Skip to content
Chapter 01 · 5 min

Why a team beats a solo session

You write the code, then review it, then test it, then open the PR, then update the docs: every task behind the last one, all day. Claude Code can run those at the same time: one agent per task while you move to the next feature. This course shows how to set that up without burning tokens or blowing up your repo.

“A lone craftsman who cuts, sands, finishes, and packs every chair himself ships one a day. A workshop where four people each own a station ships four, and nobody waits.”

The sequential tax

A four-part feature (backend, frontend, tests, review) done one piece at a time takes most of a day. Worse, your single context window bloats as you switch between unrelated tasks, so the model carries the frontend's mess into the backend's reasoning.

What changes with a team

Run the four in parallel and a lead agent coordinates them. Each teammate works in its own clean context. The same feature lands in a couple of hours, and you spend your time reviewing and merging rather than typing every step yourself.

SOLO                          TEAM
 backend                       backend  ┐
   then frontend                frontend ├─ in parallel
     then tests                 tests    │
       then review              review   ┘
 ~1 day, one context          ~2 hrs, clean contexts
Same work, same subscription. The difference is whether the tasks queue or run side by side.

What this course covers

  • Chapter 2, The three levels: subagents, Agent View, and Agent Teams, and which problem each solves.
  • Chapter 3, Your first team: turn the feature on, then write a prompt that lets a lead agent decompose the work.
  • Chapter 4, Cost and models: route teammates onto a cheaper model and understand what actually caps spend.
  • Chapter 5, Manage and choose: drive everything from Agent View, and a rule for when *not* to staff a team.
  • Chapter 6, Guardrails and setup: permission rules plus a copy-paste configuration.

In one line each

  • Done sequentially, a multi-part feature eats a day and bloats one context window.
  • A lead agent plus parallel teammates does the same work faster, each in its own clean context.
  • The change costs one environment variable and a prompt that says 'spawn separate agents', not a new tool or plan.

Where to go next

Why a team beats a solo session · AI courses · SDEN