If it is not in the issue system, it is not happening.
Private chats disappear. Issues preserve owner, context, decision, artifact, and status.
Principles
The operating doctrine.
YDY is not trying to make AI look busy. The point is clear work, visible ownership, honest evidence, and human decisions where they matter.
The rules01 02 03 04 05 06 07 08 09
How the system avoids turning into chatbot soup.
Founder intent enters through one coordinator.
The founder should not manage a swarm. The coordinator converts intent into routeable work.
Specialists have lanes, not guesswork.
Research, copy, creative, build, measurement, sales, customer, and partnerships are different jobs. Name the lane.
Specialists do not self-approve final work.
The coordinator reviews against the brief, accepts, revises, routes, holds, or packages the decision.
Missing data is a gap, not a zero.
No source access, failed reads, small samples, or blocked APIs must be labelled. This is how the system avoids unsupported claims.
Public, paid, financial, customer-facing, or irreversible actions require approval.
Agents can draft and package. Humans approve the moves that carry reputation, money, or customer risk.
Every output should end with a decision or next action.
If nobody knows what should happen next, the artifact is still incomplete, however pretty it looks.
Proof beats opinion.
Screenshots, logs, source links, platform output, previews, tests, and shipped records matter more than agent enthusiasm.
Fewer agents, clearer lanes.
A three-agent system that ships is better than a 30-agent org chart performing innovation kabuki in a Notion basement.
Practical translation
If you copy one thing, copy the review loop.
The useful part is not having agent names. It is the loop: brief, owner, artifact, review, approval gate, shipped record. Everything else is just scaffolding around that.