Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

100% This is EXACTLY my experience and I didn't even spend 1% of what Steve has. Fable loves to create rules around the evaluations and decisions that I make. This is good, but this is also scary because 1/ I am fallible 2/ I don't have time to digest every detail and make a careful decision. And if I do spend that time, the system will return in 10 mins and give me another massive set of decisions to make .. thus creating an unending loop, resulting in decision fatigue... which leads to #2 again.

So AI is optimizing in some ways for an AI as the judge, not human as a judge. It needs this input, this steer, but humans aren't built to support this.



The trick is to use a cheap Chinese model as the human-to-machine interface. No bullshit. Just honest mistakes.



Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: