Good analogy! Would be interesting to read more details about how you’re getting very high quality outputs
It's basically (1) don't make decisions for the LLM - force it to think (2) make it explore extensively (3) make it pressure test it's own thinking.
alt Hacker News
It's basically (1) don't make decisions for the LLM - force it to think (2) make it explore extensively (3) make it pressure test it's own thinking.