Thrilled to share I’ve joined OpenAI for Science, a new team building AI systems to advance scientific reasoning and accelerate discovery in math and physics. 🧵
As a theoretical physicist, I used to think it would be a long time before AI could touch the research frontier. Then GPT-5 Pro arrived, and it completely changed my mind.
This summer I wrote a paper on new symmetries in black-hole perturbation theory:
It is relatively easy to unfold the physical implications of these symmetries once they are known, but the hard part is finding their precise form (or if they even exist).arxiv.org/abs/2506.05298
It is relatively easy to unfold the physical implications of these symmetries once they are known, but the hard part is finding their precise form (or if they even exist).arxiv.org/abs/2506.05298
This took me days of tough calculation, as you might guess from the complicated form of the generators of these new symmetries (Eq. 7 of the paper): 
After GPT-5 Pro launched, I gave it that same problem. To my utter shock, it rediscovered the result in <30min!
See for yourself:
It’s not flawless (it needs priming on the flat-space case before tackling the full problem) but the leap is incredible.chatgpt.com/share/68b006eb…
See for yourself:
It’s not flawless (it needs priming on the flat-space case before tackling the full problem) but the leap is incredible.chatgpt.com/share/68b006eb…
And it’s not just theory. GPT-5 Pro can also tackle hard questions in observational astrophysics, producing answers that would take a bright grad student several days of research time: chatgpt.com/share/68d154c9…
These examples and many others have convinced me: AI will transform scientific research.
There’s still huge room to grow, and I’m excited to help push those boundaries at OpenAI for Science!
There’s still huge room to grow, and I’m excited to help push those boundaries at OpenAI for Science!
• • •
Missing some Tweet in this thread? You can try to
force a refresh
