• [object Object]@lemmy.ca
    link
    fedilink
    English
    arrow-up
    21
    ·
    2 days ago

    Inline completions are genuinely useful, I’m mostly replacing them with local models though. They are slower but free as in beer (once you pay the hardware cost).

    • sobchak@programming.dev
      link
      fedilink
      English
      arrow-up
      1
      ·
      11 hours ago

      I tried AI auto-complete for maybe 6 months. I found that mistakes and hard to find bugs increased, actually lowering my productivity. I find that actually just typing everything (well, with traditional LSP) gives me more time to think about my code, the architecture, etc while I’m typing, leading to better results. “Slow is smooth, smooth is fast.”

      • [object Object]@lemmy.ca
        link
        fedilink
        English
        arrow-up
        4
        ·
        edit-2
        1 day ago

        I’m using vim with minuet-ai, and it plugs the AI suggestion into my completion module. I found the Copilot style virtual text interfaces all janky.

      • [object Object]@lemmy.ca
        link
        fedilink
        English
        arrow-up
        5
        ·
        2 days ago

        Sure, but hopefully small code completion (2-4b range) models can run locally on a lot of things. They’re just less good.