Super excited about all the possible applications for these new protein diffusion generative models from @namrata_anand2 nanand2.github.io/proteins/! The contrast between diffusion for 3D structures vs images is fascinating. Great to see this progress on the structural side.
Tudor Achim
235 posts
- Pleased to share that @HarmonicMath has raised $75M in Series A financing, led by @sequoia and with significant participation from @IndexVentures, along with firms like @GreatPointVC and Jasper Lau's Era Funds, and angels like @nikesharora and @JaredLeto. 1/
- I’m incredibly proud of the progress the Harmonic team has made thus far and even more excited about what’s ahead. This funding enables us to continue doing what we do best: explore the frontiers of human understanding. /1
- 👋 We started Harmonic less than a year ago to build Mathematical Superintelligence (MSI). Here is what we’ve done so far. 1/
- Couldn’t be more proud of our teamThis year, @HarmonicMath tested our advanced reasoning model Aristotle on the International Mathematics Olympiad (IMO), achieving Gold Medal-level performance and producing formally-verified proofs in Lean. We livestreamed our results from the HP Garage, check it out!
- Excited to see @namrata_anand2 work on generative AI models for protein design featured in the New York Times. Lots more to come from her new company, Diffuse Bio!
- Math is the essential part of reasoning. When we teach a model to do math, we are incepting the ability to reason. @sonyatweetybird and I had the pleasure of speaking to @tachim and @vladtenev about @HarmonicMath, an AI research lab with the goal of pushing the frontiers of
00:00 - Super proud of the team!We are excited to share three major updates today on our path to mathematical superintelligence 🦾 1. A new state-of-the-art of 90% on the MiniF2F benchmark. This beats our previously announced 83% from under a month ago.
- Replying to @tachimOur first result is Aristotle, a language model which advances the state of the art on MiniF2F, the standard dataset of formal mathematics problems statements. We are using Lean 4 for formal verification. 2/
- Replying to @tachimIf you’re interested in helping us accelerate the advent of Mathematical Superintelligence, check out our open roles at harmonic.fun 6/
- Replying to @tachimWe’re grateful to @kleinerperkins , @paradigm and our other new and repeat investors that have confidence our team is forging a meaningful path towards a future of verifiable and trustworthy intelligence. /2
- Replying to @tachimIf you’re passionate about AI that is verifiable, interpretable, and scalable, check out our open roles: jobs.ashbyhq.com/Harmonic /3
- Replying to @tachimAristotle currently achieves an 83% success rate with computer algebra assistance, and a 63% proof rate at producing full Lean proofs. For more detail, please see harmonic.fun/news.html 3/
- Replying to @tachimWe were able to achieve this state of the art performance predominantly with synthetic data, which gives us confidence that this approach can scale to true superintelligence. 4/





