A lot of folks who are having a bad experience are using GPT-5 in agent harnesses that aren't yet optimized for it. For every new model release, there's a time lag between release + when companies that integrate the model have it truly working well. Agent companies rush to switch the model param in their codebase to `gpt-5` or whatever the newest model is, and it works well enough, but it takes days to weeks to actually get the model running as intended (prompt, harness, tool tweaks are almost always necessary, to reflect how the model was trained). If you had a bad experience using GPT-5 in a coding harness (@cline, @cursor_ai, etc.), give it a week, and try again. I think you'll be pleasantly surprised.

REALMENTE emocionado de anunciar que Sunflower ha recaudado más de $ 3.88 millones ahora de @flybridge @ycombinator @speedrun. 🌻 Vamos a usar este $$ para construir el producto # 1 para los 1.2 mil millones de humanos que luchan contra la adicción. Recaudaremos más en YC DD en diciembre. De vuelta a la construcción ahora. 🫡