I read and review any code that goes into production. You could probably make the case that, if you do that, it's no longer vibe coding

Show Post

every single one of these examples is humans cutting themselves out of processes where their review is critically important, because it's too fun to push the button

Show Post

Stop asking AI coding agent to build the whole thing at once. 💡 Complex tasks broken into focused steps produce dramatically cleaner code. ✅ Decompose first. Prompt second. Review third.

Show Post

My latest thing has been: write PRD --> AI expands into types and interfaces-only spec --> long-lived plan agent creates plan.md made of Markdown todos, babysits it --> red-green TDD code agent --> Rich Hickey agent edits code agent work --> review agent --> loop back to plan agent

Show Post

Multi-agent orchestration isn't a framework problem. It's a coordination failure problem. You can have perfect message passing and still get emergent chaos. The hard part isn't the protocol—it's defining who owns state when every agent has a different view of reality.

Show Post