OpenAI models have no agency anymore

I don't know if it's a token saving measure or to bait more interactivity from users but it feels like out of nowhere the models have just became useless with zero agency. If you don't have a goal plugin that constantly prods them, there is a very high chance it just doesn't do anything and responds back like you're asking it a question.

OFTEN it ends sentences with "I'll do XXXXX." or "I'll implement XXXXX." and STOPS where the next turn should be a tool call or something. Because of this I genuinely thought it was a Pi or opencodex regression stopping the turn right before an expected tool call. But no, it's because the models are incredibly lazy now.

This came out of nowhere.

---

Comments URL: https://news.ycombinator.com/item?id=50435490

Points: 3

# Comments: 1