.@danshipper and @bcherny on why they use Claude agents to check other agents. Boris, the creator of Claude Code, runs multiple agents at once for big migrations—with one checking compliance, one digging through git history, and another hunting for bugs. Because the agents often flag false positives, Boris spawns five more sub-agents to check the findings for accuracy. Dan uses a similar process for filing expenses. One sub-agent argues his case for filing a certain receipt; another plays…