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

Currently building @OpenPipeAI (acquired by @CoreWeave). Formerly @ycombinator, @google. "Because I have been given much I too must give."

269 following17k followers

The Innovator

Kyle Corbitt is a tech trailblazer passionately advancing AI technology through open-source projects and cutting-edge research. With a rich background at Google and Y Combinator, Kyle thrives on pushing boundaries and sharing knowledge generously. Always curious and hands-on, Kyle turns complex tech into accessible breakthroughs.

Impressions
265.3k-109k
$49.74
Likes
1.8k-380
61%
Retweets
71-19
2%
Replies
115-19
4%
Bookmarks
974-102
33%

Top users who interacted with Kyle Corbitt over the last 14 days

@ankit2119

RL and long horizon agents. I dont talk much, just have curious questions on latest arxiv papers.

1 interactions
@redtachyon

p/hd | Big RL energy | 0.71 |research⟩ + 0.71 |engineer⟩ leaving Meta | Soon to be MoE elsewhere | Prev. lead maintainer of Gym(nasium)

1 interactions
1 interactions
@GUT_AI_F

GUT-AI Foundation — AI/acc | Pro-Innovation |Pro-Tech | Mass adoption of AI | #AI #ML #OpenSource #COSS #OpenStartup #OpenScience #Privacy #CensorshipResistance

1 interactions
@vikhyatk

teaching computers how to see @moondreamai

1 interactions
@arb8020

s̶k̶i̶l̶l̶ ̶i̶s̶s̶u̶e̶ will issue | prev: @morph_labs, five rings

1 interactions
@CaoMinhWeb3

Web3 Marketing | Investor | Content Creation · Lead @MBMweb3 · Founder @Trend_DAO 💎 Pudgy Penguins 💎

1 interactions
@CarryFaze

computer engineering @UCSB class of 2027, UCSB NLP lab | prev: swe intern @Cisco

1 interactions

Kyle’s so deep in AI clusters and infiniband links, he probably dreams in GPU cores—just don’t ask him to simplify it unless you want a full-stack explanation that crashes your brain like a rogue infinite loop.

Successfully launched agent.exe, a multi-platform app harnessing Claude 3.5 Sonnet’s API to control computers via AI agents, showcasing his knack for blending cutting-edge research with practical real-world applications.

To revolutionize AI development by pioneering innovative tools and frameworks that empower both developers and users, fostering open collaboration and advancing the field of artificial intelligence.

Kyle believes that knowledge should be shared freely and that the power of AI is best realized through open collaboration and community-driven innovation. He values transparency, practical experimentation, and continuous learning as key drivers of technological progress.

Kyle’s strengths lie in his deep technical expertise, ability to innovate within complex AI landscapes, and his commitment to open-source accessibility. His hands-on approach and effective communication make cutting-edge AI approachable for a wider audience.

His laser focus on technical innovation sometimes risks alienating non-expert audiences, and his critical eye can come off as overly blunt or bearish, potentially limiting broader community engagement.

To grow his audience on X, Kyle should blend his deep technical insights with more storytelling and approachable explanations, showcasing real-world impact and user stories. Engaging in regular Q&A sessions and sharing behind-the-scenes development journeys can humanize his brand while maintaining authority.

Fun fact: Kyle isn’t just building AI models—he’s fine-tuning and testing them on his own fine-tuning test suite to ensure they outperform giants like GPT-4o mini, proving that smaller, open models can punch above their weight.

Top tweets of Kyle Corbitt

"RL from a single example works" "RL with random rewards works" "Base model pass@256 can match RL model pass@1" "RL updates a small % of params" Recent papers all point in the same direction: RL is mostly just eliciting latent behavior already learned in pretraining, not teaching new behavior. Yann Lecun's "RL is the cherry on top" was right after all. Is this bearish for RL? Perhaps not! Maybe we should think about RL as the last mile of on-the-job training for your specific task. When you hire a customer support rep, you start with someone who's already smart and capable. But you *also* watch them take a few calls and give them feedback. That helps them align them to the nuances of your org and task, and can improves their performance dramatically. RL is exactly this, but for LLMs. I expect it to become standard practice across orgs serious about deploying reliable agents.

101k

Most engaged tweets of Kyle Corbitt

I am working with @mattshumer_ to get to the bottom of what happened with Reflection. He is providing access to all serving code and weights with the goal of replicating the strong reasoning performance @ArtificialAnlys was able to see over the weekend. I will report back.

131k

"RL from a single example works" "RL with random rewards works" "Base model pass@256 can match RL model pass@1" "RL updates a small % of params" Recent papers all point in the same direction: RL is mostly just eliciting latent behavior already learned in pretraining, not teaching new behavior. Yann Lecun's "RL is the cherry on top" was right after all. Is this bearish for RL? Perhaps not! Maybe we should think about RL as the last mile of on-the-job training for your specific task. When you hire a customer support rep, you start with someone who's already smart and capable. But you *also* watch them take a few calls and give them feedback. That helps them align them to the nuances of your org and task, and can improves their performance dramatically. RL is exactly this, but for LLMs. I expect it to become standard practice across orgs serious about deploying reliable agents.

101k

People with Innovator archetype

The Innovator

NPMing the universe since 2012: 🇨🇿 Invoicing solution: faktorio.cz 👂Speech-to-text: callnotes.fyi 🗝️ Password manager: @authierpm

1k following1k followers
The Innovator

Seeding superhumans @bio_hacker_dao | Deep-tech Investor superhuman.fund | Podcast host 👇

2k following12k followers
The Innovator

Currently { Crypto and Web3 } Curious. Ex AntlerVC, Wingify. 100% Techie. 1X FOUNDER (Unicom Techlabs acquired by @knowlarity).

6k following1k followers
The Innovator

nyc | swe | talk to me about ai/ml infra | opinions my own | 🇵🇰🇺🇸

58 following13 followers
The Innovator

Co-Founder @andthenchat (a16z SR005), Emmy-Winning Showrunner, co-creator @aiforhumansshow made a career of hurtling emergent media into established spaces

4k following24k followers
The Innovator

ML researcher (@primeintellect), speculator • extremely silly jester

2k following20k followers
The Innovator

Applied ML Specialist | Turning research into real-world products. Follow for insights, demos & deep dives into the latest in AI. New: videosearch.animikh.me

435 following358 followers
The Innovator

Building intelligent memory for AI and humans to work together | Waymaker One | Beta launching soon

485 following6k followers
The Innovator

26 . @auramaxdotfun | exploring DeFi w @metromxyz | @techstars alum | @ns v2 | @jump_pit 🇸🇬 🏆 | Public Goods - volunteer @alwayscareindia | Ex. @buildoorxyz

1k following406 followers
The Innovator

🪂 空投猎人|Yap氛围制造机|加密梦行者 携手前行,共建Web3新纪元 🚀 #互粉 #YAPPER #CONNECT #WEB3 #KAITO Airdrop Seeker | Yap Soul Active | Dream Trader 📊 Follow & explore the Web3 future

264 following440 followers
The Innovator

accelerating @aztecnetwork ⚡️ | views are my own | zcash is planetary scale, unstoppable, encrypted money!

1k following4k followers
The Innovator

bd at @overtime_io

1k following566 followers

Explore Related Archetypes

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

Supercharge your 𝕏 game,
Grow with SuperX!

Get Started for Free