3 ms·

I think the answer is obvious so maybe you're looking for something different. More perf means more attempts in parallel with some sort of arbiter model decidi

I think the answer is obvious so maybe you're looking for something different. More perf means more attempts in parallel with some sort of arbiter model deciding what to pick. This can happen at the token, prompt, or agent level or all of them.

the__alchemist 1y ago
This hits on a point I'm noticing more and more. I'll ask an LLM to QC a segment of code (Or ask for help, and it'll QC nearby parts). It suggests I change from A to B. I ask it another question later, or maybe troubleshoot. It tells me B is causing the problem, and I should do A instead.