The method
Six stages, and a way out of each one.
A system for tracking what matters, writing down what you expect before you act, installing reversibly, inspecting against the expectation, versioning when reality finds the edges, and expiring what stops working.
The six stages
Track
Point a system at a domain. Scheduled and automated, not manual browsing.
Tracker output, daily log
Hypothesize
Write down what you expect before acting. What changed, why it matters, what success looks like.
Decision record
Run
Install the smallest reversible version. Never the full commitment first.
Vertical slice, pilot protocol
Inspect
Measure against the hypothesis. Not against how you feel about it.
Review note with numbers
Version
Reality found an edge. Encode the boundary, bump the rule, loop back.
Boundary case, versioned rule
Expire
Every adoption carries a kill date. Nothing is permanent by default.
Expiry date on every decision
The three exits
Inspect has three ways out. One loops.
Most frameworks stop at Inspect and offer two outcomes. It worked, or it did not. The third is the common one, and it is the only one that compounds.
Why version is the point
A system with fifty documented edges cannot be copied by reading its description.
Binary pass or fail is not what happens in practice. What happens is: it mostly works, then something at the edge breaks, and the rule gets tightened. That is the third exit, it is the most common one, and it is the one that accumulates.
A published case states the rule that was added and what that rule still fails to catch. A case without both is marketing, not evidence.