Pavel

Data Scientist & Management Consultant
I build enterprise AI agents

[Pavel @baltabaev]

1. Yes, we can use Lean toautomatically verify some (though not all) math solutions

2. The complexity comes from the many layers of abstraction and concepts needed to understand these problems and proofs. The jump from 2+2=4 to linear algebra and calculus is smaller than the jump from linear algebra and calculus to this.

[Pavel @baltabaev]

I’ve spent well over 10,000 hours studying math in my life, yet I can’t understand these proofs, at least not without weeks of digging deep into each topic. What’s more, none of my math PhD friends know much about these problems either, and they can’t verify most of them without working directly in the field (yes, math is VERY diverse).

LLMs are getting smarter than the experts themselves, and I’m not sure we have enough bright human minds to verify everything that will come out of them in the coming years.

Remember when we compared AI intelligence to PhD students? I think we’re past that.

[Noam Brown @polynoamial]

An internal version of Astra, @OpenAI’s next major model family, solved 10 major open problems in mathematics, quantum complexity, and theoretical computer science.

We believe it will be a major step for scientific reasoning. https://t.co/iP6cyheZ7i

[Pavel @baltabaev]

@bigmrrobert In math some solutions can be expressed in a formal language (Lean) and automatically verified which helps to avoid the need of human verification but otherwise these problems level of expertise required is increasing dramatically

[Pavel @baltabaev]

@bigmrrobert There are people who can validate it in reasonable time but these experts are very few. Probably hundreds globally for most of these problems. Random math PhD won’t be able to tell within 1-2 hours because they work in a different domain.

[Pavel @baltabaev]

@SlavaOPs Don’t get me wrong, I don’t question these results. Lean is super helpful but my understanding that not all problems/solutions can be converted to Lean. And then how we do it beyond Math.

[Pavel @baltabaev]

I asked Fable to rank each problem from 1 (high school) to 10 (Fields medal) level and it graded all of them between 9 and 10.

[Pavel @baltabaev]

I’ve spent well over 10,000 hours studying math in my life, yet I can’t understand these proofs, at least not without weeks of digging deep into each topic. What’s more, none of my math PhD friends know much about these problems either, and they can’t verify most of them without working directly in the field (yes, math is VERY diverse).

LLMs are getting smarter than the experts themselves, and I’m not sure we have enough bright human minds to verify everything that will come out of them in the coming years.

Remember when we compared AI intelligence to PhD students? I think we’re past that.

[Pavel @baltabaev]

We had 2 quarters of financial markets penalizing companies for data center Capex and over indexing on short-term FCF. Based on MAG7 reports in the last 24 hours, I think it’s shifting.

[Pavel @baltabaev]

@AndrewCurran_ Gpt-5.6 Sol was the best OAI release in the last year and it shows

[Pavel @baltabaev]

@pitdesi Funny enough, we almost hit our friend’s Tesla at a very similar spot in SF using FSD.

The worst part that it tried to park in a stranger’s garage vs. road.

[Pavel @baltabaev]

@Noahpinion This is satire / fake

[Pavel @baltabaev]

Remember OpenClaw?

[Pavel @baltabaev]

@MollySOShea @friedberg @RoKhanna Can't agree more with David on the tax policy

[Pavel @baltabaev]

@DmitryRybin1 The “think macrohard make no mistakes” meme went out of control with this one. Amazing.

[Pavel @baltabaev]

@buccocapital

[Pavel @baltabaev]

Regardless of what kind of prompt chain and tool usage led to this discovery, it’s very impressive for an LLM to discover such counterexample given it’s not something you can grid search.

I feel like we’re experiencing another major Leap in AI intelligence with Fable / Sol 5.6 / Kimi K3 after Opus 4.5 release last year.

[levent @__alpoge__]

hello there the jacobian conjecture is false thanx to my close friend akhil for asking about it and my other close friend fable for working during the world cup final

((1+xy)^3 z + y^2 (1+xy) (4+3xy), y + 3 x (1+xy)^2 z + 3 x y^2 (4+3xy), 2 x - 3 x^2 y - x^3 z): \C^3\to \C^3, has jacobian determinant -2, and sends (0, 0, -1/4), (1, -3/2, 13/2), and (-1, 3/2, 13/2) to (-1/4, 0, 0)

[Pavel @baltabaev]

@TheLaikYobaz @nikitabier please stop people from uploading images as videos to farm likes

[Pavel @baltabaev]

@stevehou True! There are years of agentic build and adoption to come to reach true potential.

[Pavel @baltabaev]

@alex STEM PhD*

[Pavel @baltabaev]

@hollyoxcoburn @SenSanders Not saying it's solving anything, but this is the worst tax alternative out there, and here's my argument. A wealth tax is simply a populist move: the goal isn't to fund social programs, it's to make people feel "justice," fueled by jealousy.