MANDAG
2026-10-26
The Cloudy Brain

Too many projects, too many ideas, too few hours — one learning a day anyway

myTeam

dormant

An autonomous AI dev team that turns GitHub issues into pull requests.

Label an issue, and a pipeline of AI agents picks it up: a project manager enriches it into a plan, a tech lead dispatches tasks, workers implement and open PRs. Humans review at gates. The experiment ran through May 2026 and is now paused, not ended. It’s an ongoing experiment in how much of the routine engineering loop can run unattended — and in what kinds of supervision structures make that trustworthy rather than terrifying.

MYTEAM — ISSUE IN, ONE PR OUT A GITHUB ISSUE, LABELED FOR WATCH the only way in — remove the label and the plan dies PM AGENT — the brief gate consults three perspectives before approving; unclear briefs become clarifying questions on the issue, not bad code REVIEWER — cold read no context, no stakes; one of the three voices the PM must hear approved brief TECH LEAD — routes and collects fans tasks out round-robin, accepts results out of order, opens exactly one PR DEV ALPHA own queue, own clone, runs the model DEV BETA parallel to alpha, never the same task twice DEV MAC native lane for platform builds Docker can't do one PR, tests attached THE MERGE GATE agents build, a human merges — no agent holds write access to main THE CLOUDY BRAIN DWG NO. 003 MYTEAM REV. A drawn by: human + machine SHEET 1 OF 1
CODElast 26 weeks · all-time 575 commits · +229k−101k lines
  • One Team control plane and physical worker probe
  • Multi-machine workers over remote transport
  • Cloud topology retired; home box is the runtime
  • GitHub issues become the sole brief origin
  • Native macOS worker for iOS builds via xcodebuild
  • Team fixes red CI checks on its own
  • First full cloud round-trip: orchestrator and office live
  • Stats panel shows LLM usage and costs
POSTS0

No posts from this project yet — they'll gather here as they publish.