Zihan Yan

PhD graduand in theoretical physics at Cambridge. Transitioning into AI research.

[Zihan Yan @physicscat0x7d]

My new paper "Dynamical entropy of charged black objects" with Manus Visser is now on arXiv!
https://t.co/nzujZCZ5pm

We develop a general, gauge-invariant framework for understanding the p-form electromagnetic potential-charge contributions to the first law of near-stationary black objects in general diffeomorphism invariant theories of gravity with (non-minimally and non-linearly coupled) p-form gauge fields. The charged black objects in consideration could have various horizon topology: holes, rings, folds, strings, branes, etc. We illustrate our framework with dyonic AdS black holes, dipole black rings, and charged black branes, and we look forward to its applications to string theory solutions with α'-corrections.

This is another work that I expected to finish with 20-30 pages but turned out to be 60+😂

[Zihan Yan @physicscat0x7d]

@ThomasVanRiet2 Ah, I forgot another factor: Focus. In industry you could do full-time research and no worries about teaching and admin stuff.

[Zihan Yan @physicscat0x7d]

Several angles here:
1. Scale: to speed up individual projects, there are plenty of good open-source agents + cheap SOTA subscriptions (GPT Pro, Claude Opus, etc.). But if you want a systematic sweep across millions of problems, academia just can't offer the compute or money.
2. Value: depends how you weigh academic freedom/openness + personal visibility vs. massive resources + huge package. At an AI company you can still publish or open-source, but in a limited, controlled way, though if they announce a megapackage for you, your personal visibility probably only goes up.
3. Frontier: some people join the big AI companies mostly to witness the whole revolution from the inside.

But overall, tenured academics can have both via a sabbatical at an AI company. Junior folks mostly have to pick one.

[Zihan Yan @physicscat0x7d]

@CuiMao 哈哈哈哈哈 可以

[Zihan Yan @physicscat0x7d]

@ShenRaphael Congratulations!

[Zihan Yan @physicscat0x7d]

Inclusiveness is most important. But I think a major source of tension is linguistic hegemony: inter-cultural communication is always *expected* to be in English by default everywhere in the world. People whose native language is not English always have to do extra work to be inclusive and to feel included. While we endeavour to make the most inclusive environment for all, I hope everyone could realise and acknowledge the extra effort involved.

[Zihan Yan @physicscat0x7d]

@CarinaLHong Impressive! Looking forward to more breakthroughs by you guys!

[Zihan Yan @physicscat0x7d]

@gary_shiu @UWMadPhysics @arnal_charles @Meta @f_charton @axiommathai Looks so cool. Will study it closely!

[Gary Shiu @gary_shiu]

It's my pleasure giving the opening talk of the Strings and Geometry Conference: https://t.co/ri2tYXMmpP where I presented joint work with Jacky Yip and Alessandro Mininno @UWMadPhysics and collaborators @arnal_charles @Meta @f_charton @axiommathai

[Zihan Yan @physicscat0x7d]

@Renet29304 @periodiclabs Many congratulations!

[Jiayi Weng @Trinkle23897]

Very exciting to see the cool result! Same pattern, different physics: Codex-grown heuristics matching or beating DRL agents in fluid dynamics, while staying readable, maintainable, and transferable.

Heuristics were not dead. They were under-maintained.

[Zihan Yan @physicscat0x7d]

@yiliuai Better than AI? Prove it and we will distill you.

[Zihan Yan @physicscat0x7d]

@Renet29304 @atheorist Thanks!

[Zihan Yan @physicscat0x7d]

@Renet29304 @atheorist Very cool! Looking forward to reading your new papers soon! I recently started a job in AI so barely had time to do physics. Hopefully I will write something about Cauchy slice holography or so with my AI agents when I have time…

[Zihan Yan @physicscat0x7d]

@LAkinjole Just curious, why are you all asking about “Zhang et al.”? Is that some magical buzzword?

[Zihan Yan @physicscat0x7d]

@BiblengCo I have already done my defence. The hard copy is the final version

[Zihan Yan @physicscat0x7d]

Finally submitted the hard copy of my PhD thesis!

[Zihan Yan @physicscat0x7d]

@undapav Can’t find a job in academia lol

[Zihan Yan @physicscat0x7d]

Any opinion on OpenCode+local model such as DeepSeek for heavy coding? Having token anxiety when using Claude Code & Codex…

[Zihan Yan @physicscat0x7d]

@quantum_short Yes. The stationary case is solved long ago. You could check Wald 1993 and related works.