The claim

Three statements, not one. Let C be the collection, x the item, τ the deadline, and S_τ the part of C a search has examined by τ.

  • Present: x ∈ C.
  • Found by τ: x ∈ S_τ, and the searcher recognized it.
  • Available by τ: the chance that the system, given the cue it has, deploys x for the task before τ (TERM-055’s α).

“If I can’t find it, it isn’t there” is said when Found by τ fails. The owner’s version in CASE-094 says what is meant: “if I can’t find it quickly enough, it may as well not be there”. That is a statement about availability, put as a statement about presence for emphasis. TERM-055 already records the reading: “literally false as a claim about inventory, but potentially correct as an operational equivalence.”

Read as unavailability, the rule needs no default. A search that has not found x by τ has not deployed it by τ. That is an outcome, not an inference, and it is what the task needs to know: stop, substitute, or change the request. It is also one observation of α for this request, and repeated failures are evidence that α is low for this arrangement and this cue.

Read as absence, it needs a certificate. Not found by τ licenses x ∉ C only when something shows that x ∉ C \ S_τ. Three things can:

  1. Exhaustion. S_τ = C. This is Clark’s finite failure: negation as failure is sound only after “the exhaustive but unsuccessful search for a proof” (THEORY-341).
  2. An invariant of the arrangement. A sorted table, an index or a labelled compartment bounds what the remainder can hold, so part of a search can prove absence (THEORY-340, THEORY-363; CLAIM-311).
  3. A closed-world default. Reiter’s assumption that what is not entailed is false (THEORY-341). It is a default, not a proof, and it is inconsistent when the database holds a disjunction such as “the drug is in one of these compartments”. It is not even cheap in general: by Halevy’s account of later work, deciding what follows from stored partial views is co-NP-hard under the closed-world assumption even for conjunctive views, where it is often polynomial under the open-world one (THEORY-374, from a survey’s report).

CLAIM-306 assumed only the first and third, and so concluded that absence needs exhaustion and that the rule, read literally, is stronger than either the closed-world assumption or negation as failure. Under the literal reading with no certificate, that conclusion stands.

When the gap between the readings matters. By Bayes, P(present | not found by τ) = π(1 − h) / (π(1 − h) + 1 − π), where π is the prior that x is present and h the chance a search finds a present item by τ. That is the probability that “it isn’t there” is false as an inventory claim while true as an availability claim. The hyperbole is harmless when the act it prompts is one unavailability licenses. In CASE-085 the medic stopped searching and asked for something else. It is harmful when the statement travels to someone who acts on presence, such as the next shift, which then does not look, or a stock record, which then reorders. The same sentence is then a closed-world claim made on someone else’s behalf.

What it does not say

  • That the medic was wrong to move on. Under a deadline, stopping is part of using a resource (CLAIM-296; CASE-085).
  • That availability is all that matters. Whether a further search could pay depends on presence, through π.
  • That every speaker of the rule means availability. The claim is about the owner’s usage and the medic’s case, where the context makes the operational reading the natural one.
  • That Reiter or Clark considered deadlines. The extension to search under a deadline is the record’s.

What it replaces

CLAIM-306 is superseded. Its account of the two assumptions, of the disjunction in CASE-085, and of the Bayes term is kept above as the analysis of the literal reading. Its claim that absence is shown “only by exhausting the collection” is withdrawn (CLAIM-311).

The owner, at U10 of the exchange of 2026-10-26, on A9’s “correctability frontier”: “reminds me of the ‘if I can’t find it, it isn’t there’ stuff. let’s poke out the resource-boundedness here.” A10 read this claim and its predecessors correctly (CLAIM-306 Superseded, CLAIM-311’s sorted index) and carried the distinction to remedies:

  • “We have not found a lawful means of returning this person” reports a failed search, and establishes operational unavailability.
  • “No lawful means of returning this person exists” asserts absence, and needs “a justified exclusion of alternatives”, the certificate this claim requires.
  • “We cannot find and execute a lawful means before the deadline” is a resource- and time-relative capability claim.

What the item case lacks is a searcher who is a party. A10: “one actor may influence the search environment faced by the actor responsible for reviewing or correcting its behavior”, and an institution’s budget “is itself partly an outcome of political and organizational decisions”. QUESTION-279 asks what the certificate would be for a remedy, and TERM-085 is the availability. A10 declined to file its own claim candidate, “since this is substantially an application of CLAIM-296”, and so does this filing.