[robrenaud]

I think you underestimate how powerful lean is, and close it is to the tedious part of formal math. A theorem prover needs consult no outside resource. A formal math LLM-like generator need only consult the theorem prover to get rid of hallucinations. This is why it's actually much easier than SWE to optimize/hill climb on.

Low level, automated theorem providing is going to fall way quicker than most expected, like AlphaGo, precisely because an MCTS++ search over lean proofs is scalable/amendable to self play/relevant to a significant chunk of professional math.

Legit, I almost wish the US and China would sign a Formal Mathematics Profileration Treaty, as a sign of good will between very powerful parties who have much to gain from each other. When your theorem prover is sufficiently better than most Fields medalists alive, you share your arch/algorithms/process with the world. So Mathematics stays in the shared realm of human culture, and it doesn't just happen to belong to DeepMind, OpenAI, or Deepseek.

[baq]

On the contrary I think we're low key on the verge of model checkers being widely deployed in the industry. I've been experimenting with Opus 4.5 + Alloy and the preliminary results I'm getting are crossing usability thresholds in a step-function pattern (not surprising IMHO), I just haven't seen anyone pick up on it publicly yet.

The workflow I'm envisioning here is the plan document we're all making nowadays isn't being translated directly into code, but into a TLA+/Alloy/... model as executable docs and only then lowered into the code space while conformance is continuously monitored (which is where the toil makes it not worth it most of the time without LLMs). The AI literature search for similar problems and solutions is also obviously helpful during all phases of the sweng process.

[robot-wrangler]

> The workflow I'm envisioning here is the plan document we're all making nowadays isn't being translated directly into code, but into a TLA+/Alloy/... model as executable docs and only then lowered into the code space while conformance is continuously monitored

I'm sure we've agreed on this before, but I agree again ;) There are dozens of us at least, dozens! There's also a recent uptick in posts with related ideas, for example this hit the front-page briefly ( https://news.ycombinator.com/item?id=46251667 ).

I was tempted to start with alloy/tla for my own experiments along these lines due to their popularity, but since the available training data is so minimal for everything in the space.. I went with something more obscure (MCMAS) just for access to "agents" as primitives in the model-checker.

[baq]

> available training data is so minimal for everything in the space

Haven't tried anything other than Alloy yet, but I've got a feeling Anthropic has employed some dark arts to synthesize either Alloy models or something closely related and trained Opus on the result - e.g. GPT 5.1 commits basic syntax errors, while Opus writes models like it's just another day at the office.

[gaigalas]

Is it trivial for any mathematician to understand lean code?

I'm curious if there is a scenario in which a large automated proof is achieved but there would be no practical means of getting any understanding of what it means.

I'm an engineer. Think like this: a large complex program that compiles but you don't understand what it does or how to use it. Is such a thing possible?

[JulianWasTaken]

It's not trivial for a mathematician to understand Lean code, but it's something that's possible to learn to read and interpret in a day (without then necessarily being proficient in how to write it).

That's true though of Lean code written by a human mathematician.

AI systems are capable (and generally even predisposed to) producing long and roundabout proofs which are a slog to decipher. So yes the feeling is somewhat similar at times to an LLM giving you a large and sometimes even redundant-in-parts program.

[gus_massa]

With very difficult human generated proof, it's common that it take like 10 or 20 to make it understandable for mortals. The idea is to split the proof, create new notation, add intermedite steps that are nice, find a simpler path. It's like refactoring.

Sometimes the original proof is compleyely replaced, bit by bit, until there is an easy to understand version.

[gus_massa]

Too late to edit:

"10 or 20" -> "10 or 20 years"

[gaigalas]

Wow!

If curl developers are overwhelmed by AI PRs, imagine how mathematicians will feel verifying a huge backlog of automated proofs.

Or isn't there such a thing? Can someone make a very complicated automated proof that ultimately reveals itself to be useless?

[gus_massa]

For math it's easy, everyone just ignore it. There is no Daniel to blame. There are a few post about P=/!=NP or the Riemann conjeture or rewriting physics each week that are posted to HN. I'm just ignoring them. Other mathematicians are just ignoring them. But you will not find anyone to blame.

There are a few "solutions" of conjetures that may be correct, like https://en.wikipedia.org/wiki/Abc_conjecture I'm not sure about the current state. There may be a few mathematicians trying to read some parts, or perhaps no. Perhaps in a few years the easy parts will be refactored and isolated, and published as special cases. And after a while, it may be verified or someone will find a gap and perhaps fix it. Just wait a few decades.

> Can someone make a very complicated automated proof that ultimately reveals itself to be useless?

It depends, on what you consider insightful. Take a look at "Determination of the fifth Busy Beaver value" https://news.ycombinator.com/item?id=45273999 in particular the first comment. Is that an heroic result that opens a lot of paths or a useless combination of tricks that no one will ever understand? (In my opinion a proof is a proof [standing applause emoji].)

[gaigalas]

What I (personally) consider insightful is irrelevant. It's about what mathematicians consider insightful.

Mathematicians are obviously not ignoring automated proofs. Terry's post is an evidence of that.

Consider LK99 instead of crackpot P vs NP proofs. That wasted a lot of academia time.

It seems that it could happen to math.

[QuesnayJr]

There's a couple of problems that were solved that way a while ago, and they have been formalized, but not in Lean:

https://en.wikipedia.org/wiki/Four_color_theorem

https://en.wikipedia.org/wiki/Kepler_conjecture

[voxl]

Yes.. yes.. sure, of course... You neglect this one little detail: theorem proving IS programming. So if an AI can be "better than a fields medalist" (a laughable claim akin to basically calling it AGI) then it will be better than all software engineers everywhere.

But see you neglect something important: it's the programmer that is establishing the rules of the game, and as Grothendieck taught us already, often just setting up the game is ALL of the work, and the proof is trivial.

[robrenaud]

What is harder, beating Lee Sedol at Go, or physically placing stones on a Go board? Which is closer to AGI?

Because AlphaGo can only do one.

AI could very well be better at formal theorem proving than fields medalists pretty soon. It will not have taste, ability to see the beauty in math, or pick problems and set directions for the field. But given a well specified problem, it can bruteforce search through lean tactics space at an extremely superhuman pace. What is lacks in intuition and brilliance, it will make up in being able to explore in parallel.

There is a quality/quantity tradeoff in search with a verifier. A superhuman artificial theorem prover can be generating much worse ideas on average than a top mathematician, and make up for it by trying many more of them.

It's Kasparov vs DeepBlue and Sedol vs AlphaGo all over.

It's also nowhere near AGI. Embodiment and the real world is super messy. See Moravec's paradox.

Practical programs deal with the outside world, they are underspecified, their utility depends on the changing whims of people. The formal specification of a math problem is self contained.

[riku_iki]

> But given a well specified problem, it can bruteforce search through lean tactics space at an extremely superhuman pace. What is lacks in intuition and brilliance, it will make up in being able to explore in parallel.

right, but because math is not well specified and formalized yet, it could be a problem, and that's where humans with intuition and more rigid reasoning still necessary.

[voxl]

Your analogy completely misses the point. What is harder? Designing a game that has relevant implications in physical reality, or playing a game already designed given an oracle that tells you when you make an incorrect move?