Skip navigation
Never correct Claude Code more than twice on the same problem.
0:00

Never correct Claude Code more than twice on the same problem.

Never correct Claude Code more than twice on the same problem. This is in Anthropic's own documentation and almost nobody follows it. When Claude does something wrong and you correct it that correction goes into the context. Do it again and the whole session is now anchored to every wrong path it already tried. You are not making progress. You are making it worse. The right move is /clear. Reset the context entirely. Then rewrite your original prompt using what you learned from those two failures. More specific. Better scoped. With the edge cases you now know about. A clean session with a better prompt almost always outperforms a long session with accumulated corrections. That is a direct quote from Anthropic's docs. Not my opinion. Two corrections max. Then clear and restart smarter. Join AI Staff Room at aistaffroom.com for the exact prompt structure I use when I restart.

Comments

Comments from the original post on Instagram

No comments yet