@dtgeek @bsky.app@bsky.brid.gy is currently down But here’s one of their devs bragging about Claude writing all their code: https://fed.brid.gy/r/https://bsky.app/profile/did:plc:vpkhqolt662uhesyj6nxm7ys/post/3meomclcfss2w Post provided via @bsky.brid.gy@bsky.brid.gy mirror of course as, well, Bluesky is down.

Out of the 10 recent problems OpenAI solved, the existence of a non-sofic group lies in my area of math. It answers a genuinely impressive question, however the framing "problems that have seen no progress on the main result for at least a decade" in their blogpost is dishonest, as quite a bit of theory has recently been building up. One of their paper's main ingredients (Proposition 2.3) seems to be a generalization of results by Kun and Thom from 2019. This is especially ironic since in their blogpost OpenAI claims to have deep respect for [..] the Leiden declaration on AI and mathematics, where correct attribution is explicitly listed as one of the main points. Francesco Fournier-Facio wrote a fantastic note on the missing context of this discovery. He also gives another (short) construction of a non-sofic group which may be more natural to geometric group theorists. The note can be found on Francesco's webpage: https://www.fffmaths.com OpenAI blogpost: https://openai.com/index/ten-advances-in-mathematics #math #ai #sofic #geometricgrouptheory #openAi

Salesforce published a detailed writeup on going agentic with Claude Code. A couple things jumped out. A migration they'd scoped at 231 days shipped in 13. One PR delivered 21 endpoints at 100% test coverage.