Christopher Weir coordinates about 90 Claude Code worker sessions from one "ringleader" thread for more than 60 merged PRs in five weeks
Updated on Set published to 2026-11-11 from the page's time datetime="2026-11-11".
Christopher Weir describes coordinating coding agents in a large work codebase. Over about five weeks, around 90 Claude Code worker sessions produced more than 60 merged pull requests, while he sent about 200 messages, typically around 15 words each, nearly all to one parent thread he calls the ringleader. The ringleader worked through an agreed plan in phases, turned decisions into briefs of a few thousand characters, started workers in separate Git worktrees, waited for prerequisites to merge, and then checked each change itself by reading diffs, rerunning checks against the combined code and exercising user flows. Workers implemented changes, reviewed code, tested behaviour or investigated failures. He turned the approach into a Claude Code skill and first used it on a navigation refactor.
The numbers (as reported)
| Worker sessions | About 90 Claude Code sessions |
|---|---|
| Merged pull requests | More than 60 |
| Duration | About five weeks |
| Human messages | About 200, typically around 15 words |
| Share of model output in workers | About nine-tenths (transcript estimate) |
Facts
- Company / subject
- Christopher Weir
- Tools
- Claude Code
- Codebase
- brownfield
- Work type
- Refactor / maintenance
- Stack
- —
- Published
- 2026-11-11
- Last checked
- 2026-11-16
- Updated
- 2026-11-16
What the source claims Practitioner (own blog/paper)
Claimed by the builder, writing about their own work. Not checked independently.
- More than 60 merged PRs from about 90 worker sessions over about five weeks, steered with about 200 short messages 60+ PRs, ~200 messages
- Hands-on checks by the ringleader caught a double-submit bug and wrong keyboard focus after hundreds of tests had passed
Caveats
Self-reported by the developer about a work codebase; the company, product, stack and the kind of change behind the 60+ PRs are not named, so the work type is uncertain (the only named project is a later navigation refactor). Figures are rounded on purpose. He has not measured total cost or elapsed time against doing the project in one thread. In one case parallel workers filled the disk and stopped the local container runtime.
More stories by tool: Claude Code
Related topics: Refactoring and maintenanceBrownfield codebases