• Serinus@lemmy.world
      link
      fedilink
      arrow-up
      12
      arrow-down
      1
      ·
      7 hours ago

      If you want the real version of this…

      The AI can produce its own tests after writing its own code, allowing for safe and incredibly efficient development cycles that can be put straight into production with little to no human intervention.

      • aksdb@lemmy.world
        link
        fedilink
        arrow-up
        3
        ·
        3 hours ago

        Just have a second agent review the output of the first and have a third agent try to break the implementation. Easy. /s

      • uSeRnAmE396@lemmy.world
        link
        fedilink
        arrow-up
        5
        arrow-down
        1
        ·
        edit-2
        5 hours ago

        You can write as many tests as you want. Does it mean they’re any good? or are they testing what they’re supposed to be testing ? No !

        I’ve seen the tests the actual developers write and most of the times they don’t stand a chance to someone whom is specialised in this field.

        But you do you and trust your “assert(true) tests” that your magic llm is regurgitating.
        Your vibe coded software is an abomination.

        Later edit :

        Also, when you’re done smelling your own farts, you might put 2 and 2 together and realise that all the failures that happened in the last weeks (banks, airports, etc …) are caused by your “can be put straight into production with little to no human intervention.”