👋We’re excited to launch DSG2-mini, our newest AI protein design model, now available in DiffuseSandbox⏳🎁, our new app for protein binder design. Click through to design a protein yourself! https://t.co/h8CKF4kwGI 1/

It’s wild to me that you can do this on a public API. This is the exact same endpoint that got 10 on the Putnam straight from the natural language questions that humans got. We’re going from math copilots to math autopilots 🚀🚀

BREAKING: Consumer version of Aristotle smashes Putnam, solving and formally verifying 10/12 problems in @leanprover. Congrats to amateur mathematician @namrata_anand2, who used the consumer-grade version of Aristotle with an early public release of the problems. It appears that

Check out the verified proofs here https://t.co/bZgIcYaI9A 2/

Aristotle from @HarmonicMath got 10/12 problems within 7 hours on Putnam, the world’s hardest undergrad math competition! How do I know this? Well, I ran Aristotle myself via Harmonic’s public API. And it took me about 10 minutes to get started and kick it off! 9 out of 10

I made it into Terry Tao’s blog! https://t.co/n466FaLpTT One cool part of this experience is that I *would not have made the Claude Deep Research query resulting in the connection to Erdos 106 if not for Aristotle’s exact implementation*. i.e. Aristotle, an AI tool, contributed

@elnomic I've always been interested in the big questions, like ▪️was there one interaction that unified gravity with the other forces, and what was it like? ▪️what happened before the Big Bang? ▪️can every even number be written as the sum of exactly two primes, and if so, why? These

Aristotle

Nice guest post on Buzzard’s blog by Boris Alexeev on his and others’ use of AI systems to attack Erdos-type problems (thanks to Bloom’s @thomasfbloom website collating them): https://t.co/XBgHLly41H In particular, with systems like Harmonic’s @HarmonicMath Aristotle,

BREAKING: “Aristotle has now written the majority of the public formalized solutions on https://t.co/IkaHak8lfY.” — Boris Alexeev, mathematician

Aristotle has learned to code

Been very interesting to see the system's capabilities evolve! You can check these results for yourself using the public API :)

Beyond math: Aristotle achieves SOTA 96.8% proof generation on VERINA: Benchmarking Verifiable Code Generation. You can read more about this performance on our engineering blog linked in bio

Congrats to @fthumpasery, @Samuelklam, and @PermitFlow on their $54M Series B! One of the biggest roadblocks in construction is the gnarly permitting process—dealing with hundreds of documents, thousands of details, and hyper-localized rules that aren't easily accessible.

Another one bites the dust! Last night, @AcerFur used Aristotle from @HarmonicMath to autonomously prove in @leanprover Erdos Problem #481, which had been open for over 45 years. A systematic approach targeting all unsolved Erdos problems is in the works.

Another one bites the dust! BTW, anyone can try this at home. Just sign up for an API key on our website.

Aristotle proved another Erdos problem autonomously last night, concurrently with @AcerFur, and at @AcerFur's request. As far as we are aware, this is not an easier version of another problem Erdos intended. Before @AcerFur published his proof, he made two requests to Aristotle:

The pace of progress in AI for mathematics right now is incredible, and we're excited that #LeanLang provides the foundation for results that can be trusted, shared, and extended. Huge congratulations to @vladtenev @tachim and @HarmonicMath!

@AcerFur @HarmonicMath oh interesting. that means Aristotle independently solved the problem without your proof

I have seen some takes that the problem that @HarmonicMath’s Aristotle recently solved was competition-level, maybe even trivial. Maybe so! I’d be interested to see it administered on an exam. Another data point though: Gemini Deep Think messed it up:

We are on the cusp of a profound change in the field of mathematics. Vibe proving is here. Aristotle from @HarmonicMath just proved Erdos Problem #124 in @leanprover, all by itself. This problem has been open for nearly 30 years since conjectured in the paper “Complete sequences