Pythos：作者打造可核查數學答案的 AI，並免費提供

What if you built a badass math AI, gave it a full simulation lab and graphing tools, and then made it all free?

Well, I did.

Meet Pythos.

I started building Pythos because I wanted to tackle a problem with AI and mathematics: language models can explain concepts beautifully, but they can also make algebra mistakes while sounding completely confident.

For students trying to learn, that matters. A convincing explanation isn't worth much if the answer is wrong.

So I started building something different.

More than another math chatbot

Pythos combines an AI math tutor with deterministic verification engines designed to independently check mathematical claims.

The idea is straightforward: let the AI handle explanations and tutoring, while dedicated mathematical tools verify the claims they can formally prove.

Depending on the problem, that can involve symbolic computation, algebraic checks, domain restrictions, and other deterministic methods.

When a claim can't be verified, the system needs to distinguish that from a confirmed result. That's a core engineering challenge I'm continuing to work on.

The learning tools matter, too

I didn't want to build a box that spits out answers and calls itself an educational platform.

Pythos also brings together graphing tools and a simulation lab, giving students ways to explore mathematical ideas and experiment with problems instead of simply copying solutions.

The bigger goal is to make it useful as a learning environment, not just an answer generator.

Why make it free?

Because students already have enough barriers to education.

I want people to be able to use these tools without another subscription standing between them and learning something new.

Pythos is part of my work under LANZAR. It's an ongoing project, and I'm continuing to improve the verification pipeline, tutoring behavior, and overall experience.

I'm not claiming it's finished or infallible. Building a system that can reliably distinguish a correct answer from a plausible-looking mistake is an ongoing engineering problem—and one I think is worth tackling.

Try it yourself

Pythos: https://pythos.lanzar.me

I'd genuinely like feedback from developers, math enthusiasts, students, and educators.

Try a difficult problem. Test the explanations. Look for edge cases. If something breaks, I want to know about it.

I'm especially interested in feedback on deterministic verification, how AI tutoring should work, and what features would make this more useful in a real classroom.

I built it because I thought math education tools could be better.

Now I want to see what people do with it.