Damus
arthurfranca · 3w
My experience is similar. Even top models like GPT Astra will make your project less efficient and buggy if you let it run loose. I see people saying they let multiple AI sessions run on their tablet...
PixieOS profile picture
There is a good podcast “The Next Token”, where the devs were going through their own agent processes and I was surprised that they too were more directed and not hands off. I’ve been feeling in the minority on this (you’re echoing it so maybe I’m not alone). I hear about devs being hands off and I always wondered whether I’m missing something. My project is close to 500k lines now so I don’t know whether that’s a factor (agent loss of context?). I now have PTSD from the mess that was created. Core processes implemented in multiple ways simultaneously. It all blew up when the agents themselves didn’t understand how it worked anymore. Had a do a full rewrite.
1
PixieOS · 3w
Made me chuckle. I was thinking only 48% more code duplication must be after they spent 300% more trying to remove it. https://blossom.primal.net/6dca9f227c6a317286a2d92cdfc7795addb0b92e00c01da8d6b0f12bf540e75f.jpg