• Beej Jorgensen@lemmy.sdf.org
    link
    fedilink
    English
    arrow-up
    1
    ·
    7 days ago

    Building software at scale has always been an exercise in managing complexity. We carefully architect systems to be maintainable even at huge scales. And no one can do this quickly. I think writing code by hand gave us some think-time to ponder these architectural decisions.

    But with AI, the thinking time is gone, but the problem of managing complexity remains. Typically, however, nobody above the devs gives a shit about managing complexity–that just sounds like something that eats into the bottom line. It is, as certain people might say, “loser thinking”.

    A lot of software engineers like to work on projects where they can be proud of what they’ve made. A beautiful architecture doesn’t only shine in its own right, but it also makes the company run smoothly, makes the customers happy, and improves the bottom line.

    And when you don’t have those things for the product that you’re pouring your heart and soul into 996, is anyone truly surprised that devs are unhappy?

    “The night of the fight, you may feel a slight sting. That’s pride fucking with you. Fuck pride. Pride only hurts, it never helps.” --Marsellus Wallace, Pulp Fiction

    • ZeDoTelhado@lemmy.world
      link
      fedilink
      English
      arrow-up
      1
      ·
      7 days ago

      I do resonate a lot with this. Also, another detail: with the speed of ai churning mostly not correct by itself results most of the time gives 2 big things:

      • a much bigger effort of reviewing spaghetti code and hope it makes sense and nothing bad comes out of it (keep in mind generated code can be heavily verbose and without need to do so)
      • without the thinking hard part of the code, making a cohesive architecture is a lot harder.

      Problem is, when the solution is “press button and hope the slot machine gives the good spaghetti back to you” is the answer, you’re fucked.

  • stopforgettingit@lemmy.world
    link
    fedilink
    English
    arrow-up
    1
    ·
    7 days ago

    The actual writing of correct syntax was never the bottleneck in development projects, but management is not ready to hear that they are the problem.

  • Treczoks@lemmy.world
    link
    fedilink
    English
    arrow-up
    1
    ·
    7 days ago

    Like one developer reported: Management demanded AI use, the most senior developers took retirement, and the management hired kids with AI experience, but no knowledge of the actual product. The codebase got bloated by a factor of three, bugs by the factor of ten, and nobody actually knows anymore how the product actually works.

  • eicker@lemmy.worldOP
    link
    fedilink
    English
    arrow-up
    0
    ·
    7 days ago

    Turns out giving every developer five tireless AI interns doesn’t eliminate work. It turns them into unpaid middle managers of five interns who confidently submit broken code at 3 a.m. Management sees five times the output; developers see five times the pull requests. Congratulations, we’ve automated coding and scaled up the headache.

    • thejml@sh.itjust.works
      link
      fedilink
      English
      arrow-up
      1
      ·
      7 days ago

      As a manager of 4, I now get 20x the AI slop PRs I have to read through and try to convince the developers they’re doing it wrong. AI use has made my job way more exhausting and while productivity increased, it’s fairly obvious that understanding has not. Even my most senior team members turn their brains off and become meat proxies.

      I have to keep reminding them “I don’t care what Claude found, I care what you think about it and how you’re personally going to own a fix for it”.

      They run Claude it makes a PR and even if you pay close attention, no one knows what a PR someone else wrote as well as the person who wrote it.