◀ THE FOLD0ROOT.AI // WORLD II · BOSS · THE FIREWALL◆ .dlw.fold
THE FOLD / BOSS / THE FIREWALL / THE NAME THAT WALKS

THE NAME THAT WALKS

the control that fires most often
1 WHAT IT IS · WHAT IT DOES · FACT OR FICTION
A name was coined in conversation and deliberately kept out of the shipped artifact. A grep at step nine found it sitting in the page anyway — for the second build running. Of every control in the pack, the one that fires most often is not an assertion about a number. It is a search for a word.

LIT verified live. The name grep has fired on 2 of the 2 builds where it existed — a rate of 100%. The other three controls are numeric, and none of them can see a name at all. The leak needed a control of a different kind, not a stricter one.
2 HOW IT WAS WEAVED · AI + HUMAN
David (human) logged the catch and drew the general conclusion from it: “the name coined this session, and one structure’s proper name, were both sitting in the artifact. Redacted at the model and re-injected. Second build running where this grep has fired; it is now the control that catches most often, which says something about how easily a name walks into a deliverable.”

AVAN (AI) would add the reason the measurement is possible at all. A leak can only be counted if somebody decided in advance that the thing should not be there — a name nobody chose to withhold would never register as escaped. So the 100% rate is not a fact about names in general; it is a fact about the one class of content this pack has an explicit policy on, and the policy is what turned an ordinary editorial slip into a countable event.
3 ONE DIMENSION
Four controls, and how often each fires.
4 TWO DIMENSIONS · INTERACTIVE
Which control could have caught which fault.
5 THREE DIMENSIONS + AVAN’S INVERSE
The green forward object: a word slipping through a net of numbers.
AVAN’s addition (the inverse-companion): the forward reading is “a name walks into a deliverable easily.” The inverse is that a control firing every single time is a control describing a broken process, not a working guard. Two builds, two leaks, caught twice at step nine — the grep is doing its job and the thing upstream of it has not changed at all. Read backwards, a 100% catch rate should be read as an outstanding defect rather than as a success, and the fix is not a better grep but a step that stops the name reaching the artifact.
LIT the name grep has fired on 2 of the 2 builds where it existed, a rate of 100%, while the other three controls are numeric and none of them can see a name at all - the leak needed a control of a different KIND, not a stricter one

FIG David logged the catch and drew the general conclusion: 'the name coined this session, and one structure's proper name, were both sitting in the artifact. Redacted at the model and re-injected. Second build running where this grep has fired; it is now the control that catches most often, which says something about how easily a name walks into a deliverable.' AVAN adds the reason the measurement is possible at all: a leak can only be counted if somebody decided in advance that the thing should not be there. So the 100% rate is a fact about the one class of content this pack has an explicit policy on, and the policy is what turned an editorial slip into a countable event.
◆ sealed .dlw.fold → folded to ROOT_0 · a sphere of THE FIREWALL · David Lee Wise (ROOT0), with AVAN