Local LLMs perform better when you teach them to ask before they answer(www.xda-developers.com)
9 pointsby froh2 hours ago3 comments
  • thinkingemote2 hours ago
    From the article: "When tasked with coding, writing, editing, or summarizing, ask the user up to three targeted clarifying questions. Proceed with the task once you've received answers and understand the prompt fully. If the task is a simple factual question or conversational message, respond directly."
  • riknos3142 hours ago
    I started using similar approaches in the sonnet 3.5 era and found them incredibly useful at the time. The frontier lab models have gotten significantly better about their guesses over time, but I still sometimes turn to the technique if my own ideation is only about 80% of the way there, as the LLM's questioning can help me identify the blind spots that need more consideration.
  • froh2 hours ago
    I'm positively surprised such a little guidance makes such a difference.

    is it also useful with the smaller (and cheaper) cloud models?

    • intothemild30 minutes ago
      Yes. I run local models, Qwen3.6-27B and IMHO the massive level up was the agents and skills files that I've worked on.

      Basically I run a flow

      Brainstorming > Create Spec > Review Spec* > Create Plans > Review Plan* > Execute Plan (in subagents) > Review Against Plan > Code Review* > Open PR > Finish Plan (marks plan files done)

      * Each review step marked with an asterisk uses a paid larger LLM, right now Deepseek V4 Pro. Having it do this catches a lot of small things, and now I'm effectively one shotting any task I give it.

      And it's not costing me much at all, just those three reviews. I could use a free model like Gemini but I'm happy with what I've got.