• Wrrzag@lemmy.ml
    link
    fedilink
    arrow-up
    4
    ·
    4 days ago

    Dude, writing code is the fun part, reviewing code written by others is time-consuming, difficult and boring. Why would you automate the fun out of coding and force yourself to review even more shit?

  • Carrot@lemmy.today
    link
    fedilink
    arrow-up
    11
    ·
    5 days ago

    I’m not strictly against using LLMs for coding. However, every time I try, it makes me waste my time. I’ve tried auto complete, and it will often almost get it right, but not quite. So instead of getting to write the code, I go to write the code, stop to check if the auto complete is doing what I was thinking, see that it isn’t, and then have to type it out anyway. I’ve tried using a chatbot for coding questions, but often times it will claim there isn’t a way to use a given SDK to do what I want, but I will check the docs and there will be a way, so it just lies. I’ve tried having it generate code from a design, but having to write a strict design, and review all the code it produces, only to have to make pretty drastic changes tends not to save me any time, and means I have to spend the most time doing the parts of coding I hate the most: writing designs and code review.

    So far, the AI coding tools make my life worse and don’t save time, at least when writing code. It’s pretty good at writing git messages and making my operational summaries sound nice, so I primarily use it for that

  • dohpaz42@lemmy.world
    link
    fedilink
    English
    arrow-up
    12
    arrow-down
    2
    ·
    edit-2
    6 days ago

    Ok, here’s a hot take. I’m avidly against generative AI. But I’m not against using LLMs to help assist with coding. Jetbrains offers a free LLM (they have a separate paid AI service too) to help with code completion, and it’s fairly good for repetitive stuff. It’s also hosted locally on your machine and not a cloud service.

    That said, if people want to write boilerplate/template code with AI, fine. But it must be edited, tweaked, and reviewed by a human. This is no different than blindly copying and pasting from places like StackOverflow. You need to know and understand what’s being presented to you so you can know and understand what it’s going to do in your application.

    Edit: forgot to lament that writing code is not the time-consuming part. It’s the editing, tweaking, and reviewing process that is. But what do I know? I’ve only been doing this for ~20 years professionally. 🤷‍♂️

    • Vinny_93@lemmy.world
      link
      fedilink
      arrow-up
      4
      ·
      6 days ago

      I don’t disagree with having it help out a little or be better at autocomplete but having worked with Databricks’s Genie or whatever it’s called I am just now annoyedly hitting backspace when it suggests something stupid again but enter also autocompletes rather than just tab.

      • dohpaz42@lemmy.world
        link
        fedilink
        English
        arrow-up
        2
        ·
        6 days ago

        I get it. As much as I love autocomplete, it’s not perfect. I like the LLM, but I’ve never used (nor do I plan to) an AI service.

        Though I’ve been toying with the idea of creating my own local LLM. Maybe that would be better than Copilot or those other public services. 🤷‍♂️

  • ISO@lemmy.zip
    link
    fedilink
    arrow-up
    2
    ·
    6 days ago

    this is the future

    This is not necessarily wrong. This is a potential future of sorts.

    But, as always, the usual implicit “future = good” is assumed by all involved.

    One shall neither subscribe to progressive optimism when it comes to the future, nor expect the future to have a Whiggish outlook about us.