What to close first — COGNI.BUILD

Fourteen items from the conformance assessment, with the dependencies between them. Pick what you want to close; the planner enforces the ordering, sequences it, and prices what you leave out. Effort figures are estimates from the assessment, not measurements — treat them as relative, not absolute.

capacity

10 engineer-days

the plan, in dependency order

what this leaves open

backlog

The four half-day items are one afternoon between them, and three of them get structurally more expensive the longer they wait — not because the work grows, but because the thing they constrain grows around them. Declaring an effect flag costs half a day against four write tools and a fortnight against forty.

The dependency chain is the real finding. Selection probes need a case set; a case set needs a test harness; the harness is half a day that has never been prioritised because on its own it does nothing. That single unglamorous item is what gates the entire evaluation layer.

What that showed

The ordering is the finding. Four of these are a single afternoon between them, and three of those four get structurally more expensive the longer they wait — not because the work grows, but because the thing they constrain does.