推文

@harden_369741 · 2026-10-12 14:17

stole a prompting trick from the Chinese AI community and it's embarrassingly effective. three gears: MAX reasoning to plan (dump everything, voice notes and all), MEDIUM to execute (but only if your spec is detailed), MAX again to review. the review pass catches ~20% of issues at a third of the cost of regenerating. I've been running MAX on everything like a caveman. never again. #PromptEngineering #AIagents #LLM

曝光 12 · 评论 0 · 点赞 0 · 书签 0 · 曝光/时 9.88483086359983