> What's going on with this developer message vs. system message?
Yeah, this. I was surprised I had to find the o1 release thread via the search box; I expected it to stay on the frontpage for couple days, accumulating gazillion points.
Anyway, that question is the most important and most worrying part to me. It basically reads like they're building up a hierarchy of prompt importance, where the API users are now demoted to second-tier role, while the most important system prompt will be under control of whoever is serving the model (i.e. OpenAI, or Microsoft via Azure, or Apple). This... sucks.
What is this poorly documented reasoning_effort parameter and how can I use it?
Why is re-enabling markdown part of the prompt and not a parameter?
What's going on with this developer message vs. system message?
Any thoughts about the model spec's chain of command? https://cdn.openai.com/spec/model-spec-2024-05-08.html#follo...
Please HN, don't make me use X.