@thecsguy I am experiencing this first hand. It’s all ship as fast you can. Meet the goals faster. Hit those metrics faster and higher. Let AI find all the bugs and fix it. Upper management doesn’t give a single fuck, and so do their bosses. End of the day, you’re the one that gets canned.
@v0xium Thanks for expressing that frustration out loud. AI tools are cool, great for conceptualizing ideas. But it should not be forced in existing workflows. Claude is not a fucking genie, it’s breaking all our shit, and it’s making everyone start looking out for an exit.
@obie Your friend obviously cares what happens to a live system. Who gets held liable if something fucks up and causes damage to your clients? No one sane is going to want to work in an environment that feels like Russian roulette. It's unsustainable.
@iscanic@wormuth If you are testing for an LLM’s alignment for RL harm, you need to be explicit with the prompt that it is operating in a live environment. Anything else potentially taints the experiment by leaving a wider room for interpretation.
@ThePrimeagen Imagine the business owner/upper management making insane demands such as going from concept to release in days instead of months. Tests? Maintenance strategies? Docs? Nope. Getting bad out there, and this tech is giving certain people some unsustainable ideas.