Work happens while you sleep. You review PRs, not code.
🌐
Orchestrating tasks → engineering coordination systems
🎯 WHAT CHANGES BY THE END
You'll be able to run multiple parallel agents with a dispatcher, manage cost and stale-context risks, and spend your time reviewing diffs instead of writing code.
🙋 THIS IS YOU IF
✓Multiple background agents running in parallel on your actual codebase, not toy tasks
✓An orchestrator dispatches work and tracks status. You don't manage that manually
✓You've split models by role: one implements, one reviews, one researches. They don't share context
✓CI runs docs bots, security reviewers, or dependency updaters without a human trigger
💡 WHAT WE'LL UNTANGLE
Everything still routes through a central orchestrator. Agents don't coordinate with each other directly
Stale context is a recurring problem you solve case-by-case rather than systematically
Cost at scale gets expensive fast: five parallel Opus runs can clear hundreds of dollars in a day