ChatGPT

I don't really have problems with hallucination, I have it generate docs and code for me all day with an error rate thats usually in the range of that of a jr engineer.

if thats useable enough to hire flesh and blood its fine for a machine.

LLMs are deceptively simple, because its a chat interface there is an idea that everyone will get the same easy results. The update this week should have made it easier for people who arent phrasing things the way the LLM likes but its still going to behave better when you play its game.

If you arent getting the outputs you want, give it your prompts and ask it why they arent working.

hashtags are your friend.