San Francisco, CA
- Tweets 554
- Following 1,087
- Followers 6,087
- Likes 3,535
Conversation with my father.
Me: We should make a no AI writing to family rule
Dad: If I own a beautiful shirt, why should I settle for wearing something plain?
Me: I disagree with this characterization; I think there's something very human about writing to one another—imperfections and all—that gets erased when you plaster it with the makeup of AI.
Especially because AI's rouge is so recognizable.
Dad: *reacted 🙏 to your last message*
5
83
12,792
Previewing Ultrafast mode for @OpenAI's GPT 5.6 Sol, powered by Cerebras.
GPT-5.6 Sol Ultrafast generates responses at up to 750 tokens per second. That's the full, GPT-5.6-Sol model -- up to 14× faster than the same model on Standard processing.
It speedran Humanity’s Last Exam in 11h 11m, nearly 7× faster than Claude Fable 5 with comparable accuracy.
157
363
4,447
1,289,928
It's hard to find as A tier of a crew as this... Jess and Rakesh were some of the brightest folks I had the pleasure of meeting at Harvard! Very excited for them & Moritz & the team :)
Announcing Hone
Intelligence has become abundant. Yet the world looks remarkably similar to how it did five years ago. With every model release, the gap between what frontier AI can do and the economic value derived from it widens.
Closing the gap requires re-organizing work around organizational outcomes, not individual tasks. Hone builds AI that creates, orchestrates, and improves agents and software continuously to own organizational outcomes over weeks and months.
We are ex-founders and early core contributors to Cognition, Mercor, Ramp, and OpenAI. We obsess over real-world value, not theoretical benchmarks.
Our core beliefs on closing the gap in the thread below.
2
62
10,303
milks & matcha & mechanisms & markets
Can AI outperform human quant traders while inheriting our biases?
@jeffreygwang of @OpenAI connects game theory, mechanism design and language-model research to explain why even a powerful reasoning model may not be the perfect trader.
Big Chip Club with @cerebras // @alyciazcary
2
1
32
4,188
“Getting the maximum amount that you can out of a given chip is a super high priority for us.”
@jeffreygwang of @OpenAI explains why training performance depends on moving parameters efficiently and reducing communication overhead, not just arithmetic.
8
9
136
173,103
Jeffrey Wang retweeted
This is a surreal moment. Few people could have predicted that the AI will advance to solving math problems at the highest level only a few years after GPT-2/3. The models then couldn't reliably solve grade school math problems. They barely were good enough to draft emails. They still very much felt like statistical parrots.
All things that looked like fundamental limitations slowly faded with some advances (e.g. high scale RL) in a span of a few years, which is really a short time. We should behold this moment both in awe and disbelief. What will a few more years of progress bring? How is it going to impact the world and society at large? Are we ready for the tsunami of intelligence at our fingertip?
More than any other moment, this feels to me like the eve of singularity. Glasswing & huggingface incident further increase the gravity.
A few years ago, deep down I felt working on alignment is premature. It's nice to do if it's your passion, but the shapes of things weren't clear enough for it to be critical in my opinion. The chances that you end up working on things that are useless for aligning the actual AGI was high. It's different now. Now, it feels like it's the most critical thing facing us.
openai.com/index/ten-advance…
41
86
967
245,863
Jeffrey Wang retweeted
yes, nonsofic groups exist: this statement is one of many new beautiful results proved by Astra, our next major model.
We're releasing 10 such Astra proofs, complete with lean certificates and CoT walkthroughs for each of them. The results are wide-ranging, from von Neumann algebras (disproof of Connes' Rigidity Conjecture) to better bounds for high dimensional sphere packing, for circuit complexity, for monochromatic triangles in multicolored graphs, and more.
More thoughts here: openai.com/index/ten-advance…
277
944
6,678
4,279,578
Jeffrey Wang retweeted
Don't be a doomer. It is possible to get AI alignment right. It is possible to build a society where everyone benefits from this technology, one without disempowerment and upsetting inequality. And the future this would bring, if we succeed, makes it worth trying.
5
6
50
3,304
real ones trained their first NNs in octave 🤠
From training his first neural net in high school to working at OpenAI.
@jeffreygwang followed a five-year trail through Andrew Ng’s course, early GAN papers + OpenAI research.
By the time ChatGPT arrived, “it all pointed toward OpenAI.”
Presented with @cerebras // @alyciazcary
1
1
39
7,523
was amazing chatting with Sarah and the @cerebras team!
(visceral excitement about a future of ultra-fast, ultra-intelligent inference)
“Whatever hardware allows that to happen is the hardware that we want.”
@jeffreygwang from @OpenAI on a future where agents consume far more tokens.
Cerebras is building the infrastructure to deliver those tokens at speed.
1
4
48
6,296
GPT 5.6 Pro finds counterexample to 30 yr old open problem via 22-word initial prompt to find a counterexample, followed by three prompts to, in effect, "keep going."
Dinitz-Garg-Goemans conjecture is false. This graph theory problem was open for ~30 years.
The graph below has fractional flow cost 58. Any unsplittable flow (with capacity violation <=15) has cost at least 60.
Chat with GPT 5.6 Pro where this was found: chatgpt.com/share/6a60b2eb-0…
1
22
3,721
When I first moved to SF almost two years ago, Tinah was already prototyping and ideating what has now become Andera.
Since then, their trajectory has gone vertical! It's been a privilege to watch Andera grow—the team, the company, and the scope of their ambition.
Tech has decided that regulation is the enemy of progress, but innovation requires capital
Regulation builds the trust necessary to inject billions into markets
To enable free innovation, Andera raised a $37M series A led by Lightspeed to scale financial oversight with agents
3
33
4,747
Nim, Linda, and the Sable team have been killing it since their Harvard days. An incredible team here—go Sable!
Introducing Aidan, your newest AI employee
Aidan is the first computer-using AI built for realtime conversation
We’ve raised $45M from @sequoia and @8VC to bring Aidan to the world
@NotionHQ and @DecagonAI already use Aidan to run customer interactions. This is how @withsableai works:
2
29
4,034
Jeffrey Wang retweeted
Today we’re releasing DeepSWE, a new standard for agentic coding benchmarks.
On public leaderboards, top models often look relatively close in capability. DeepSWE shows where they actually diverge, reflecting the realistic experience of developers in their day-to-day work.
516
729
6,032
1,984,399
neo is a fabulous community, full of camaraderie and was home to many of my first friends in sf. cannot recommend applying enough :)
🏆 Neo Scholar applications are open!
Are you a college student who excels at CS?
Follow in the footsteps of Neo Scholars who founded Cursor, Chai Discovery, Applied Compute, Flint, Cognition, & more.
Apply to join one of tech’s strongest communities. neo.com/scholars
3
2
28
6,267