Hacker News new | ask | show | jobs
by jumploops 96 days ago
Funnily enough, with the most recent models (having reduced sycophancy), putting in the wrong assumptions often still leads to the right output.