Did Ilya just pop the bubble?

We need to do research, to go back to ideas, yet virtually no frontier labs have the courage to step away from the rat race to do meaningful research.

PhD student in my office today: "I am realizing that people just want theory even though it's useless theory".

I’m applying to PhD programs and will be at the NeurIPS workshops! I’m interested in generalization theory, interpretability, and reasoning. Hope to meet profs and students working in these areas! Please see my website (hrithikravi.github.io) for more details about my

I totally buy that AI has made you more productive. And I buy that if other lawyers were more agentic, they could also get more productivity gains from AI. But I think you're making my point for me. The reason it takes lawyers all this schlep and agency to integrate these models

Cristiano Ronaldo Perplexity

Last night, Ankit Gupta @ NeurIPS and I hosted a great dinner with 14 professors at #NeurIPS2025 from leading academic labs across the US, and many cited compute in academia as "abhorrent". Out of curiosity I just pulled these stats. This is insane. To do meaningful AI research today you need

Age of Research would be a beautiful name for an AI winter.

adding floats with different precisions gives different results depending on element order (a+b)+c !== a+(b+c) GPUs batch and schedule operations based on workload, which adds variance to element order therefore outputs will vary even with 0 temp x.com/SebbeBroman/st…

Trump Announces 5,000% Increase In All Numbers theonion.com/trump-announce…

Remote workers at 1:37 PM after replying "Sounds good" on Slack

"I still remember the first morning I biked to university in Copenhagen, the February air reddening my cheeks. I was thrilled to be on an exchange semester overseas, but I saw it as just a detour from my imagined career path. I didn’t realize I was already pedaling toward a

I repeat: We have to split lectures / practice / exams and people should be allowed to pick and mix whatever elements from whatever academic institution. In particular universities must allow LLM-self-taught people to get "real world credentials".

will ye Holy shit bro this is so based. You guys grind so hard. So crazy the waitress knew how to debug. Only in sf brother 🤙

Claude Shannon revolutionized the world not by just grinding, but also by playing. Research demands the confidence to "waste time" pursuing the new and unknown. His work underpins everything from hardware to AI. Had he lived to see the era he built, he’d be a Nobel laureate.

If a model can not afford to simulate the atoms, it is forced to invent the 'chemistry' to predict the outcome. Check out this fascinating new preprint that provides an information theoretical foundation along with quantifiable measures "From Entropy to Epiplexity" by Marc Finzi

In that moment, I decided to stop shutting off my curiosity. I would become an applied mathematician, follow my random intellectual interests (wherever they took me) without worrying about whether they were "pure" enough. One of the best things that ever happened to me. (10/10)

Can you read 900 words per minute? Try it.

On 15 January, 1950, John Nash published a one-page paper, “Equilibrium Points in N-Person Games,” in the Proceedings of the National Academy of Sciences. Based on his PhD work and citing only two references, it introduced the Nash equilibrium and later earned him the Nobel Prize