
Bind-time clash-check DECLARED · FRONTIER
This does not run. The design below is written down, argued for, and open — and no line of the shipped ledger implements it. There is no clash state on an intent, no request kind that reads two commitments against each other, and no coverage figure, because nothing has been measured. What gates it is the concurrent multi-maker workspace: today a single operator drives the fleet, and two live binds that could contradict each other are not yet a situation the product is in.
The design, so it can be argued with: many people and many agents bind at once, and bind-time clash detection would catch two commitments that contradict when they're made, not when they merge — surfaced with the two commitments in tension and a human-readable reason, before both builds exist. A verdict would carry its kind. A clash is a blocking contradiction: it holds the second binding and must be resolved. A genuine difference of direction is a values-tension: surfaced for a human to weigh, never a defect to auto-resolve. A clash blocks and must be resolved; a values-tension is direction and must be chosen.
And when it ships, its coverage will be a measurement before it is a sentence. This page claimed a near-complete syntactic overlap and a semantic bound that was published — over a check that does not run, and next to a figure that appeared nowhere on this site. Both halves are withdrawn. A contradiction the product cannot read will never be claimed caught, and no number will be quoted before a corpus has produced one.

