Hacker News new | past | comments | ask | show | jobs | submit
Like if you're using claude code on a feature tangential to your training pipeline it's allowed to nerf itself and damage your AI work.