I came across this commentary on the AZsPCs programming contest list. https://simonwillison.net/2026/Sep/8/on-navier-stokes/
Two mathematicians, Tristan Buckmaster at NYU and Levant Alpöge at Anthropic had made significant progress on Navier Stokes including solving the Euler case for non viscous liquids. They had not yet published their results because they were still iterating their paper with Claude (Anthropic) and Codex (OpenAI). After the OpenAI announcement, Buckmaster quickly posted their unfinished, partially AI generated papers. https://mastodon.social/@tristanbuckmaster/117233413705701198 It's over 200 pages in 3 papers. You can read their AI statement on page 7 of the Bousinnesq paper. OpenAI said they would decline the $1 million Clay prize. Simon Willison estimates they spent $15 million on compute, based on 300 million output tokens at Astra 6 prices. This is happening on the Hutter prize too. Data compression developers openly share their work on the encode forum, then a huge surge of progress, then people accusing others of stealing their work when prize money is involved. I think the big news is 10,000 agents essentially each doing a lifetime's worth of Ph.D. level work in 88 hours at less than 1/1000 of the cost. Imagine reading this thread 4 years ago. It would have been pure science fiction. -- Matt Mahoney, [email protected] ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/Tf866419d43752e9c-M6a9b81408600550e83c081f6 Delivery options: https://agi.topicbox.com/groups/agi/subscription
