Multi-agent orchestration.

Multi-agent orchestration is running a bunch of AI agent together under one view, with shared context, instead of babysitting each one in its own window. You spin up Subagents in parallel, and the work gets routed to whichever agent is right for it.

If one agent hits something outside its scope, it passes the job to the right agent automatically. You don't sit there copying output from one chat into another. The system handles the handoff.

When you're building something real, this is what lets a job span several specialists without you being the glue between them. It's also Tokens-hungry, since every agent is reading and thinking at once, so I save it for work that needs more than one brain.