thilakan

@tao@mathstodon.xyz are there any parallels in software engineering now that software generation has become abundant?

@tao I just watched Bubeck’s IPAM talk on GPT-5 coming up with a proof of a nontrivial extremal combinatorics problem. He talks about having designed a specific harness for the AI to work long term and that producing several thousands of pages of intermediate steps that could possibly contain novel ideas. What are your thoughts on this experiment?