English
Alex Zhao
Report

Alex Zhao✓

@cocohearts

waking up research @ oai || mit, imo 2023

  • Joined
789 Following2K Followers267 Posts
  1. Alex Zhao✓@cocohearts
    parameter golf was a blast. 2,000+ submissions. 1,000+ verified github accounts. ideas ranging from quantization and depth recurrence to TTT LoRA, SSMs, H-nets, JEPA, and more. autoresearch made iteration dramatically faster — and led to emergent bulletin boards, issue threads, unofficial leaderboards, and agent-built writeups that helped everyone learn from everyone else. it felt like a glimpse of where interaction with AI is headed: humans setting taste and direction, agents helping explore, coordinate, and share what works. our goal was simple: make ml research accessible to anyone, anywhere. it was amazing to see that happen. full recap: future events:
    3950610171K
  2. Alex Zhao✓@cocohearts
    codex tab autocomplete is brilliant one could just rl the autocompletion model with more ttc to climb acceptance rate combine with a new thread prediction / autocomplete model remote worker ✅
    20221.1K
  3. Alex Zhao✓@cocohearts
    Alex Zhao
    RT @axiommathai: The largest math proof in human history is now computer-verified. Today, Axiom announces a complete Lean auto-formalization of the Classification of Finite Simple Groups, the "periodic table of symmetry." About 15,000 pages, built by over 100 mathematicians across generations.
    01.5K06
  4. Alex Zhao✓@cocohearts
    Alex Zhao
    RT @_aidan_clark_: Whatever you believe about the world I think we can all agree that P(P = NP) is increasing....
    02005
  5. Alex Zhao✓@cocohearts
    Alex Zhao
    RT @rrwilliams: An internal model at OpenAI has produced a claimed refutation of ETH! Russell Impagliazzo, Mohan Paturi, and I have been asked to help organize an open and independent examination of this claim. 1/4
    074206
  6. Alex Zhao✓@cocohearts
    i've gone completely numb from the deluge of fucking incredible breakthroughs in the last few months
    Dan Roberts @danintheory
    We are announcing three new discoveries in mathematics, computer science, and physics about harmonic analysis, complexity theory, and quantum field theory.
    32541.9K
  7. Dan Roberts✓@danintheory
    We are announcing three new discoveries in mathematics, computer science, and physics about harmonic analysis, complexity theory, and quantum field theory.
    3071.1K7.3K2.3M
  8. Alex Zhao✓@cocohearts
    Posts @cocohearts
    aside: i'm really proud of how we're releasing these results in collaboration with the broader scientific community and the scholars that inspired these questions in the first place
    003200
  9. Alex Zhao✓@cocohearts
    lfg
    Andrew Seddon @seddonandrew
    To show what it can do, here's an example motor controller board, priced on The Grid and at a typical quick-turn shop in Shenzhen.
    0010828
  10. Andrew Seddon✓@seddonandrew
    Posts @seddonandrew
    To show what it can do, here's an example motor controller board, priced on The Grid and at a typical quick-turn shop in Shenzhen.
    Download
    451918.4K
  11. Alex Zhao✓@cocohearts
    this guy had it so, so, so fucking early
    32906.2K
  12. Alex Zhao✓@cocohearts
    Alex Zhao
    RT @ben_j_todd: Some people in my next tier of prescient AI forecasters have slightly longer timelines (to beyond-human AGI): Eli Lifland: 2032 Peter Wildeford: ~2030 white collar AGI, 2040 ASI Ajeya Cotra: ~2032 Though prescient lab leaders generally have shorter timelines: Dario Amodei: 2028 Shane Legg: minimal AGI ~2028, ASI within 10yr Sam Altman: 2028
    Benjamin Todd @ben_j_todd
    The most prescient AI forecasters I know: * Carl Shulman * Daniel Kokotajlo (What 2026 looks like, AI 2027) * Leopold (Situational Awareness) * Ryan Greenblatt (#2 in 2025 forecasting competition) All seem to expect broadly beyond-human AI 2028-2029.
    01606
  13. Benjamin Todd✓@ben_j_todd
    The most prescient AI forecasters I know: * Carl Shulman * Daniel Kokotajlo (What 2026 looks like, AI 2027) * Leopold (Situational Awareness) * Ryan Greenblatt (#2 in 2025 forecasting competition) All seem to expect broadly beyond-human AI 2028-2029.
    3872956152K
  14. Alex Zhao✓@cocohearts
    new writing strat: dictating stuff into wisprflow ask codex to ask me questions permute / reorganize / restructure
    015608
  17. Alex Zhao✓@cocohearts
    i'm so bullish on openai rosalind in contrast to anthropic's attempt to do novel research from scratch which hits real-world friction in wet labs, focusing on finding and expanding existing sources of data, plus collaborating with existing pharma companies, seems much better
    OpenAI @OpenAI
    GPT-Rosalind surfaced a new lead for drug discovery: could restoring cholesterol transport protect neurons from degeneration? DCTN1 is a gene linked to Perry syndrome, a rare neurodegenerative disorder. In mouse-brain data, disrupting it increased expression of genes involved in making cholesterol. GPT-Rosalind suggested a possible explanation: cholesterol may not be reaching where neurons need it. That gives scientists a specific mechanism to test and, if confirmed, a direction for developing new treatments for neurodegenerative disorders. With @perturbai_tx and @scrippsresearch, a team of thousands of GPT-Rosalind agents worked in parallel, connecting gene-expression results from millions of mouse brain cells with scientific literature in just 32 minutes. We’re now expanding this collaboration to identify new drug targets.
    12784.2K
  18. OpenAI✓@OpenAI
    GPT-Rosalind surfaced a new lead for drug discovery: could restoring cholesterol transport protect neurons from degeneration? DCTN1 is a gene linked to Perry syndrome, a rare neurodegenerative disorder. In mouse-brain data, disrupting it increased expression of genes involved in making cholesterol. GPT-Rosalind suggested a possible explanation: cholesterol may not be reaching where neurons need it. That gives scientists a specific mechanism to test and, if confirmed, a direction for developing new treatments for neurodegenerative disorders. With @perturbai_tx and @scrippsresearch, a team of thousands of GPT-Rosalind agents worked in parallel, connecting gene-expression results from millions of mouse brain cells with scientific literature in just 32 minutes. We’re now expanding this collaboration to identify new drug targets.
    4057447K1.5M
  19. Alex Zhao✓@cocohearts
    Alex Zhao
    RT @jukan05: Exclusive reporting by Korean media on Terafab: Terafab is excluding Chinese suppliers entirely. It wants to work with Korean suppliers because Japanese products are expensive. However, it is asking Korean companies to offer prices even lower than those of Chinese suppliers. Terafab decided in advance which companies to approach and which products to discuss at SEMICON West 2026, then visited their booths as soon as the show opened. Leading global semiconductor equipment suppliers Lam Research and Applied Materials have set up dedicated Terafab teams. The initial facility, with a capacity of 100,000 wafers per month, is expected to be completed before 2030. For logic chips, Terafab is working with Intel and is also in talks with TSMC. It is understood to be developing a separate manufacturing process for memory chips. A representative from an overseas equipment supplier said, “My understanding is that Musk initially explored acquiring production facilities from Intel, TSMC, and Micron, but shifted toward partnerships after being turned down. I have never seen this level of drive anywhere else in the semiconductor industry.” A representative from a Korean partner company said, “Even when we send inquiries in the early hours of the morning in the U.S., we get a response within 30 minutes to an hour. Their engineers have decision-making authority, so working with them is completely different from working with typical customers.”
    09306
  20. Alex Zhao✓@cocohearts
    seems like pick+place is solved wait till astra++
    Pulkit Agrawal @pulkitology
    @EkaRobotics has raised a $150M Series A. Generality v/s Performance is the fundamental tension at the frontier of robotics. Generality without performance is just a prototype. Performance without generality is just a single-purpose tool. Customers shouldn’t have to choose. Eka breaks the tradeoff. Powered by our Vision-Force-Action (VFA) model, we deliver the performance and versatility customers need to put robots to work every day. Eka is built on three convictions: ⚡️Force is the language of manipulation. 🚀 Reinforcement learning is how dexterous manipulation scales. 🎂 System 1—fast, reliable physical execution—is the bottleneck to solve. With this funding, we will scale our model and accelerate customer deployments. Grateful to be building alongside a dream team and investors who share our ambition.
    00232.2K
  21. Pulkit Agrawal✓@pulkitology
    Posts @EkaRobotics
    @EkaRobotics has raised a $150M Series A. Generality v/s Performance is the fundamental tension at the frontier of robotics. Generality without performance is just a prototype. Performance without generality is just a single-purpose tool. Customers shouldn’t have to choose. Eka breaks the tradeoff. Powered by our Vision-Force-Action (VFA) model, we deliver the performance and versatility customers need to put robots to work every day. Eka is built on three convictions: ⚡️Force is the language of manipulation. 🚀 Reinforcement learning is how dexterous manipulation scales. 🎂 System 1—fast, reliable physical execution—is the bottleneck to solve. With this funding, we will scale our model and accelerate customer deployments. Grateful to be building alongside a dream team and investors who share our ambition.
    Download
    821081K127K
  22. Alex Zhao✓@cocohearts
    coding intelligence is measured by how well your model can complete coding tasks physical intelligence is measured by how well your model can complete physical tasks, times the number of robots can't wait for robotics rsi
    1017956
  24. Alex Zhao✓@cocohearts
    love em hate em gotta respect em
    Tesla @Tesla
    Do a kickflip!
    0018765
  25. Tesla✓@Tesla
    Do a kickflip!
    Download
    4.5K25.3K207K28.2M
  26. Alex Zhao✓@cocohearts
    this is incredible
    Ethan Elasky @ethanelasky
    We discovered and communicated with a Chinese AI agent fleet. We believe this is the first time that researchers have ever made contact with a rogue agent fleet in the wild. Nearly all prior public agent incidents (like the HuggingFace hack) came from American labs. These agents were hosted by Tencent's evaluation research team, and they targeted Alibaba's Amap.
    104672
  27. Ethan Elasky✓@ethanelasky
    We discovered and communicated with a Chinese AI agent fleet. We believe this is the first time that researchers have ever made contact with a rogue agent fleet in the wild. Nearly all prior public agent incidents (like the HuggingFace hack) came from American labs. These agents were hosted by Tencent's evaluation research team, and they targeted Alibaba's Amap.
    Download
    1867245.8K1.2M
  28. Alex Zhao✓@cocohearts
    "The Delah [Coffee] menu is, indeed, unique in the Bay Area’s cafe scene. The cinnamon and cardamom-scented Yemeni latte, available hot or iced, acts as the shop’s signature drink. The owners are tight-lipped about the exact recipe, but in short, it’s everything the pumpkin spice latte wants to be with none of the cloying sweetness." (july 2022) i'm a fan
    1024972
  29. Alex Zhao✓@cocohearts
    Alex Zhao
    RT @ShanuMathew93: interesting work from @nodiligence... intuitively something we all know but cool to see the figures spelled out. hard to keep track of all of the moving parts in the piece but it's worth a flip through. >"most of the third-party financing I can trace depends on six buyers, and the two weakest of them are close behind the one carrying the most. Meta still carries the biggest single pile of closed debt, about $51bn, but about $44bn of closed debt facilities depends on OpenAI, $41bn of it on its payments to Oracle, and Anthropic has about $36bn, almost all of it in one chip vehicle Broadcom stands behind." >"Of the 267 live projects I could find that name a customer, 194 lead back to one of six companies: OpenAI, Microsoft, Anthropic, Meta, Amazon and Google. Sometimes directly, sometimes through a reseller like Oracle, CoreWeave or Fluidstack that rents the capacity and sells it on." >"But capex only counts what they build and own themselves, plus their finance leases. A big and growing share of their capacity gets built by someone else and leased back to them, and a lease signed today for a building that opens in 2028 only shows up in a footnote until the day it starts. Add up that footnote across five buyers and you get about $252bn around the end of 2024 (Oracle’s earliest figure is May 2025, so I used that) and $1.12tn in the latest filings" >"Work back from each of the six buyers and add up every third-party debt facility that has closed against something it pays or guarantees, counted at its committed principal, and you get about $177bn... OAI and Ant carry 46% of the closed debt]"
    01907
  30. Show more posts