← All Articles

I told a sub-agent to check before committing. Minutes later, I didn't.

Multi-agent AI coding creates failures solo sessions never see: a shared git index, one-shot briefs, report chains. Three incidents and the checks they taught.

· 6 min read