Kevin Weil &#x1f1fa;&#x1f1f8;(@kevinweil)

The new codex computer use is shockingly good

Kevin Weil 🇺🇸 (@kevinweil) — Fri, 17 Apr 2026 05:02:23 GMT

The new codex computer use is shockingly good

Ari Weinstein@AriX

This is the first time I've ever seen an LLM operate a GUI as fast as a person, and it's surreal.

Your browser does not support the video tag.

💬13🔄7❤️337👀43719📊36

See our blog post for videos and demos: https://t.co/YDE0XkFuUO Request access to GPT-Rosalind here...

Kevin Weil 🇺🇸 (@kevinweil) — Thu, 16 Apr 2026 19:37:34 GMT

See our blog post for videos and demos: openai.com/index/introduc…

Request access to GPT-Rosalind here: openai.com/form/life-scie…

Life Sciences plugin: github.com/openai/plugins…

💬3🔄5❤️62👀8815📊14

💥 Super excited to launch GPT-Rosalind, our first frontier model built for scientific research acro...

Kevin Weil 🇺🇸 (@kevinweil) — Thu, 16 Apr 2026 19:37:34 GMT

💥 Super excited to launch GPT-Rosalind, our first frontier model built for scientific research across biology, drug discovery, and translational medicine. 

The model is trained in chemistry, protein engineering, genomics and more, including built-in knowledge of relevant databases and tools researchers use.

We want to make these capabilities available to scientists and researchers while maintaining strong safeguards against biological misuse. GPT-Rosalind is launching through a trusted access deployment structure for qualified customers. 

We're also launching a Life Sciences plugin for Codex to everyone today, which works with our mainline models as well as GPT-Rosalind.

💬32🔄69❤️900👀73663📊143

Could not agree more with @gdb.

Kevin Weil 🇺🇸 (@kevinweil) — Tue, 14 Apr 2026 06:10:15 GMT

Could not agree more with @gdb.

Greg Brockman@gdb

The world is transitioning to a compute-powered economy.

The field of software engineering is currently undergoing a renaissance, with AI having dramatically sped up software engineering even over just the past six months. AI is now on track to bring this same transformation to every other kind of work that people do with a computer.

Using a computer has always been about contorting yourself to the machine. You take a goal and break it down into smaller goals. You translate intent into instructions. We are moving into a world where you no longer have to micromanage the computer. More and more, it adapts to what you want. Rather doing work with a computer, the computer does work for you. The rate, scale, and sophistication of problem solving it will do for you will be bound by the amount of compute you have access to.

Friction is starting to disappear. You can try ideas faster. You can build things you would not have attempted before. Small teams can do what used to require much larger ones, and larger ones may be capable of unprecedented feats. More and more, people can turn intent into software, spreadsheets, presentations, workflows, science, and companies.

People are spending less energy managing the tool and more energy focusing on what they are actually trying to create. That shift brings a kind of joy back into work that many people haven’t felt in a long time. Everyone can just build things with these tools.

This is disruptive. Institutions will change, and the paths and jobs that people assumed were stable may not hold. We don’t know exactly how it will play out and we need to take mitigating downsides very seriously, as well as figuring out how to support each other as a society and world through this time. But there is something very freeing about this moment. For the first time, far more people can become who they want to become, with fewer barriers between an idea and a reality. OpenAI’s mission implies making sure that, as the tools do more, humans are the ones who set their intent and that the benefits are broadly distributed, rather than empowering just one or a small set of people.

We're already seeing this in practice with ChatGPT and Codex. Nearly a billion people are using these systems every week in their personal and work lives. Token usage is growing quickly on many use-cases, as the surface of ways people are getting value from these models keeps expanding.

Ten years ago, when we started OpenAI, we thought this moment might be possible. It’s happening on the earlier side, and happening in a much more interesting and empowering way for everyone than we’d anticipated (for example, we are seeing an emerging wave of entrepreneurship that we hadn’t previously been anticipating). And at the same time, we are still so early, and there is so much for everyone to define about how these systems get deployed and used in the world.

The next phase will be defined by systems that can do more — reason better, use tools better, plan over longer horizons, and take more useful actions on your behalf. And there are horizons beyond, as AI starts to accelerate science and technology development, which have the potential to truly lift up quality of life for everyone. All of this is starting to happen, in small ways and large, today, and everyone can participate. I feel this shift in my own work every day, and see a roadmap to much more useful and beneficial systems. These systems can truly benefit all of humanity.

💬3🔄9❤️225👀24036📊21

This is a great take. Human intelligence and computer intelligence are not different points on a lin...

Kevin Weil 🇺🇸 (@kevinweil) — Mon, 13 Apr 2026 02:04:30 GMT

This is a great take. Human intelligence and computer intelligence are not different points on a line, but in a high dimensional space—and we should explore more of the space.

Michał Podlewski@trajektoriePL

Terence Tao proposes what he calls a "Copernican view of intelligence". 

Instead of buying into the common, one-dimensional narrative that artificial intelligence will simply evolve from "subhuman" to "superhuman" and ultimately make humanity entirely redundant, Tao urges us to look at the bigger picture.

Much like the Copernican revolution proved the Earth is not the center of the universe, Tao suggests we need to realize that human intelligence isn't the only, or necessarily the highest, form of intellect. Historically, we have treated other forms of storing or creating knowledge—like animals, books, and computers—as secondary. However, we actually exist within a much richer universe of intelligence.

Both human intelligence and computer intelligence possess their own distinct strengths and weaknesses. The true potential lies not in viewing them as direct competitors, but rather in focusing on collaboration. By working together, humans and computers can achieve additional things that neither could accomplish on their own, requiring us to think in much wider terms than just what humans or computers can do alone.

Your browser does not support the video tag.

💬10🔄27❤️229👀27621📊34

Five Erdos problems at once! The proofs are getting more elegant as the models improve 👀

Kevin Weil 🇺🇸 (@kevinweil) — Thu, 09 Apr 2026 02:55:25 GMT

Five Erdos problems at once! The proofs are getting more elegant as the models improve 👀

Mehtaab Sawhney@mehtaab_sawhney

We’ve just released another paper solving five further Erdős problems with an internal model at OpenAI: arxiv.org/abs/2604.06609.

Several of the proofs were especially enjoyable to digest while writing the paper. My personal favorite was the solution to Erdős Problem 1091. The question asks: if a graph G has chromatic number 4, while every small subgraph has chromatic number at most 3, must it contain an odd cycle with many diagonals? The internal model gives a very enlightening counterexample to this conjecture, and the proof was a pleasure to understand.

For those so inclined, a really fun exercise is to try to reconstruct the proof from Figure 5 of the paper, which was of course produced by Codex.

💬5🔄11❤️258👀23545📊26

Btw: @hemal built this as a simple skill in a few hours, since Prism is powered by Codex. More to co...

Kevin Weil 🇺🇸 (@kevinweil) — Tue, 07 Apr 2026 19:01:00 GMT

Btw: @hemal built this as a simple skill in a few hours, since Prism is powered by Codex. More to come.

And if you love the idea of building AI-native products to accelerate science, we're hiring a lead designer for Prism: apply here! jobs.ashbyhq.com/openai/7e265ae…

Kevin Weil 🇺🇸@kevinweil

💥 New in Prism today: Paper Review, an AI workflow for reviewing technical and scientific papers. 

This is the opposite of AI slop: we're using AI to improve scientific rigor, correctness, and reproducibility.

💬6🔄7❤️119👀19389📊20

Try it out at https://t.co/4MBs45PNiU today and let me know what you think!

Kevin Weil 🇺🇸 (@kevinweil) — Tue, 07 Apr 2026 17:48:20 GMT

Try it out at prism.openai.com today and let me know what you think!

💬1🔄6❤️59👀6300📊12

Paper Review is built to act more like a careful technical reviewer than a grammar checker. It looks...

Kevin Weil 🇺🇸 (@kevinweil) — Tue, 07 Apr 2026 17:48:19 GMT

Paper Review is built to act more like a careful technical reviewer than a grammar checker. It looks for issues in math, derivations, notation, units, structure, and whether a paper’s claims are actually supported by the results. It also catches inconsistencies across sections, figures, tables, captions, and appendices. The goal is to help authors strengthen technical rigor and correctness before submission.

Inside Prism, it runs as a structured workflow and writes its output as an editable LaTeX review file directly in the project. That means the review becomes part of the paper workspace itself, so authors can revise and work through feedback in the same environment as the manuscript. It also flags unclear phrasing, citation issues, and proofreading mistakes before submission.

💬2🔄8❤️85👀8398📊13

💥 New in Prism today: Paper Review, an AI workflow for reviewing technical and scientific papers. ...

Kevin Weil 🇺🇸 (@kevinweil) — Tue, 07 Apr 2026 17:48:19 GMT

💥 New in Prism today: Paper Review, an AI workflow for reviewing technical and scientific papers. 

This is the opposite of AI slop: we're using AI to improve scientific rigor, correctness, and reproducibility.

💬60🔄86❤️855👀229665📊216

Not only is AI solving more open problems—its proofs are getting more elegant as the models improve

Kevin Weil 🇺🇸 (@kevinweil) — Wed, 01 Apr 2026 04:38:05 GMT

Not only is AI solving more open problems—its proofs are getting more elegant as the models improve

Mehtaab Sawhney@mehtaab_sawhney

We are excited to share a new paper solving three further problems due to Erdős; in each case the solution was found by an internal model at OpenAI. Each proof is short and elegant, and the paper is available here: arxiv.org/pdf/2603.29961

💬24🔄13❤️267👀18257📊38

"An open source workflow for producing a personalized mRNA cancer vaccine"

Kevin Weil 🇺🇸 (@kevinweil) — Sun, 29 Mar 2026 20:29:55 GMT

"An open source workflow for producing a personalized mRNA cancer vaccine"

Yishan@yishan

My friend @philfung was inspired by the man who built a personalized cancer vaccine for his dog, so he wrote a guide to DIY mRNA vaccine production.  

Phil used to run a lab startup, and the guide covers the entire process - from sequencing to synthesis, using open-source software and benchtop lab equipment.  

Note: This is for educational purposes only and is not intended for medical use

um unless you have cancer

💬9🔄60❤️667👀73822📊119

Paul used ChatGPT + AlphaFold to create a personalized mRNA vaccine protocol for his dog's cancer. ...

Kevin Weil 🇺🇸 (@kevinweil) — Fri, 27 Mar 2026 06:44:18 GMT

Paul used ChatGPT + AlphaFold to create a personalized mRNA vaccine protocol for his dog's cancer. 

This is a glimpse of the future, with AI accelerating personalized medicine. We have to make it easier for people to do this for people, not just dogs!

Paul S. Conyngham@paul_conyngham

x.com/i/article/2036…

💬20🔄14❤️190👀37818📊34

👀

Kevin Weil 🇺🇸 (@kevinweil) — Wed, 25 Mar 2026 14:16:16 GMT

👀

Jared Duker Lichtman@jdlichtman

In the latest paper of Terry Tao

💬3🔄6❤️134👀11092📊14

I’m inspired by young people putting frontier AI to work in science in new and surprising ways. We ...

Kevin Weil 🇺🇸 (@kevinweil) — Mon, 23 Mar 2026 15:01:00 GMT

I’m inspired by young people putting frontier AI to work in science in new and surprising ways.

We recently met a young astronomer who used AI to analyze massive datasets. He identified 1.5 million new celestial objects in space, and the head of NASA offered him a job and a ride in a fighter jet as a signing bonus! (x.com/rookisaacman/s…)

To celebrate stand-out students and recent grads like this—and help keep their ideas moving—we’re launching a new program: ChatGPT 26. If you're selected, we'll host you at OpenAI HQ in SF, give you access to our latest tech, and give you a $10K cash grant to keep pushing your ideas forward.

If this sounds like you, we’d love to meet you and hear your story. Apply or nominate someone here: chatgpt26.com

💬20🔄44❤️493👀65374📊105

Technology is so rad. @neuralink is amazing.

Kevin Weil 🇺🇸 (@kevinweil) — Mon, 23 Mar 2026 06:42:24 GMT

Technology is so rad. @neuralink is amazing.

Jon L. Noble🇬🇧@CheckCanopy

It’s hard to believe it’s already been 100 days since I received my Neuralink N1 implant. Looking back, the whole journey feels like science fiction that somehow became my everyday reality.

The surgery on Day 0 was surprisingly easy. A quick general anaesthetic, a small incision, and the robotic system did the rest — precisely placing the 1,024 ultra-thin threads into my motor cortex. I woke up alert and in good spirits and went home the next afternoon. By Day 3 I was feeling a lot better, and by Day 7 the little scar was already starting to fade. Recovery was genuinely minimal; I felt sharper and more positive than I had been in years after the BCI was turned on.

The real fun started in Week 2 when we paired the implant with my brand-new Apple MacBook (my very first Mac). The @neuralink engineers walked me through calibration sessions, and within a couple of minutes I was moving the cursor just by thinking. At first it felt like trying to remember a dream, but by Week 3 it was second nature. Scrolling, clicking, typing — all mind-controlled. The Mac integration was buttery smooth; I went from total Mac newbie to power-user faster than I ever expected.

By Day 80 I was ready for the big leagues. That’s when I fired up @Warcraft of Warcraft for the first time with pure thought control. The first raid felt clunky, but once my brain and the BCI synced, it was pure magic. I’m now raiding, and exploring Azeroth hands-free at full speed — no mouse, no keyboard, just intention. It’s honestly brilliant. The freedom is addictive.

The social-media side has been just as surprising. Every update I’ve shared has been met with genuine excitement rather than scepticism. Thousands of messages from people with disabilities, gamers, students, and scientists — all asking real questions about the tech and what it could mean for the future. The positivity has been overwhelming and incredibly motivating.

100 days in and I already can’t imagine life without it. The N1 didn’t just give me a new way to use a computer — it gave me a new way to live. Can’t wait to see what the next 100 days bring.

Thank you all so much for your support and I will keep you all updated as we continue this journey together.

Your browser does not support the video tag.

💬1🔄2❤️103👀16807📊9

I'll put a lot of money on the over, @jpatel41 :)

Kevin Weil 🇺🇸 (@kevinweil) — Tue, 17 Mar 2026 09:57:43 GMT

I'll put a lot of money on the over, @jpatel41 :)

Jeetu Patel@jpatel41

Thanks for the kind words @rohanvarma. The partnership between @Cisco and @OpenAI has been nothing short of fabulous. Especially over the past 75 days. Our team is pretty stoked with the progress being made with the use of Codex. 

Let’s keep pushing on both sides. Appreciate you leaning in. The goal is to have 6 products 100% written with AI by end of 2026 and 70% of our products 100% written with AI by end of 2027. @kevinweil thinks I am sandbagging. I hope to prove him right ;-).

💬2🔄0❤️65👀15389📊6

This is the right take.

Kevin Weil 🇺🇸 (@kevinweil) — Sun, 15 Mar 2026 20:57:36 GMT

This is the right take.

Trung Phan@TrungTPhan

💬9🔄9❤️331👀15201📊31

This is beautiful, and true

Kevin Weil 🇺🇸 (@kevinweil) — Sat, 14 Mar 2026 03:39:17 GMT

This is beautiful, and true

maddie rune🪰@maddierune

No one tells you that parenting is just relearning the world through someone who thinks worms are friends & birds are miracles. It’s the most healing thing I’ve ever done. My daughter looked out the window this morning & said, everything is green & growing. I told her, you too. And something inside me whispered, so are you. Now I’m watching her hold flowers up to the sun while the light bends like it recognizes her. It’s funny, every spring I think I’m teaching my child about the world & every spring she proves she’s the one teaching me how to see it.

💬3🔄5❤️155👀43389📊16

So excited to work together! I have a feeling it's going to be a productive summer :)

Kevin Weil 🇺🇸 (@kevinweil) — Tue, 10 Mar 2026 22:01:09 GMT

So excited to work together! I have a feeling it's going to be a productive summer :)

Acer@AcerFur

I guess now is as good a time as any to announce that I shall be joining the AI for Science team at @OpenAI this summer. This has been in the works since January, and I thank @SebastienBubeck and @kevinweil for their personal interest in making this happen.

💬1🔄7❤️185👀20917📊15

If true, this would be the first of @EpochAIResearch's Frontier Math open problems to be resolved by...

Kevin Weil 🇺🇸 (@kevinweil) — Tue, 10 Mar 2026 14:37:44 GMT

If true, this would be the first of @EpochAIResearch's Frontier Math open problems to be resolved by AI.

"The result emerged from a single GPT-5.4 Pro run and was subsequently refined into Lean with GPT-5.4 XHigh which ran for a few hours."

spicylemonade@spicey_lemonade

We believe we have fully resolved, in Lean and python, one of @EpochAIResearch  Frontier Math open problems: a Ramsey-style problem on hypergraphs.

The result emerged from a single GPT-5.4 Pro run and was subsequently refined into Lean with GPT-5.4 XHigh which ran for a few hours.

github.com/spicylemonade/…@Jsevillamol

💬15🔄36❤️509👀75540📊65

A look at the future/present

Kevin Weil 🇺🇸 (@kevinweil) — Tue, 10 Mar 2026 04:04:25 GMT

A look at the future/present

Andrej Karpathy@karpathy

Three days ago I left autoresearch tuning nanochat for ~2 days on depth=12 model. It found ~20 changes that improved the validation loss. I tested these changes yesterday and all of them were additive and transferred to larger (depth=24) models. Stacking up all of these changes, today I measured that the leaderboard's "Time to GPT-2" drops from 2.02 hours to 1.80 hours (~11% improvement), this will be the new leaderboard entry. So yes, these are real improvements and they make an actual difference. I am mildly surprised that my very first naive attempt already worked this well on top of what I thought was already a fairly manually well-tuned project.

This is a first for me because I am very used to doing the iterative optimization of neural network training manually. You come up with ideas, you implement them, you check if they work (better validation loss), you come up with new ideas based on that, you read some papers for inspiration, etc etc. This is the bread and butter of what I do daily for 2 decades. Seeing the agent do this entire workflow end-to-end and all by itself as it worked through approx. 700 changes autonomously is wild. It really looked at the sequence of results of experiments and used that to plan the next ones. It's not novel, ground-breaking "research" (yet), but all the adjustments are "real", I didn't find them manually previously, and they stack up and actually improved nanochat. Among the bigger things e.g.:

- It noticed an oversight that my parameterless QKnorm didn't have a scaler multiplier attached, so my attention was too diffuse. The agent found multipliers to sharpen it, pointing to future work.
- It found that the Value Embeddings really like regularization and I wasn't applying any (oops).
- It found that my banded attention was too conservative (i forgot to tune it).
- It found that AdamW betas were all messed up.
- It tuned the weight decay schedule.
- It tuned the network initialization.

This is on top of all the tuning I've already done over a good amount of time. The exact commit is here, from this "round 1" of autoresearch. I am going to kick off "round 2", and in parallel I am looking at how multiple agents can collaborate to unlock parallelism.
github.com/karpathy/nanoc…

All LLM frontier labs will do this. It's the final boss battle. It's a lot more complex at scale of course - you don't just have a single train. py file to tune. But doing it is "just engineering" and it's going to work. You spin up a swarm of agents, you have them collaborate to tune smaller models, you promote the most promising ideas to increasingly larger scales, and humans (optionally) contribute on the edges.

And more generally, *any* metric you care about that is reasonably efficient to evaluate (or that has more efficient proxy metrics such as training a smaller network) can be autoresearched by an agent swarm. It's worth thinking about whether your problem falls into this bucket too.

💬13🔄7❤️232👀26688📊30

Agh, GPT 5.4 I mean 🤦‍♂️

Kevin Weil 🇺🇸 (@kevinweil) — Sat, 07 Mar 2026 04:07:21 GMT

Agh, GPT 5.4 I mean 🤦‍♂️

💬0🔄0❤️27👀3045📊2

This is a pretty mind-blowing look inside GPT-4 and its reasoning abilities. Definitely worth your t...

Kevin Weil 🇺🇸 (@kevinweil) — Sat, 07 Mar 2026 00:37:36 GMT

This is a pretty mind-blowing look inside GPT-4 and its reasoning abilities. Definitely worth your team to read.

Hanson Wang@hansonwng

x.com/i/article/2029…

💬7🔄1❤️135👀33218📊21

Codex for Open Source is an awesome idea. OSS maintainers get API credits, 6 months of ChatGPT Pro...

Kevin Weil 🇺🇸 (@kevinweil) — Fri, 06 Mar 2026 19:20:11 GMT

Codex for Open Source is an awesome idea. 

OSS maintainers get API credits, 6 months of ChatGPT Pro with Codex, and access to Codex Security as needed.

OpenAI Developers@OpenAIDevs

We’re launching Codex for Open Source to support the contributors who keep open-source software running.

Maintainers can use Codex to review code, understand large codebases, and strengthen security coverage without taking on even more invisible work.

developers.openai.com/codex/communit…

Your browser does not support the video tag.

💬3🔄7❤️127👀14920📊14

💥 GPT 5.4 is launching today! It's our best model ever, and it's also the most capable scientific m...

Kevin Weil 🇺🇸 (@kevinweil) — Thu, 05 Mar 2026 18:18:23 GMT

💥 GPT 5.4 is launching today! It's our best model ever, and it's also the most capable scientific model we've ever released. 

GPT 5.4 Pro in particular is 🤯 based on early testing with scientists and mathematicians.

💬63🔄28❤️829👀41208📊131

We integrated the Codex harness into Prism (https://t.co/cmEsJYRtW9) — this means you get skills, re...

Kevin Weil 🇺🇸 (@kevinweil) — Wed, 04 Mar 2026 19:21:34 GMT

We integrated the Codex harness into Prism (prism.openai.com) — this means you get skills, reasoning levels, and the raw tenacity of the Codex model in your LaTeX environment. 

Oh and we also built version mgmt into Prism, which was one of the top requests. 

See below thread from @vicapow for some great examples of the power of Codex inside Prism 👇

Victor Powell@vicapow

🧵1/ We've brought the most advanced AI to Prism by introducing Codex to Prism. Prism is already the best place for scientific writing to happen—and with Codex, now you can write, compute, analyze, and iterate all in one place.

Your browser does not support the video tag.

💬8🔄23❤️330👀40856📊46

💥 AI accelerating high energy physics Just a few weeks after the gluon scattering paper, this mor...

Kevin Weil 🇺🇸 (@kevinweil) — Wed, 04 Mar 2026 18:14:42 GMT

💥 AI accelerating high energy physics 

Just a few weeks after the gluon scattering paper, this morning we posted the more complicated graviton scattering analogue. See below for more from @ALupsasca 👇

Alex Lupsasca@ALupsasca

We just posted a new preprint: “Single-minus graviton tree amplitudes are nonzero.”

Yes: a helicity sector long assumed to vanish in quantum gravity can actually appear under well-defined kinematics. Preprint: cdn.openai.com/graviton/gravi…

💬9🔄17❤️174👀20223📊25

🇺🇸

Kevin Weil 🇺🇸 (@kevinweil) — Sat, 28 Feb 2026 05:57:06 GMT

🇺🇸

Sam Altman@sama

Tonight, we reached an agreement with the Department of War to deploy our models in their classified network.

In all of our interactions, the DoW displayed a deep respect for safety and a desire to partner to achieve the best possible outcome.

AI safety and wide distribution of benefits are the core of our mission. Two of our most important safety principles are prohibitions on domestic mass surveillance and human responsibility for the use of force, including for autonomous weapon systems.  The DoW agrees with these principles, reflects them in law and policy, and we put them into our agreement.

We also will build technical safeguards to ensure our models behave as they should, which the DoW also wanted. We will deploy FDEs to help with our models and to ensure their safety, we will deploy on cloud networks only.

We are asking the DoW to offer these same terms to all AI companies, which in our opinion we think everyone should be willing to accept. We have expressed our strong desire to see things de-escalate away from legal and governmental actions and towards reasonable agreements.

We remain committed to serve all of humanity as best we can. The world is a complicated, messy, and sometimes dangerous place.

💬54🔄0❤️201👀24827📊49

So excited for the new world where everyone can build

Kevin Weil 🇺🇸 (@kevinweil) — Wed, 25 Feb 2026 21:23:46 GMT

So excited for the new world where everyone can build

Brian Armstrong@brian_armstrong

Beware, it's 2026 and the CEO is back in the code base thanks to AI.

💬18🔄1❤️159👀16363📊23

Instantly one of the great sports photos of all time.

Kevin Weil 🇺🇸 (@kevinweil) — Sun, 22 Feb 2026 18:47:38 GMT

Instantly one of the great sports photos of all time.

SportsCenter@SportsCenter

🇺🇸 🇺🇸 🇺🇸

💬7🔄15❤️588👀40578📊48

Daniel has a really interesting perspective on the future of Math and AI—well worth a read 👇

Kevin Weil 🇺🇸 (@kevinweil) — Sat, 21 Feb 2026 20:25:35 GMT

Daniel has a really interesting perspective on the future of Math and AI—well worth a read 👇

Daniel Litt@littmath

Some thoughts on AI and mathematics, inspired by "First Proof."

💬8🔄10❤️133👀14222📊22

More on the gluon scattering/GPT 5.2 paper from @ALupsasca below 👇 If you're in the Boston area on...

Kevin Weil 🇺🇸 (@kevinweil) — Mon, 16 Feb 2026 15:39:58 GMT

More on the gluon scattering/GPT 5.2 paper from @ALupsasca below 👇

If you're in the Boston area on Tuesday, go see his lecture at Harvard!

Alex Lupsasca@ALupsasca

Our recent preprint on gluon amplitudes has sparked a lot of discussion, so I want to share the backstory — including how AI helped crack a problem that had stumped us for a year.

I'll also be giving a public lecture at Harvard this week. Details at the end.

💬4🔄3❤️80👀9312📊10

AI 🤝 Mathematics. The First Proof challenge was a great idea and I'm looking forward to seeing the ...

Kevin Weil 🇺🇸 (@kevinweil) — Sat, 14 Feb 2026 04:52:06 GMT

AI 🤝 Mathematics. The First Proof challenge was a great idea and I'm looking forward to seeing the results! 

Great color on OpenAI's submission from @merettm below 👇

Jakub Pachocki@merettm

Very excited about the "First Proof" challenge. I believe novel frontier research is perhaps the most important way to evaluate capabilities of the next generation of AI models.

We have run our internal model with limited human supervision on the ten proposed problems. The problems require expertise in their respective domains and are not easy to verify; based on feedback from experts, we believe at least six solutions (2, 4, 5, 6, 9, 10) have a high chance of being correct, and some further ones look promising.

We will only publish the solution attempts after midnight (PT), per the authors' guidance - the sha256 hash of the PDF is d74f090af16fc8a19debf4c1fec11c0975be7d612bd5ae43c24ca939cd272b1a .

This was a side-sprint executed in a week mostly by querying one of the models we're currently training; as such, the methodology we employed leaves a lot to be desired. We didn't provide proof ideas or mathematical suggestions to the model during this evaluation; for some solutions, we asked the model to expand upon some proofs, per expert feedback. We also manually facilitated a back-and-forth between this model and ChatGPT for verification, formatting and style. For some problems, we present the best of a few attempts according to human judgement.

We are looking forward to more controlled evaluations in the next round!

1stproof.org#1stProof

💬6🔄6❤️183👀10081📊18

Not gonna lie it's kinda amazing to see gluon scattering trending on Twitter... my worlds colliding ...

Kevin Weil 🇺🇸 (@kevinweil) — Fri, 13 Feb 2026 21:00:25 GMT

Not gonna lie it's kinda amazing to see gluon scattering trending on Twitter... my worlds colliding 

x.com/i/trending/202…

💬9🔄8❤️236👀10944📊32

👇💥

Kevin Weil 🇺🇸 (@kevinweil) — Fri, 13 Feb 2026 19:46:08 GMT

👇💥

Patrick OShaughnessy@patrick_oshag

I spent last night with Andrew Strominger and Alex Lupsasca, two of the top physicists in the world

They just released a paper, co-authored with OpenAi, that seems to me like ASI

Andrew, who helped develop string theory, told me that a year ago, his view was that he didn’t know how helpful AI was going to be.  

A year later, after some back and forth with GPT 5.2 pro, they submitted a final query to an internal model which solved AND proved a previously unsolved problem in quantum field theory…in 12 hours. 

A model, doing something two of the smartest people in the world in their field couldn’t do. And, when I was with them, they were giddy with excitement for what might lay ahead. 

Andy said “It is the first time I’ve seen AI solve a problem in my kind of theoretical physics that might not have been solvable by humans.”

They said, “two things changed: the model improved and we figured out how to talk to it.”

Andy also told me “I also now feel that with the recent advances, most physicists who want to keep up with the frontiers of progress will need to learn how to talk to it. That wasn’t true a year ago.”

ASI is here, just not evenly distributed.

💬2🔄2❤️76👀10339📊8

The preprint is here: https://t.co/DSXMu4kLEQ More on the OpenAI blog here: https://t.co/7DjxupEZ0Z...

Kevin Weil 🇺🇸 (@kevinweil) — Fri, 13 Feb 2026 19:12:01 GMT

The preprint is here: arxiv.org/pdf/2602.12176

More on the OpenAI blog here: openai.com/index/new-resu…

Also: I'm so lucky to get to work with @ALupsasca.

💬2🔄8❤️88👀6009📊14

GPT 5.2 derived a new result in theoretical physics. For decades it's been assumed that certain gl...

Kevin Weil 🇺🇸 (@kevinweil) — Fri, 13 Feb 2026 19:12:00 GMT

GPT 5.2 derived a new result in theoretical physics. 

For decades it's been assumed that certain gluon amplitudes ("single minus") were zero, and that the maximally helicity violating amplitudes had two gluons of one helicity and n-2 of the other. It turns out that isn't necessarily true!

Andy Strominger realized this a year ago, and with Alfredo Guevara, David Skinner, and @ALupsasca they had shown this up to n=6 by hand. The expressions were getting incredibly complicated though, and Alex invited them to OpenAI to see what we could do together. 

In short order, GPT-5.2 Pro suggested a beautiful and general formula for arbitrary n—but couldn't prove it. An internal scaffolded model, thinking continuously for over 12 hours, proved it. 

This is exciting because when complicated calculations reduce to something simple, it implies there is yet-to-be-understood physics waiting to be discovered. There are multiple follow-ons to this paper, and we hope other physicists (maybe with AI!) explore the implications as well. 

One particular fun thing for me was getting to work with Andy Strominger, someone I looked up to throughout my time as a physics student in undergrad and grad school. To hear him talk about how he's been accelerated by AI was incredibly motivating. 

AI 🤝 Physics. Here's to bringing the science of the future into the present!

💬29🔄71❤️631👀45079📊111

Coding at 1000 tokens/sec is a mind-expanding experience. You have to try this.

Kevin Weil 🇺🇸 (@kevinweil) — Thu, 12 Feb 2026 18:25:43 GMT

Coding at 1000 tokens/sec is a mind-expanding experience. You have to try this.

OpenAI@OpenAI

GPT-5.3-Codex-Spark is now in research preview.

You can just build things—faster.

Your browser does not support the video tag.

💬24🔄16❤️626👀94531📊73

Could not agree more. Curious for everyone's take on how we best address this, given the relative sp...

Kevin Weil 🇺🇸 (@kevinweil) — Tue, 10 Feb 2026 05:19:14 GMT

Could not agree more. Curious for everyone's take on how we best address this, given the relative speed of model improvement vs peer review.

Kevin Roose@kevinroose

i am begging academics to study AI capabilities using frontier models. 

the models used in this study (which is going to be cited for years as proof that "AI is bad at health advice") are GPT-4o, Llama 3, and Command R+, two obsolete models and one i've never heard of.

💬33🔄8❤️193👀34218📊40

Can confirm

Kevin Weil 🇺🇸 (@kevinweil) — Mon, 09 Feb 2026 16:33:33 GMT

Can confirm

Sam Altman@sama

.@SebastienBubeck is so so good; incredible researcher and leader.

💬6🔄3❤️262👀26212📊23

Coding models are amazing, but "GPT-5 lowers the cost of cell-free protein synthesis by 40%" is also...

Kevin Weil 🇺🇸 (@kevinweil) — Thu, 05 Feb 2026 22:29:24 GMT

Coding models are amazing, but "GPT-5 lowers the cost of cell-free protein synthesis by 40%" is also pretty rad. This was all done in a robotic lab in partnership with @Ginkgo. AI + Science!

OpenAI@OpenAI

We worked with @Ginkgo to connect GPT-5 to an autonomous lab, so it could propose experiments, run them at scale, learn from the results, and decide what to try next. That closed loop brought protein production cost down by 40%.

Your browser does not support the video tag.

💬8🔄18❤️354👀20393📊37

So many AI graphs are in log scale, you forget how wild it really is.

Kevin Weil 🇺🇸 (@kevinweil) — Thu, 05 Feb 2026 03:34:02 GMT

So many AI graphs are in log scale, you forget how wild it really is.

💬28🔄91❤️1019👀133493📊157

🤣🤣🤣

Kevin Weil 🇺🇸 (@kevinweil) — Thu, 05 Feb 2026 03:27:31 GMT

🤣🤣🤣

Jon Matzner@MatznerJon

This is either going to be the best or worst idea I've ever had.

Hooked up my OpenClaw to all of our internet connected cameras at the house.

Got this this (OUT OF NOWHERE) this morning.

💬9🔄10❤️323👀56823📊35

AI doesn't just accelerate science in big ways, it accelerates it in a thousand incremental compound...

Kevin Weil 🇺🇸 (@kevinweil) — Wed, 04 Feb 2026 15:34:04 GMT

AI doesn't just accelerate science in big ways, it accelerates it in a thousand incremental compounding ways too.

Derya Unutmaz, MD@DeryaTR_

I've now created, using the OpenAI Codex app, a massive Cancer Mutation Catalog of driver mutations! It contains more than 90,000 mutations across 1,600 genes. You can filter by cancer type, limited now, but will soon expand it to include all cancer types
cancer-mutations-derya.netlify.app

💬10🔄16❤️238👀26683📊29

If you're a scientist and/or write in LaTeX, try Prism at https://t.co/4MBs45PNiU and let us know wh...

Kevin Weil 🇺🇸 (@kevinweil) — Tue, 03 Feb 2026 03:28:04 GMT

If you're a scientist and/or write in LaTeX, try Prism at prism.openai.com and let us know what you think! 

Taking feedback and feature requests here and turning them into code daily with Codex.

OpenAI@OpenAI

Much of today’s scientific tooling has remained unchanged for decades. Prism changes that.

@ALupsasca joins @kevinweil and @vicapow to walk through what it looks like when GPT-5.2 works inside a LaTeX project with full paper context.

Your browser does not support the video tag.

💬26🔄13❤️196👀27120📊37

I've gotten to the point where if I'm sitting in a meeting and I don't have a Codex prompt running, ...

Kevin Weil 🇺🇸 (@kevinweil) — Mon, 02 Feb 2026 20:19:16 GMT

I've gotten to the point where if I'm sitting in a meeting and I don't have a Codex prompt running, I feel anxious for "wasting" an hour I could have been making parallel progress.

It's happened super quickly and totally changed how I think.

💬88🔄53❤️1239👀117065📊219

💥 The Codex app launches today, first for Mac and soon Windows. It's awesome and I can't recommend ...

Kevin Weil 🇺🇸 (@kevinweil) — Mon, 02 Feb 2026 20:14:33 GMT

💥 The Codex app launches today, first for Mac and soon Windows. It's awesome and I can't recommend it highly enough. 

I've "written" more code in the last few weeks than I have in years and shipped multiple bug fixes and feature requests in Prism with it. It's been so much fun.

💬21🔄2❤️225👀22427📊40