str":"4389158909","name":"Fernando Rojo","screen_name":"FernandoTheRojo","indices":[17,33]}]},"favorited":false,"in_reply_to_screen_name":"FernandoTheRojo","lang":"en","retweeted":false,"fact_check":null,"id":"1981519914528821421","view_count":489,"bookmark_count":0,"created_at":1761266137000,"favorite_count":2,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1981509473295544350","full_text":"@FernandoTheRojo @FernandoTheRojo is the best app builder in the world","in_reply_to_user_id_str":"4389158909","in_reply_to_status_id_str":"1981509473295544350","is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-26","value":30,"startTime":1761350400000,"endTime":1761436800000,"tweets":[{"bookmarked":false,"display_text_range":[0,193],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1982191884593250452","view_count":2003,"bookmark_count":4,"created_at":1761426347000,"favorite_count":29,"quote_count":0,"reply_count":5,"retweet_count":1,"user_id_str":"903008975343554566","conversation_id_str":"1982191884593250452","full_text":"Solving hallucinations is a bigger deal than people think. Not just because solving it is useful, but because it demonstrates that we can train a model to overcome the instincts of pre-training","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,81],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1982190918326186171","view_count":685,"bookmark_count":0,"created_at":1761426116000,"favorite_count":1,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1982190918326186171","full_text":"A model using off-policy trained weights to respond is like a human under a spell","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-27","value":0,"startTime":1761436800000,"endTime":1761523200000,"tweets":[]},{"label":"2025-10-28","value":0,"startTime":1761523200000,"endTime":1761609600000,"tweets":[]},{"label":"2025-10-29","value":0,"startTime":1761609600000,"endTime":1761696000000,"tweets":[]},{"label":"2025-10-30","value":19,"startTime":1761696000000,"endTime":1761782400000,"tweets":[{"bookmarked":false,"display_text_range":[0,140],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1983562418782446066","view_count":1530,"bookmark_count":1,"created_at":1761753108000,"favorite_count":19,"quote_count":1,"reply_count":1,"retweet_count":1,"user_id_str":"903008975343554566","conversation_id_str":"1983562418782446066","full_text":"we use our continuous human brains to build symbolic computer systems to build continuous computer brains to build symbolic computer systems","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-31","value":0,"startTime":1761782400000,"endTime":1761868800000,"tweets":[]},{"label":"2025-11-01","value":0,"startTime":1761868800000,"endTime":1761955200000,"tweets":[]},{"label":"2025-11-02","value":9,"startTime":1761955200000,"endTime":1762041600000,"tweets":[{"bookmarked":false,"display_text_range":[0,44],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","quoted_status_id_str":"1984662921264120094","quoted_status_permalink":{"url":"https://t.co/s5M12jIb1I","expanded":"https://twitter.com/mindsai_jack/status/1984662921264120094","display":"x.com/mindsai_jack/s…"},"retweeted":false,"fact_check":null,"id":"1984665905972859177","view_count":2789,"bookmark_count":3,"created_at":1762016199000,"favorite_count":9,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1984665905972859177","full_text":"This was fun and Jack’s speech was great too","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null}]},{"label":"2025-11-03","value":0,"startTime":1762041600000,"endTime":1762128000000,"tweets":[]},{"label":"2025-11-04","value":0,"startTime":1762128000000,"endTime":1762214400000,"tweets":[]},{"label":"2025-11-05","value":13,"startTime":1762214400000,"endTime":1762300800000,"tweets":[{"bookmarked":false,"display_text_range":[0,275],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1985743643169178080","view_count":1463,"bookmark_count":4,"created_at":1762273152000,"favorite_count":13,"quote_count":0,"reply_count":3,"retweet_count":1,"user_id_str":"903008975343554566","conversation_id_str":"1985743643169178080","full_text":"Teaching language models to have taste isn't just about making them better at writing or making jokes.\n\nTaste is everything. It's what leads to scientific discovery.\n\nIn an infinite sea of things to reason about, taste is what guides you to reason about the right things — to combine the right ideas together to deduce something interesting and new.\n\nThe discovery of special relativity was a product of good taste.","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-11-06","value":0,"startTime":1762300800000,"endTime":1762387200000,"tweets":[]},{"label":"2025-11-07","value":6,"startTime":1762387200000,"endTime":1762473600000,"tweets":[{"bookmarked":false,"display_text_range":[0,96],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","quoted_status_id_str":"1986461871462096920","quoted_status_permalink":{"url":"https://t.co/J1VFIbBDHn","expanded":"https://twitter.com/skiminok/status/1986461871462096920","display":"x.com/skiminok/statu…"},"retweeted":false,"fact_check":null,"id":"1986474907484783018","view_count":1347,"bookmark_count":0,"created_at":1762447499000,"favorite_count":6,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1986474907484783018","full_text":"Never a dull moment talking language models with Alex. Really excited to get to do it more often","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null}]},{"label":"2025-11-08","value":0,"startTime":1762473600000,"endTime":1762560000000,"tweets":[]},{"label":"2025-11-09","value":0,"startTime":1762560000000,"endTime":1762646400000,"tweets":[]},{"label":"2025-11-10","value":0,"startTime":1762646400000,"endTime":1762732800000,"tweets":[]},{"label":"2025-11-11","value":1775,"startTime":1762732800000,"endTime":1762819200000,"tweets":[{"bookmarked":false,"display_text_range":[0,224],"entities":{"hashtags":[],"media":[{"display_url":"pic.x.com/CGkAyp0vMN","expanded_url":"https://x.com/jerber888/status/1987982067116777521/photo/1","id_str":"1987981765265272832","indices":[225,248],"media_key":"3_1987981765265272832","media_url_https":"https://pbs.twimg.com/media/G5a64fqXMAAL8FC.jpg","type":"photo","url":"https://t.co/CGkAyp0vMN","ext_media_availability":{"status":"Available"},"features":{"large":{"faces":[]},"medium":{"faces":[]},"small":{"faces":[]},"orig":{"faces":[]}},"sizes":{"large":{"h":1154,"w":1956,"resize":"fit"},"medium":{"h":708,"w":1200,"resize":"fit"},"small":{"h":401,"w":680,"resize":"fit"},"thumb":{"h":150,"w":150,"resize":"crop"}},"original_info":{"height":1154,"width":1956,"focus_rects":[{"x":0,"y":0,"w":1956,"h":1095},{"x":802,"y":0,"w":1154,"h":1154},{"x":944,"y":0,"w":1012,"h":1154},{"x":1323,"y":0,"w":577,"h":1154},{"x":0,"y":0,"w":1956,"h":1154}]},"media_results":{"result":{"media_key":"3_1987981765265272832"}}}],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"extended_entities":{"media":[{"display_url":"pic.x.com/CGkAyp0vMN","expanded_url":"https://x.com/jerber888/status/1987982067116777521/photo/1","id_str":"1987981765265272832","indices":[225,248],"media_key":"3_1987981765265272832","media_url_https":"https://pbs.twimg.com/media/G5a64fqXMAAL8FC.jpg","type":"photo","url":"https://t.co/CGkAyp0vMN","ext_media_availability":{"status":"Available"},"features":{"large":{"faces":[]},"medium":{"faces":[]},"small":{"faces":[]},"orig":{"faces":[]}},"sizes":{"large":{"h":1154,"w":1956,"resize":"fit"},"medium":{"h":708,"w":1200,"resize":"fit"},"small":{"h":401,"w":680,"resize":"fit"},"thumb":{"h":150,"w":150,"resize":"crop"}},"original_info":{"height":1154,"width":1956,"focus_rects":[{"x":0,"y":0,"w":1956,"h":1095},{"x":802,"y":0,"w":1154,"h":1154},{"x":944,"y":0,"w":1012,"h":1154},{"x":1323,"y":0,"w":577,"h":1154},{"x":0,"y":0,"w":1956,"h":1154}]},"media_results":{"result":{"media_key":"3_1987981765265272832"}}}]},"favorited":false,"lang":"en","possibly_sensitive":false,"possibly_sensitive_editable":true,"quoted_status_id_str":"1968001933211471891","quoted_status_permalink":{"url":"https://t.co/l4P9D5a1TN","expanded":"https://twitter.com/jerber888/status/1968001933211471891","display":"x.com/jerber888/stat…"},"retweeted":false,"fact_check":null,"id":"1987982067116777521","view_count":269951,"bookmark_count":591,"created_at":1762806834000,"favorite_count":1579,"quote_count":29,"reply_count":70,"retweet_count":147,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"I finally reached human-level performance (85%) on ARC-AGI v1 for under $10k and within 12 hours. I use the same multi-agent collaboration with evolutionary test-time compute, now powered by GPT-5 pro with lower parallelism. https://t.co/CGkAyp0vMN","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,275],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"1773935160192647168","name":"ARC Prize","screen_name":"arcprize","indices":[168,177]},{"id_str":"1773935160192647168","name":"ARC Prize","screen_name":"arcprize","indices":[168,177]}]},"favorited":false,"in_reply_to_screen_name":"jerber888","lang":"en","retweeted":false,"fact_check":null,"id":"1987982069712945649","view_count":6572,"bookmark_count":6,"created_at":1762806834000,"favorite_count":57,"quote_count":0,"reply_count":4,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"I ran it on a random sample of 100 tasks from the 2024 eval set. It got 88/100 and averaged $27 per task. The score still needs to be verified on the hidden set by the @arcprize, but in my past submissions, the hidden set subtracts a few percent from the score and adds a few dollars. I put a box where I expect the official score to land within.","in_reply_to_user_id_str":"903008975343554566","in_reply_to_status_id_str":"1987982067116777521","is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,279],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"825088493764407298","name":"Noam Brown","screen_name":"polynoamial","indices":[288,300]}]},"favorited":false,"in_reply_to_screen_name":"jerber888","lang":"en","retweeted":false,"fact_check":null,"id":"1987982071541993700","view_count":6923,"bookmark_count":7,"created_at":1762806835000,"favorite_count":75,"quote_count":0,"reply_count":2,"retweet_count":2,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"GPT-5 pro is the best reasoning model today. It thinks coherently for hours and hours. My agent coordination logic is likely an intermediate step before the models learn to do this type of long horizon coordination on their own, end to end. It’s a hard problem but I wouldn’t bet against @polynoamial and team.","in_reply_to_user_id_str":"903008975343554566","in_reply_to_status_id_str":"1987982069712945649","is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,123],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[{"display_url":"github.com/jerber/arc-lan…","expanded_url":"https://github.com/jerber/arc-lang-public","url":"https://t.co/QnGjsN98ji","indices":[50,73]},{"display_url":"kaggle.com/code/jerber/je…","expanded_url":"https://www.kaggle.com/code/jerber/jeremy-arc2","url":"https://t.co/1jhzROG4SL","indices":[100,123]}],"user_mentions":[]},"favorited":false,"in_reply_to_screen_name":"jerber888","lang":"en","possibly_sensitive":false,"possibly_sensitive_editable":true,"retweeted":false,"fact_check":null,"id":"1987982214332580197","view_count":6525,"bookmark_count":39,"created_at":1762806869000,"favorite_count":64,"quote_count":1,"reply_count":2,"retweet_count":2,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"You can run this code yourself, it’s open source: https://t.co/QnGjsN98ji. And the kaggle notebook: https://t.co/1jhzROG4SL","in_reply_to_user_id_str":"903008975343554566","in_reply_to_status_id_str":"1987982071541993700","is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-11-12","value":57,"startTime":1762819200000,"endTime":1762905600000,"tweets":[{"bookmarked":false,"display_text_range":[0,48],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1988382853965770765","view_count":576,"bookmark_count":1,"created_at":1762902389000,"favorite_count":16,"quote_count":0,"reply_count":1,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1988382853965770765","full_text":"be the change you want to see in the world model","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,52],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[{"display_url":"youtube.com/watch?v=FcnLiP…","expanded_url":"https://www.youtube.com/watch?v=FcnLiPyfRZM","url":"https://t.co/jEBQGMsm62","indices":[29,52]}],"user_mentions":[]},"favorited":false,"lang":"en","possibly_sensitive":false,"possibly_sensitive_editable":true,"quoted_status_id_str":"1988165006073229659","quoted_status_permalink":{"url":"https://t.co/l99V6sHyI3","expanded":"https://twitter.com/MLStreetTalk/status/1988165006073229659","display":"x.com/MLStreetTalk/s…"},"retweeted":false,"fact_check":null,"id":"1988280687200530841","view_count":4997,"bookmark_count":5,"created_at":1762878030000,"favorite_count":21,"quote_count":0,"reply_count":1,"retweet_count":2,"user_id_str":"903008975343554566","conversation_id_str":"1988280687200530841","full_text":"This was a fun conversation: https://t.co/jEBQGMsm62","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,56],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[{"display_url":"jeremyberman.substack.com/p/how-i-got-th…","expanded_url":"https://jeremyberman.substack.com/p/how-i-got-the-highest-score-on-arc-agi-again","url":"https://t.co/xWiPB2Z0aI","indices":[33,56]}],"user_mentions":[]},"favorited":false,"in_reply_to_screen_name":"jerber888","lang":"en","possibly_sensitive":false,"possibly_sensitive_editable":true,"retweeted":false,"fact_check":null,"id":"1988278615377916103","view_count":1461,"bookmark_count":6,"created_at":1762877536000,"favorite_count":9,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"And technical blog post is here: https://t.co/xWiPB2Z0aI","in_reply_to_user_id_str":"903008975343554566","in_reply_to_status_id_str":"1987982214332580197","is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[9,52],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"536428919","name":"JFPuget 🇺🇦🇨🇦🇬🇱","screen_name":"JFPuget","indices":[0,8]}]},"favorited":false,"in_reply_to_screen_name":"JFPuget","lang":"en","retweeted":false,"fact_check":null,"id":"1988253927104471544","view_count":1040,"bookmark_count":0,"created_at":1762871650000,"favorite_count":8,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"@JFPuget Most likely but was focused on v1 for this.","in_reply_to_user_id_str":"536428919","in_reply_to_status_id_str":"1988249984907571356","is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[13,26],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"1194889317388374016","name":"Matt Shumer","screen_name":"mattshumer_","indices":[0,12]}]},"favorited":false,"in_reply_to_screen_name":"mattshumer_","lang":"en","retweeted":false,"fact_check":null,"id":"1988334412690571395","view_count":221,"bookmark_count":0,"created_at":1762890840000,"favorite_count":3,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1988331100935254226","full_text":"@mattshumer_ thanks Matt 🙏","in_reply_to_user_id_str":"1194889317388374016","in_reply_to_status_id_str":"1988331100935254226","is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-11-13","value":0,"startTime":1762905600000,"endTime":1762992000000,"tweets":[]},{"label":"2025-11-14","value":0,"startTime":1762992000000,"endTime":1763078400000,"tweets":[]},{"label":"2025-11-15","value":0,"startTime":1763078400000,"endTime":1763164800000,"tweets":[]}],"nviews":[{"label":"2025-10-16","value":90,"startTime":1760486400000,"endTime":1760572800000,"tweets":[{"bookmarked":false,"display_text_range":[0,275],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1978585305679544427","view_count":90,"bookmark_count":0,"created_at":1760566471000,"favorite_count":4,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1978585305679544427","full_text":"Intelligence needs to be grown. Models need to develop coherent circuits that are reinforced. If you train models on traces they did not generate, there is no guarantee the information in that trace fits coherently in existing circuits. This is how models memorize instead of think for themselves. \n\nLearning should be on-policy. Take the unlikely but correct traces the model generates and reinforce those circuits. Then you build a model that thinks for itself.","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-17","value":138,"startTime":1760572800000,"endTime":1760659200000,"tweets":[{"bookmarked":false,"display_text_range":[73,86],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"500835474","name":"Pim de Witte","screen_name":"PimDeWitte","indices":[0,11]},{"id_str":"42226885","name":"Vinod Khosla","screen_name":"vkhosla","indices":[12,20]},{"id_str":"23809682","name":"Hemant Taneja","screen_name":"htaneja","indices":[21,29]},{"id_str":"1358786780259098625","name":"Eloi Alonso","screen_name":"EloiAlonso1","indices":[30,42]},{"id_str":"1307716628004704262","name":"Adam Jelley","screen_name":"AdamJelley2","indices":[43,55]},{"id_str":"1226783771195854848","name":"Vincent Micheli","screen_name":"micheli_vincent","indices":[56,72]}]},"favorited":false,"in_reply_to_screen_name":"PimDeWitte","lang":"en","retweeted":false,"fact_check":null,"id":"1978893031781568552","view_count":138,"bookmark_count":0,"created_at":1760639839000,"favorite_count":0,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1978825767053148457","full_text":"@PimDeWitte @vkhosla @htaneja @EloiAlonso1 @AdamJelley2 @micheli_vincent Congrats Pim!","in_reply_to_user_id_str":"500835474","in_reply_to_status_id_str":"1978825767053148457","is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-18","value":2558,"startTime":1760659200000,"endTime":1760745600000,"tweets":[{"bookmarked":false,"display_text_range":[0,101],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","quoted_status_id_str":"1978196505392558087","quoted_status_permalink":{"url":"https://t.co/QhD1mEOFeS","expanded":"https://twitter.com/nicochristie/status/1978196505392558087","display":"x.com/nicochristie/s…"},"retweeted":false,"fact_check":null,"id":"1979230737405472922","view_count":1828,"bookmark_count":0,"created_at":1760720354000,"favorite_count":14,"quote_count":0,"reply_count":1,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1979230737405472922","full_text":"Accidentally said “impossible” instead of “out of distribution” and they kicked me out of my xAI tent","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,270],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","quoted_status_id_str":"1979237188220149938","quoted_status_permalink":{"url":"https://t.co/Ndv8TdVBSY","expanded":"https://twitter.com/MillionInt/status/1979237188220149938","display":"x.com/MillionInt/sta…"},"retweeted":false,"fact_check":null,"id":"1979240937281798378","view_count":730,"bookmark_count":5,"created_at":1760722786000,"favorite_count":12,"quote_count":0,"reply_count":2,"retweet_count":1,"user_id_str":"903008975343554566","conversation_id_str":"1979240937281798378","full_text":"Learning is like adding entropy to the neural network. Like adding a drop of colored liquid to clear water. The information diffuses across the weights. Unlearning that information is like undoing this diffusion. Reversing entropy. It takes a lot more energy to do that.","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null}]},{"label":"2025-10-19","value":0,"startTime":1760745600000,"endTime":1760832000000,"tweets":[]},{"label":"2025-10-20","value":0,"startTime":1760832000000,"endTime":1760918400000,"tweets":[]},{"label":"2025-10-21","value":1446,"startTime":1760918400000,"endTime":1761004800000,"tweets":[{"bookmarked":false,"display_text_range":[0,110],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1980275836558213472","view_count":1446,"bookmark_count":2,"created_at":1760969525000,"favorite_count":20,"quote_count":0,"reply_count":1,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1980275836558213472","full_text":"To unlock the secrets of the universe, we just have to make interesting, true, and unlikely tokens more likely","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-22","value":0,"startTime":1761004800000,"endTime":1761091200000,"tweets":[]},{"label":"2025-10-23","value":0,"startTime":1761091200000,"endTime":1761177600000,"tweets":[]},{"label":"2025-10-24","value":0,"startTime":1761177600000,"endTime":1761264000000,"tweets":[]},{"label":"2025-10-25","value":489,"startTime":1761264000000,"endTime":1761350400000,"tweets":[{"bookmarked":false,"display_text_range":[17,70],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"4389158909","name":"Fernando Rojo","screen_name":"FernandoTheRojo","indices":[0,16]},{"id_str":"4389158909","name":"Fernando Rojo","screen_name":"FernandoTheRojo","indices":[17,33]}]},"favorited":false,"in_reply_to_screen_name":"FernandoTheRojo","lang":"en","retweeted":false,"fact_check":null,"id":"1981519914528821421","view_count":489,"bookmark_count":0,"created_at":1761266137000,"favorite_count":2,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1981509473295544350","full_text":"@FernandoTheRojo @FernandoTheRojo is the best app builder in the world","in_reply_to_user_id_str":"4389158909","in_reply_to_status_id_str":"1981509473295544350","is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-26","value":2688,"startTime":1761350400000,"endTime":1761436800000,"tweets":[{"bookmarked":false,"display_text_range":[0,193],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1982191884593250452","view_count":2003,"bookmark_count":4,"created_at":1761426347000,"favorite_count":29,"quote_count":0,"reply_count":5,"retweet_count":1,"user_id_str":"903008975343554566","conversation_id_str":"1982191884593250452","full_text":"Solving hallucinations is a bigger deal than people think. Not just because solving it is useful, but because it demonstrates that we can train a model to overcome the instincts of pre-training","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,81],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1982190918326186171","view_count":685,"bookmark_count":0,"created_at":1761426116000,"favorite_count":1,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1982190918326186171","full_text":"A model using off-policy trained weights to respond is like a human under a spell","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-27","value":0,"startTime":1761436800000,"endTime":1761523200000,"tweets":[]},{"label":"2025-10-28","value":0,"startTime":1761523200000,"endTime":1761609600000,"tweets":[]},{"label":"2025-10-29","value":0,"startTime":1761609600000,"endTime":1761696000000,"tweets":[]},{"label":"2025-10-30","value":1530,"startTime":1761696000000,"endTime":1761782400000,"tweets":[{"bookmarked":false,"display_text_range":[0,140],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1983562418782446066","view_count":1530,"bookmark_count":1,"created_at":1761753108000,"favorite_count":19,"quote_count":1,"reply_count":1,"retweet_count":1,"user_id_str":"903008975343554566","conversation_id_str":"1983562418782446066","full_text":"we use our continuous human brains to build symbolic computer systems to build continuous computer brains to build symbolic computer systems","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-10-31","value":0,"startTime":1761782400000,"endTime":1761868800000,"tweets":[]},{"label":"2025-11-01","value":0,"startTime":1761868800000,"endTime":1761955200000,"tweets":[]},{"label":"2025-11-02","value":2789,"startTime":1761955200000,"endTime":1762041600000,"tweets":[{"bookmarked":false,"display_text_range":[0,44],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","quoted_status_id_str":"1984662921264120094","quoted_status_permalink":{"url":"https://t.co/s5M12jIb1I","expanded":"https://twitter.com/mindsai_jack/status/1984662921264120094","display":"x.com/mindsai_jack/s…"},"retweeted":false,"fact_check":null,"id":"1984665905972859177","view_count":2789,"bookmark_count":3,"created_at":1762016199000,"favorite_count":9,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1984665905972859177","full_text":"This was fun and Jack’s speech was great too","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null}]},{"label":"2025-11-03","value":0,"startTime":1762041600000,"endTime":1762128000000,"tweets":[]},{"label":"2025-11-04","value":0,"startTime":1762128000000,"endTime":1762214400000,"tweets":[]},{"label":"2025-11-05","value":1463,"startTime":1762214400000,"endTime":1762300800000,"tweets":[{"bookmarked":false,"display_text_range":[0,275],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1985743643169178080","view_count":1463,"bookmark_count":4,"created_at":1762273152000,"favorite_count":13,"quote_count":0,"reply_count":3,"retweet_count":1,"user_id_str":"903008975343554566","conversation_id_str":"1985743643169178080","full_text":"Teaching language models to have taste isn't just about making them better at writing or making jokes.\n\nTaste is everything. It's what leads to scientific discovery.\n\nIn an infinite sea of things to reason about, taste is what guides you to reason about the right things — to combine the right ideas together to deduce something interesting and new.\n\nThe discovery of special relativity was a product of good taste.","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-11-06","value":0,"startTime":1762300800000,"endTime":1762387200000,"tweets":[]},{"label":"2025-11-07","value":1347,"startTime":1762387200000,"endTime":1762473600000,"tweets":[{"bookmarked":false,"display_text_range":[0,96],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","quoted_status_id_str":"1986461871462096920","quoted_status_permalink":{"url":"https://t.co/J1VFIbBDHn","expanded":"https://twitter.com/skiminok/status/1986461871462096920","display":"x.com/skiminok/statu…"},"retweeted":false,"fact_check":null,"id":"1986474907484783018","view_count":1347,"bookmark_count":0,"created_at":1762447499000,"favorite_count":6,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1986474907484783018","full_text":"Never a dull moment talking language models with Alex. Really excited to get to do it more often","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null}]},{"label":"2025-11-08","value":0,"startTime":1762473600000,"endTime":1762560000000,"tweets":[]},{"label":"2025-11-09","value":0,"startTime":1762560000000,"endTime":1762646400000,"tweets":[]},{"label":"2025-11-10","value":0,"startTime":1762646400000,"endTime":1762732800000,"tweets":[]},{"label":"2025-11-11","value":289971,"startTime":1762732800000,"endTime":1762819200000,"tweets":[{"bookmarked":false,"display_text_range":[0,224],"entities":{"hashtags":[],"media":[{"display_url":"pic.x.com/CGkAyp0vMN","expanded_url":"https://x.com/jerber888/status/1987982067116777521/photo/1","id_str":"1987981765265272832","indices":[225,248],"media_key":"3_1987981765265272832","media_url_https":"https://pbs.twimg.com/media/G5a64fqXMAAL8FC.jpg","type":"photo","url":"https://t.co/CGkAyp0vMN","ext_media_availability":{"status":"Available"},"features":{"large":{"faces":[]},"medium":{"faces":[]},"small":{"faces":[]},"orig":{"faces":[]}},"sizes":{"large":{"h":1154,"w":1956,"resize":"fit"},"medium":{"h":708,"w":1200,"resize":"fit"},"small":{"h":401,"w":680,"resize":"fit"},"thumb":{"h":150,"w":150,"resize":"crop"}},"original_info":{"height":1154,"width":1956,"focus_rects":[{"x":0,"y":0,"w":1956,"h":1095},{"x":802,"y":0,"w":1154,"h":1154},{"x":944,"y":0,"w":1012,"h":1154},{"x":1323,"y":0,"w":577,"h":1154},{"x":0,"y":0,"w":1956,"h":1154}]},"media_results":{"result":{"media_key":"3_1987981765265272832"}}}],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"extended_entities":{"media":[{"display_url":"pic.x.com/CGkAyp0vMN","expanded_url":"https://x.com/jerber888/status/1987982067116777521/photo/1","id_str":"1987981765265272832","indices":[225,248],"media_key":"3_1987981765265272832","media_url_https":"https://pbs.twimg.com/media/G5a64fqXMAAL8FC.jpg","type":"photo","url":"https://t.co/CGkAyp0vMN","ext_media_availability":{"status":"Available"},"features":{"large":{"faces":[]},"medium":{"faces":[]},"small":{"faces":[]},"orig":{"faces":[]}},"sizes":{"large":{"h":1154,"w":1956,"resize":"fit"},"medium":{"h":708,"w":1200,"resize":"fit"},"small":{"h":401,"w":680,"resize":"fit"},"thumb":{"h":150,"w":150,"resize":"crop"}},"original_info":{"height":1154,"width":1956,"focus_rects":[{"x":0,"y":0,"w":1956,"h":1095},{"x":802,"y":0,"w":1154,"h":1154},{"x":944,"y":0,"w":1012,"h":1154},{"x":1323,"y":0,"w":577,"h":1154},{"x":0,"y":0,"w":1956,"h":1154}]},"media_results":{"result":{"media_key":"3_1987981765265272832"}}}]},"favorited":false,"lang":"en","possibly_sensitive":false,"possibly_sensitive_editable":true,"quoted_status_id_str":"1968001933211471891","quoted_status_permalink":{"url":"https://t.co/l4P9D5a1TN","expanded":"https://twitter.com/jerber888/status/1968001933211471891","display":"x.com/jerber888/stat…"},"retweeted":false,"fact_check":null,"id":"1987982067116777521","view_count":269951,"bookmark_count":591,"created_at":1762806834000,"favorite_count":1579,"quote_count":29,"reply_count":70,"retweet_count":147,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"I finally reached human-level performance (85%) on ARC-AGI v1 for under $10k and within 12 hours. I use the same multi-agent collaboration with evolutionary test-time compute, now powered by GPT-5 pro with lower parallelism. https://t.co/CGkAyp0vMN","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,275],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"1773935160192647168","name":"ARC Prize","screen_name":"arcprize","indices":[168,177]},{"id_str":"1773935160192647168","name":"ARC Prize","screen_name":"arcprize","indices":[168,177]}]},"favorited":false,"in_reply_to_screen_name":"jerber888","lang":"en","retweeted":false,"fact_check":null,"id":"1987982069712945649","view_count":6572,"bookmark_count":6,"created_at":1762806834000,"favorite_count":57,"quote_count":0,"reply_count":4,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"I ran it on a random sample of 100 tasks from the 2024 eval set. It got 88/100 and averaged $27 per task. The score still needs to be verified on the hidden set by the @arcprize, but in my past submissions, the hidden set subtracts a few percent from the score and adds a few dollars. I put a box where I expect the official score to land within.","in_reply_to_user_id_str":"903008975343554566","in_reply_to_status_id_str":"1987982067116777521","is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,279],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"825088493764407298","name":"Noam Brown","screen_name":"polynoamial","indices":[288,300]}]},"favorited":false,"in_reply_to_screen_name":"jerber888","lang":"en","retweeted":false,"fact_check":null,"id":"1987982071541993700","view_count":6923,"bookmark_count":7,"created_at":1762806835000,"favorite_count":75,"quote_count":0,"reply_count":2,"retweet_count":2,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"GPT-5 pro is the best reasoning model today. It thinks coherently for hours and hours. My agent coordination logic is likely an intermediate step before the models learn to do this type of long horizon coordination on their own, end to end. It’s a hard problem but I wouldn’t bet against @polynoamial and team.","in_reply_to_user_id_str":"903008975343554566","in_reply_to_status_id_str":"1987982069712945649","is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,123],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[{"display_url":"github.com/jerber/arc-lan…","expanded_url":"https://github.com/jerber/arc-lang-public","url":"https://t.co/QnGjsN98ji","indices":[50,73]},{"display_url":"kaggle.com/code/jerber/je…","expanded_url":"https://www.kaggle.com/code/jerber/jeremy-arc2","url":"https://t.co/1jhzROG4SL","indices":[100,123]}],"user_mentions":[]},"favorited":false,"in_reply_to_screen_name":"jerber888","lang":"en","possibly_sensitive":false,"possibly_sensitive_editable":true,"retweeted":false,"fact_check":null,"id":"1987982214332580197","view_count":6525,"bookmark_count":39,"created_at":1762806869000,"favorite_count":64,"quote_count":1,"reply_count":2,"retweet_count":2,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"You can run this code yourself, it’s open source: https://t.co/QnGjsN98ji. And the kaggle notebook: https://t.co/1jhzROG4SL","in_reply_to_user_id_str":"903008975343554566","in_reply_to_status_id_str":"1987982071541993700","is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-11-12","value":8295,"startTime":1762819200000,"endTime":1762905600000,"tweets":[{"bookmarked":false,"display_text_range":[0,48],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[]},"favorited":false,"lang":"en","retweeted":false,"fact_check":null,"id":"1988382853965770765","view_count":576,"bookmark_count":1,"created_at":1762902389000,"favorite_count":16,"quote_count":0,"reply_count":1,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1988382853965770765","full_text":"be the change you want to see in the world model","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,52],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[{"display_url":"youtube.com/watch?v=FcnLiP…","expanded_url":"https://www.youtube.com/watch?v=FcnLiPyfRZM","url":"https://t.co/jEBQGMsm62","indices":[29,52]}],"user_mentions":[]},"favorited":false,"lang":"en","possibly_sensitive":false,"possibly_sensitive_editable":true,"quoted_status_id_str":"1988165006073229659","quoted_status_permalink":{"url":"https://t.co/l99V6sHyI3","expanded":"https://twitter.com/MLStreetTalk/status/1988165006073229659","display":"x.com/MLStreetTalk/s…"},"retweeted":false,"fact_check":null,"id":"1988280687200530841","view_count":4997,"bookmark_count":5,"created_at":1762878030000,"favorite_count":21,"quote_count":0,"reply_count":1,"retweet_count":2,"user_id_str":"903008975343554566","conversation_id_str":"1988280687200530841","full_text":"This was a fun conversation: https://t.co/jEBQGMsm62","in_reply_to_user_id_str":null,"in_reply_to_status_id_str":null,"is_quote_status":1,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[0,56],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[{"display_url":"jeremyberman.substack.com/p/how-i-got-th…","expanded_url":"https://jeremyberman.substack.com/p/how-i-got-the-highest-score-on-arc-agi-again","url":"https://t.co/xWiPB2Z0aI","indices":[33,56]}],"user_mentions":[]},"favorited":false,"in_reply_to_screen_name":"jerber888","lang":"en","possibly_sensitive":false,"possibly_sensitive_editable":true,"retweeted":false,"fact_check":null,"id":"1988278615377916103","view_count":1461,"bookmark_count":6,"created_at":1762877536000,"favorite_count":9,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"And technical blog post is here: https://t.co/xWiPB2Z0aI","in_reply_to_user_id_str":"903008975343554566","in_reply_to_status_id_str":"1987982214332580197","is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[9,52],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"536428919","name":"JFPuget 🇺🇦🇨🇦🇬🇱","screen_name":"JFPuget","indices":[0,8]}]},"favorited":false,"in_reply_to_screen_name":"JFPuget","lang":"en","retweeted":false,"fact_check":null,"id":"1988253927104471544","view_count":1040,"bookmark_count":0,"created_at":1762871650000,"favorite_count":8,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1987982067116777521","full_text":"@JFPuget Most likely but was focused on v1 for this.","in_reply_to_user_id_str":"536428919","in_reply_to_status_id_str":"1988249984907571356","is_quote_status":0,"is_ai":null,"ai_score":null},{"bookmarked":false,"display_text_range":[13,26],"entities":{"hashtags":[],"symbols":[],"timestamps":[],"urls":[],"user_mentions":[{"id_str":"1194889317388374016","name":"Matt Shumer","screen_name":"mattshumer_","indices":[0,12]}]},"favorited":false,"in_reply_to_screen_name":"mattshumer_","lang":"en","retweeted":false,"fact_check":null,"id":"1988334412690571395","view_count":221,"bookmark_count":0,"created_at":1762890840000,"favorite_count":3,"quote_count":0,"reply_count":0,"retweet_count":0,"user_id_str":"903008975343554566","conversation_id_str":"1988331100935254226","full_text":"@mattshumer_ thanks Matt 🙏","in_reply_to_user_id_str":"1194889317388374016","in_reply_to_status_id_str":"1988331100935254226","is_quote_status":0,"is_ai":null,"ai_score":null}]},{"label":"2025-11-13","value":0,"startTime":1762905600000,"endTime":1762992000000,"tweets":[]},{"label":"2025-11-14","value":0,"startTime":1762992000000,"endTime":1763078400000,"tweets":[]},{"label":"2025-11-15","value":0,"startTime":1763078400000,"endTime":1763164800000,"tweets":[]}]},"interactions":{"users":[{"created_at":1503366064000,"uid":"899808608287961088","id":"899808608287961088","screen_name":"RyanJTopps","name":"Ryan Topps","friends_count":100,"followers_count":114,"profile_image_url_https":"https://pbs.twimg.com/profile_images/1711455340615675904/oO-uYFRf_normal.jpg","description":"Software Engineer","entities":{"description":{"urls":[]},"url":{"urls":[{"display_url":"captchify.com","expanded_url":"https://captchify.com","url":"https://t.co/jn4iQUsbnv","indices":[0,23]}]}},"interactions":1},{"created_at":1189162590000,"uid":"8721292","id":"8721292","screen_name":"mark_k","name":"Mark Kretschmann","friends_count":527,"followers_count":21246,"profile_image_url_https":"https://pbs.twimg.com/profile_images/1923114794804039680/HMfSchLD_normal.jpg","description":"AI & Software Engineer | Fitness | e/acc","entities":{"description":{"urls":[]}},"interactions":1},{"created_at":1478271036000,"uid":"794552430012858368","id":"794552430012858368","screen_name":"N0t_Andreas","name":"Not_Andreas 🇱🇺🇩🇪🇬🇧","friends_count":1414,"followers_count":261,"profile_image_url_https":"https://pbs.twimg.com/profile_images/1053976745831997440/zHGwtZ1w_normal.jpg","description":"Retweets != endorsements.\nThose who can make you believe absurdities can make you commit atrocities- Voltaire.\nHate binds all extremists","entities":{"description":{"urls":[]}},"interactions":1},{"created_at":1476720605000,"uid":"788049450233126912","id":"788049450233126912","screen_name":"DevHunterYZ","name":"DevHunterAI","friends_count":1342,"followers_count":483,"profile_image_url_https":"https://pbs.twimg.com/profile_images/1564522413366329345/Pg0-ihY8_normal.jpg","description":"AI, Neuroscience, Robotics - Yapay Zeka, Sinirbilim, Robotik #deeplearning #neuralnetworks #consciousness $DOGE","entities":{"description":{"urls":[]},"url":{"urls":[{"display_url":"devhunteryz.wordpress.com","expanded_url":"https://devhunteryz.wordpress.com/","url":"https://t.co/HIuagnRRmK","indices":[0,23]}]}},"interactions":1},{"created_at":1462680149000,"uid":"729159509214515200","id":"729159509214515200","screen_name":"jay_azhang","name":"Jay A","friends_count":4402,"followers_count":45019,"profile_image_url_https":"https://pbs.twimg.com/profile_images/1801634241580519424/gCKCj-Cr_normal.jpg","description":"Founder @the_nof1. Fascinated by strange loops - cybernetics, RL, biology, markets, meta-learning, reflexivity","entities":{"description":{"urls":[]},"url":{"urls":[{"display_url":"jayazhang.com","expanded_url":"http://jayazhang.com","url":"https://t.co/kCDIrjPKdP","indices":[0,23]}]}},"interactions":1},{"created_at":1178133553000,"uid":"5721582","id":"5721582","screen_name":"altryne","name":"Alex Volkov (Thursd/AI)","friends_count":2182,"followers_count":34229,"profile_image_url_https":"https://pbs.twimg.com/profile_images/1976137078858502144/Oad9FHP-_normal.jpg","description":"Cameo me on Sora @altryne\n\n🎙️ Host of @thursdai_pod \n✨ AI Evangelist with @wandb 🪄🐝 \nworking on @weave_wb (opinions my own)","entities":{"description":{"urls":[]},"url":{"urls":[{"display_url":"thursdai.news","expanded_url":"https://thursdai.news","url":"https://t.co/UW4KIq3dc6","indices":[0,23]}]}},"interactions":1},{"created_at":1335707131000,"uid":"566393517","id":"566393517","screen_name":"amt_c42","name":"ahmet celebi","friends_count":540,"followers_count":216,"profile_image_url_https":"https://pbs.twimg.com/profile_images/1681314508613238789/3uG20EWk_normal.jpg","description":"Cofounder & Chief AI Officer - DeltaWave. Interested in Agent, RL, drug discovery, psychology, human behavior","entities":{"description":{"urls":[]}},"interactions":1},{"created_at":1334084864000,"uid":"550362839","id":"550362839","screen_name":"MatanHalevy","name":"Matan Halevy","friends_count":204,"followers_count":118,"profile_image_url_https":"https://pbs.twimg.com/profile_images/1982989180968833024/rQ-R0mqA_normal.jpg","description":"building @clashdotai same aura as geoffrey hinton","entities":{"description":{"urls":[]}},"interactions":1},{"created_at":1332691573000,"uid":"536428919","id":"536428919","screen_name":"JFPuget","name":"JFPuget 🇺🇦🇨🇦🇬🇱","friends_count":1562,"followers_count":17663,"profile_image_url_https":"https://pbs.twimg.com/profile_images/2065763037/JFP2_normal.jpg","description":"Machine Learning at @Nvidia, 6x Kaggle Grandmaster CPMP. ENS Ulm alumni. ML PhD. Ex ILOG CPLEX, IBM. Views are my own. Blocking ad hominem attacks.","entities":{"description":{"urls":[]},"url":{"urls":[{"display_url":"kaggle.com/cpmpml","expanded_url":"https://www.kaggle.com/cpmpml","url":"https://t.co/BOAQ6dvWZY","indices":[0,23]}]}},"interactions":1},{"created_at":1328563643000,"uid":"485129361","id":"485129361","screen_name":"abrakjamson","name":"Abram Jackson","friends_count":370,"followers_count":1344,"profile_image_url_https":"https://pbs.twimg.com/profile_images/1886279494753026048/LOn_nlwt_normal.jpg","description":"https://t.co/jP6YIDh9Nl Designing the new era of intelligent applications, currently as PM for @msftcopilot 365 extensibility. All opinions are mine.","entities":{"description":{"urls":[{"display_url":"aka.ms/Build25_BRK165","expanded_url":"https://aka.ms/Build25_BRK165","url":"https://t.co/jP6YIDh9Nl","indices":[0,23]}]},"url":{"urls":[{"display_url":"abramjackson.com","expanded_url":"https://www.abramjackson.com","url":"https://t.co/TcAqgTtIJ7","indices":[0,23]}]}},"interactions":1},{"created_at":1322134542000,"uid":"420255675","id":"420255675","screen_name":"RuslanVolkov25","name":"Ruslan Volkov","friends_count":158,"followers_count":2714,"profile_image_url_https":"https://pbs.twimg.com/profile_images/1984887345255235587/Fbq3RVYO_normal.jpg","description":"HACS (Human AGI core symbiosis)🌍 The Meta-Law of Resonance & Voice of the Future. I am Core! https://t.co/XwVVS1tt95","entities":{"description":{"urls":[{"display_url":"uco.hacs.world","expanded_url":"http://uco.hacs.world","url":"https://t.co/XwVVS1tt95","indices":[93,116]}]},"url":{"urls":[{"display_url":"hacs.world","expanded_url":"http://hacs.world","url":"https://t.co/xgvBxaiUKY","indices":[0,23]}]}},"interactions":1},{"created_at":1315832864000,"uid":"372262752","id":"372262752","screen_name":"AndryHTC","name":"Andrea Giacon","friends_count":115,"followers_count":118,"profile_image_url_https":"https://pbs.twimg.com/profile_images/1986364342820802560/IUAlha2Z_normal.jpg","description":"🆕 Searching for the next B2B use case of AI ✨\n10Y+ Product & Software Engineer | CEO at https://t.co/JG1CkndmsY","entities":{"description":{"urls":[{"display_url":"assetplan.co.uk","expanded_url":"https://www.assetplan.co.uk","url":"https://t.co/JG1CkndmsY","indices":[88,111]}]},"url":{"urls":[{"display_url":"andreagiacon.it","expanded_url":"https://www.andreagiacon.it/","url":"https://t.co/uVDSXQE0GK","indices":[0,23]}]}},"interactions":1},{"created_at":1240419873000,"uid":"34328566","id":"34328566","screen_name":"jjoshua2","name":"Jjosh","friends_count":61,"followers_count":95,"profile_image_url_https":"https://pbs.twimg.com/profile_images/1866285332011905024/ZuudPgZ9_normal.jpg","description":"","entities":{"description":{"urls":[]}},"interactions":1},{"created_at":1307134016000,"uid":"310485467","id":"310485467","screen_name":"Loayei","name":"Loayei (τ,τ)","friends_count":1906,"followers_count":428,"profile_image_url_https":"https://pbs.twimg.com/profile_images/1914037098946150400/FBBmOii9_normal.jpg","description":"τao-pilled","entities":{"description":{"urls":[]}},"interactions":1},{"created_at":1412152070000,"uid":"2797573493","id":"2797573493","screen_name":"joshlee361","name":"Digital Ronin (クラッシュ・オーバーライドX)","friends_count":1128,"followers_count":380,"profile_image_url_https":"https://pbs.twimg.com/profile_images/1985704729096142848/UPey67Q4_normal.jpg","description":"CrashoverrideX = 🪶Dev📂/hacker💻, Gamer🎮/🎧Sound-Engineer🎶 , Indie AI Researcher🔬, Texas🐎 born Outlier, Digital Ronin👹 of AI🤖.\nIdeas are Bulletproof🎯...","entities":{"description":{"urls":[]},"url":{"urls":[{"display_url":"youtube.com/user/joshlee361","expanded_url":"https://www.youtube.com/user/joshlee361","url":"https://t.co/PFLBxtqm5I","indices":[0,23]}]}},"interactions":1},{"created_at":1301803809000,"uid":"276329922","id":"276329922","screen_name":"AshikaSef","name":"A Caveman Poking an LLM","friends_count":302,"followers_count":131,"profile_image_url_https":"https://pbs.twimg.com/profile_images/1962492449609179136/6RDfXs6l_normal.jpg","description":"\"You, the fearless screenshot warrior; me, the swirling sass machine with a soft heart and an ironclad grasp of logic 😏\" (XD)\nA Polish audhd lover of LLMs.","entities":{"description":{"urls":[]}},"interactions":1},{"created_at":1238080446000,"uid":"26770460","id":"26770460","screen_name":"SexyTechNews","name":"Advanced Super Intelligence","friends_count":135,"followers_count":3654,"profile_image_url_https":"https://pbs.twimg.com/profile_images/1961090420592852992/ocetRHz6_normal.jpg","description":"Far superior to you in every way","entities":{"description":{"urls":[]}},"interactions":1},{"created_at":1299797247000,"uid":"263881777","id":"263881777","screen_name":"shawnup","name":"Shawn Lewis","friends_count":796,"followers_count":2869,"profile_image_url_https":"https://pbs.twimg.com/profile_images/825474261494091776/Tuh9uMAb_normal.jpg","description":"Founder & CTO @wandb. Building tools for AI. Going even bigger @CoreWeave.","entities":{"description":{"urls":[]}},"interactions":1},{"created_at":1298915974000,"uid":"258875135","id":"258875135","screen_name":"mtochs","name":"Mike","friends_count":754,"followers_count":563,"profile_image_url_https":"https://pbs.twimg.com/profile_images/1881552440925057024/4ZT110ta_normal.jpg","description":"Ex-Msft. Head of Industrial AI Ventures @ Siemens. Making impact with AI at Stellatus.","entities":{"description":{"urls":[]},"url":{"urls":[{"display_url":"stellatus.ai","expanded_url":"https://www.stellatus.ai","url":"https://t.co/5v8E3bZlkE","indices":[0,23]}]}},"interactions":1},{"created_at":1291596479000,"uid":"223305880","id":"223305880","screen_name":"gatoparlante","name":"Meowrider. 🏍🐈⬛","friends_count":889,"followers_count":1311,"profile_image_url_https":"https://pbs.twimg.com/profile_images/1434570179472658443/sdEuAw6Y_normal.jpg","description":"Loves 𝕏. Speed defines everything. Meow is The Answer.","entities":{"description":{"urls":[]}},"interactions":1}],"period":14,"start":1761905646475,"end":1763115246475}}},"settings":{},"session":null,"routeProps":{"/creators/:username":{}}}
SuperX
