Raw LLM Responses

I'm finding there's a growing divide between architecture-minded devs who think about the system holistically and are getting great results from LLMs, and devs who are purely ticket-and-execution minded - they grab a ticket, feed it into Claude, Claude suggests whatever, they implement the whatever and move on. I'm increasingly spending my time babysitting the latter group to stop them fucking our shit up. LLMs are a genuinely dangerous tool in the hands of someone who doesn't understand and care about the system they're changing. I think this is also why so many people complain about getting shit results tbh - it's so easy to misuse. I still think AI will lead to mass unemployment in tech lol.