Get live statistics and analysis of jason's profile on X / Twitter

👽

778 following1k followers

The Analyst

Jason is a deep-thinking AI and tech enthusiast who loves dissecting complex ideas with sharp logic and clarity. His tweets mix technical insight with big-picture societal implications, inviting followers to rethink assumptions about AI, search, and innovation. Always curious and contrarian, Jason challenges prevailing narratives with nuanced arguments and thoughtful critiques.

Impressions
1.3M-806.4k
$249.63
Likes
13.3k-9.1k
95%
Retweets
121-74
1%
Replies
206-47
1%
Bookmarks
389-231
3%

Top users who interacted with jason over the last 14 days

@Yuchenj_UW

Co-founder & CTO @hyperbolic_labs cooking fun AI systems. Prev: OctoAI (acquired by @nvidia) building Apache TVM, PhD @ University of Washington.

5 interactions
@giyu_codes

Applied AI/ML & Full Stack dev. Optimizing Small Medium Enterprises with AI tooling and fundamental software. destroyer of b2b SaaS integrations

5 interactions
1 interactions
@stephsmithio

📈 Leading Growth @GroqInc 📌 Prev @a16z @HubSpot @TheHustle 💻 Chronically online: internetpipes.com 📘 Wrote doingcontentright.com 🎙 Podcast at @sydlis 👇

1 interactions
1 interactions
@rohanpaul_ai

Compiling in real-time, the race towards AGI. The Largest Show on X for AI. 🗞️ Get my daily AI analysis newsletter to your email 👉 rohan-paul.com

1 interactions
1 interactions
@PiretHappiness

WHY WERE YOU BORN? (author of How to Buy Happiness) Originally from Estonia.

1 interactions
@AetasFuturis

Covering the socio-economic trends, companies, stocks and technological advancements that will play out in this century and the next. Type 1: January 1st 2101

1 interactions
1 interactions
1 interactions
@KevinSzabo14

•Premium Ghostwriting•Helping Creators find the voice • Generated over 300k Quality followers | following back email subscribers

1 interactions
@paraschopra

life is a game 🕹️ • building @lossfunk

1 interactions
@noah_vandal

Born again Christian! BS | MS @NDSU iloveneurons.com Building @SpeechSage Passion for biomedical arenas Love a good debate, but only if there is purpose

1 interactions
@johnrushx

I run the most automated org on earth, using the AI Agents I built. @unicornplatform @indexrusher @listingbott @seobotai tinyadz.com 24 startups → johnrush.me

1 interactions
@iamgingertrash

to be made of the ilk of Seraphim

1 interactions
@SherryYanJiang

building Peek (peek.money) & codewithai.xyz, @cursor_ai ambassador, prev. @Google @Amazon, ADHD founder, DJ 🎧, poker addict 🃏

1 interactions
@Teknium

Cofounder and Head of Post Training @NousResearch, prev @StabilityAI Github: github.com/teknium1 HuggingFace: huggingface.co/teknium

1 interactions
@Joel__Collinson

Ardent futurist | Techno Optimist | Aging reversal enthusiast | Universal income proponent | Metaphysics thinker | Business Owner | Profound hearing loss

1 interactions
@plz_greencard

I want to win a green card. Let's see. Although, probably I won't go to the US. I like this country but it is run by Mossad. #ReleaseTheSnyderCut

1 interactions

Jason’s tweets are so dense and meticulous, if they were a meal, you’d need a PhD just to digest them—and a strong coffee to stay awake through it! Somehow he manages to turn even a casual chat about chatbots into a doctoral thesis, keeping everyone on their toes (or running for simpler feeds).

Successfully predicted and demonstrated a better, vector-free search pipeline for AI retrieval tasks before the AI community widely recognized the limitations of vector embeddings in large context windows.

Jason's life purpose centers on advancing understanding in technology and its societal impact, helping people see beyond hype and simplified narratives. He aims to illuminate the subtleties of AI, information retrieval, and economics to promote smarter conversations and better decisions.

He believes in transparency, reasoned debate, and the power of open knowledge to dismantle scarcity and unlock freedom. Jason holds skepticism towards manufactured consensus and values intellectual rigor and autonomy over flashy trends or surface-level hype.

Jason’s strengths lie in his analytical brilliance, clear communication of complex topics, and ability to connect technical details with broader philosophical and economic contexts. His prolific output shows dedication and stamina in sharing valuable insights consistently.

At times, his dense and nuanced style may feel inaccessible or too technical for casual audiences, potentially limiting broader engagement. His contrarian streak might come off as overly critical or dismissive to some followers.

To grow his audience on X, Jason should consider blending his deep analyses with more approachable, bite-sized content and engaging storytelling. Pairing technical threads with relatable use cases or quick myth-busting tweets can broaden appeal and spark more interactive discussions.

Jason has tweeted over 9,000 times, revealing a prolific and engaged mind. His thoughtful thread on vector search attracted over 10,000 views and sparked lively discussions, demonstrating his expertise and influence in AI tech circles.

Top tweets of jason

my thoughts on why vector search is fading in 2026 for the last week i wanted to test if i could build a production-level rag pipeline without vectors. no embeddings, no gpu burn. just bm25 (elasticsearch) for retrieval, then dump as much context as possible to gemini for reranking/filtering chunks and generating cited responses. what happened didn't really surprise me: it worked better. ingestion was near-instant (+ no overhead). retrieval was more transparent. and the reasoning model was superb at pulling out a needle from the haystack. vectors made sense in 2023 when llms had tiny context windows and reasoning didn't exist. back then you had to chunk elegantly and hope cosine distance found relevant fragments from which the llm could directly deduce an answer. but models today can intuitively reason across context windows of 1m+ tokens. the problem then goes from chunk-level retrieval to document-level retrieval and vectors simply fall apart there: > vectors are costly to ingest (huge upfront embedding step - storage scales quadratically!) > vectors are brittle for large-doc retrieval (signal gets washed out when embedding large chunks, and small chunks are simply too expensive to scale) in practice, traditional search methods (tf-idf/bm25) + reasoning beat vectors on cost, speed, and often accuracy. this paper by @orionweller from google deepmind is fascinating. turns out some docs in your index are theoretically incapable of being retrieved by vector search. plain old bm25 outperforms on recall every time. arxiv.org/abs/2508.21038

10k

jason reposted

@nikitabier london is like a live-action meme where every npc is from a different dlc

5k
Reposted @swyx

The best startup AI Engineers I've met are all building their own agents. I know it's a buzzword that's now a bit past the wave…

361k
Reposted @DrJimFan

Many of us practitioners have felt that GPT-4 degrades over time. It's now corroborated by a recent study. But why does GPT-4…

556k

Most engaged tweets of jason

my thoughts on why vector search is fading in 2026 for the last week i wanted to test if i could build a production-level rag pipeline without vectors. no embeddings, no gpu burn. just bm25 (elasticsearch) for retrieval, then dump as much context as possible to gemini for reranking/filtering chunks and generating cited responses. what happened didn't really surprise me: it worked better. ingestion was near-instant (+ no overhead). retrieval was more transparent. and the reasoning model was superb at pulling out a needle from the haystack. vectors made sense in 2023 when llms had tiny context windows and reasoning didn't exist. back then you had to chunk elegantly and hope cosine distance found relevant fragments from which the llm could directly deduce an answer. but models today can intuitively reason across context windows of 1m+ tokens. the problem then goes from chunk-level retrieval to document-level retrieval and vectors simply fall apart there: > vectors are costly to ingest (huge upfront embedding step - storage scales quadratically!) > vectors are brittle for large-doc retrieval (signal gets washed out when embedding large chunks, and small chunks are simply too expensive to scale) in practice, traditional search methods (tf-idf/bm25) + reasoning beat vectors on cost, speed, and often accuracy. this paper by @orionweller from google deepmind is fascinating. turns out some docs in your index are theoretically incapable of being retrieved by vector search. plain old bm25 outperforms on recall every time. arxiv.org/abs/2508.21038

10k

jason reposted

@nikitabier london is like a live-action meme where every npc is from a different dlc

5k
Reposted @swyx

The best startup AI Engineers I've met are all building their own agents. I know it's a buzzword that's now a bit past the wave…

361k
Reposted @DrJimFan

Many of us practitioners have felt that GPT-4 degrades over time. It's now corroborated by a recent study. But why does GPT-4…

556k

People with Analyst archetype

The Analyst

Uncursing RL! Puffer: puffer.ai Paper on Foundational Work: Playing Pokémon Red via RL Blog and Live Training Map: Beating Pokemon Red w/ RL see details

790 following3k followers
The Analyst

22 | Ex-Data engineer • coding • cooking •ML• maths

1k following3k followers
The Analyst

Predict crypto & stock prices @trylimitless Markets, tech & policy journal cjhtech.substack.com

1k following24k followers
The Analyst

|Web3 追随者| 事件解读| 行情分析|技术分析|投资心法|金融知识| *任何分享均不构成投资建议*

2k following31k followers
The Analyst

ai x crypto pleb | nothing to add here yet

1k following1k followers
The Analyst

AI & Robustness in chaos. Insight → Strategy → Operation Now,not How. 我对现实世界的认知 贫瘠到无法自知其贫瘠

493 following20k followers
The Analyst

18 // building @trycommonwealth // DEPOSIT BANKS ARE DEAD // These things aren’t worth as much as you think they are // Chicago 🇺🇸 & Hong Kong 🇭🇰

552 following1k followers
The Analyst

🧭 Navigating crypto trends | Dropping alpha daily

371 following598 followers
The Analyst

DSA + Web Dev | Code. Learn. Build.

1k following3k followers
The Analyst

Web3市场投研 | Alpha挖掘 | 热衷于撸毛 | 撸空投 | 社区运营 | 长期看好币安生态发展在加密浪潮中持续探索价值机遇 | 合作dm

3k following12k followers
The Analyst

Sharing insights, guides, and airdrop strategies | DYOR | On-chain believer | Early @KaitoAI & @BioProtocol

638 following1k followers
The Analyst

无意义和意义

430 following40 followers

Explore Related Archetypes

If you enjoy the analyst profiles, you might also like these personality types:

Supercharge your 𝕏 game,
Grow with SuperX!

Get Started for Free