Hacker News new | ask | show | jobs
by simonw 80 days ago
Thanks for this, I've added that to my write-up of the project here: https://simonwillison.net/2026/Mar/20/turbo-pascal/#hallucin...
1 comments

> Because it's amusing to loop this kind of criticism through a model

Maybe it could become a general pattern, to have an agent whose task is just to deny the output validity. GANs are a very successful technique, perhaps it could work for language models too.