• Lung@lemmy.world
    link
    fedilink
    English
    arrow-up
    4
    ·
    4 days ago

    I don’t get it, it’s just a markdown file, which is how llm specs already get written. I’m supposed to trust that the llm will “compile” it correctly every time without touching the code output? What happens when the code gets huge and complex, a “recompile” entails waiting for ages to get something similar to what others compiled? What am I missing?

    • Joe@discuss.tchncs.de
      link
      fedilink
      English
      arrow-up
      1
      ·
      edit-2
      3 days ago

      It looks like something you could run make against, building and testing one module at a time based on per-module specs.

      Cool, I guess. Except for the environment whenever someone runs make, as that’ll get slightly warmer… which might be nice in winter, if localized to my dungeon home office.

      edit: An interesting aspect seems to be the change cycle… it must detect when a spec change has occured (git diff?), and generates targeted changes so the code matches the spec. No spec change, no LLM calls, no code change. Problem in the generated code? Fix the spec instead…