Claude Code creator, Boris Cherny:
"Every six months, delete your claude.md, delete your skills, delete your hooks, see what the model does, and it might surprise you."
Anthropic deleted 80% of Claude Code's system prompt for Opus 5. It doesn't need the instructions past models
Three identity decisions that turn a generic AI assistant into an agent that thinks for itself.
Generic AI assistants give generic answers. They summarize when you want analysis. They ask permission when the answer is obvious. They sound different every session because nothing
Your AI agent starts every session from scratch unless you build these three persistence layers.
Your agent has long-term memory. It knows your contacts, your deals, your preferences. But ask it "what were we working on yesterday?" and it draws a blank. Long-term memory without
A developer wrote 1,500 lines of Claude Code skills. Claude used none of them. Then he built a hook system that made it impossible to ignore.
This comes from a dev who solo-rewrote 300K lines of code over 6 months using Claude Code. He wrote detailed skills for frontend
