Aletheia tackles FirstProof autonomously
2026/02/24 by Tony Feng, Junehyuk Jung, Sang-hyun Kim +14 · 5 voices
#cs.AI #cs.CL #cs.LG
paper · pdf
Abstract
We report the performance of Aletheia (Feng et al., 2026b), a mathematics research agent powered by Gemini 3 Deep Think, on the inaugural FirstProof challenge. Within the allowed timeframe of the challenge, Aletheia autonomously solved 6 problems (2, 5, 7, 8, 9, 10) out of 10 according to majority expert assessments; we note that experts were not unanimous on Problem 8 (only). For full transparency, we explain our interpretation of FirstProof and disclose details about our experiments as well as our evaluation. Raw prompts and outputs are available at https://github.com/google-deepmind/superhuman/tree/main/aletheia.
Citations
Discussions
- Pretty wild Ai result, that will probably get less press than the latest Ai boyfriend article: arxiv.org/abs/2602.21201 A benchmark was created with professional mathematicians, using problems natural [bsky, 150 points, 8 comments]
- Google Aletheia autonomously solved 6/10 of FirstProof challenge research-level problems. Those are so hard that it took many days for experts to evaluate solutions. One (P7) was an open problem for s [bsky, 23 points, 1 comments]
- Aletheia Tackles FirstProof Autonomously [hn, 5 points, 0 comments]
- arxiv.org/abs/2602.21201 [bsky, 3 points, 0 comments]
- here is the paper arxiv.org/pdf/2602.21201 , now it may still be possible that leaders of the field like tao, scholze etc may still be able to continue the work but i think for rest of us the machines [bsky, 1 points, 1 comments]
Related