Hacker News new | past | comments | ask | show | jobs | submit
slashdave | on: Claude Opus 4.8
RL is more than facts. Synthetic feedback is an obvious approach. Does the model suggest code that compiles and performs well?