|
|
|
|
|
|
by yonran
88 days ago
|
|
|
Seems like a lot of people’s problems with AI come from talking to the dumber models and having it not provide sufficient proof that it fixed a bug. Maybe instead of banning AI, projects should set a minimum smarts level. e.g. to contribute, you must use gpt-5.4-codex high or better for either writing it or code reviewing it. |
|
|
I use Opus 4.6 almost exclusively and it still generates nonsense if I don’t guide it.