The core primitive
A Yawn is a container for uncertainty collapse.
The root question
“What should we do next to reduce uncertainty and increase agency without violating coherence?”
Anatomy
Question
A latent or explicit question driving the yawn
Plans
Candidate paths, experiments, and moves to reduce uncertainty
Proof
What evidence counts as progress
Feedback
Channels for sensing change and learning
Promotion / Split rules
When to promote, split, retire, or kill
A Yawn is not a task list. It is a decision-control unit.
The cycle
Every Yawn runs this loop continuously. Each cycle collapses uncertainty by one move.
Spec
Define what you want and what "done" looks like
Plan
Design candidate paths with hypotheses and success criteria
Prove
Gather evidence and verify against spec criteria
Learn
Update beliefs, surface tensions, and spawn next cycle
In practice
Type whatever is on your mind. No structure needed. Your messy input is the raw material.
Yawn organizes your uncertainty into moves — concrete next steps with proof requirements.
The autonomous loop executes, proves, learns, and surfaces what to do next. Repeat until done.
The holarchy
Every Yawn is both a whole and a part. They spawn children, split when complexity grows, promote when proof is met, and retire when uncertainty is resolved.
Spawn
When uncertainty exceeds the control capacity of the parent, a child yawn is born.
Split
When working memory is exceeded — too many moves, too many agents — the yawn divides.
Retire
When uncertainty is resolved or the yawn produces no actionable proof, it dies. Death is healthy.
The core invariant
Yawn does not optimize action. Yawn optimizes the collapse of uncertainty into agency-preserving action — at every scale.