Assessment 224: owner discovery without timely source access

12 October 2026. Supporting arguments and a sustained unsuccessful extension attempt. At this assessment, construction 76 was the latest completed native result; the attempt added no native construction, root admission or external priority claim. The later construction 77 uses a different restricted interface.

Outcome and routes tested

The full argument discharges one preparation step for bounded overlapping ownership. Ordinary target input can build an exact inverse directory of retained owner names. The local support-birth flag suppresses duplicates, and once-only store groups make owner-name comparison unnecessary.

That directory does not supply final frozen store roots or source-specific lane access. An owner learned at the last target bit arrives too late for a second lookup and query replay. Short owner codes alone do not repair this: a cut argument for equal-width bit-reversed keys defeats a precisely restricted, factored activation-trie reader, including speculative traversals and free retention of nodes observed during the charged query phase.

The restriction matters. The phase starts with only the activation root and activation-independent owner data; there are no retained construction-time deep handles, source-specific shortcuts, activation aggregates or side channels. This is a representation/readout obstruction, not a native-language lower bound. Singleton stores escape it through a source-specific target trie and one pending target, with insertion paid by the next equal-width store name. Enumerating a general dense selected row cannot use that singleton budget.

The attempt tested owner inversion, final-root indirection, short width-class codes, speculative access and source-answer materialization. It did not complete the arbitrary-overlap recognizer. It narrows the missing operation to early source filtering with an explicit charge for consumed-prefix replay.

Independent adversarial review

The reviewer checked the complete owner producer, cut proof, Horn realization, singleton control and unsuccessful-extension classification. Review required explicitly excluding initially retained deep activation handles and charging replay of query-prefix bits consumed before a continuation becomes accessible. Both clarifications are in the manuscript. Approval concerns these supporting arguments relative to the stated contracts; it does not establish a new native theorem.

Independent finite evidence

The standalone audit completed the following checks. It is an independent finite semantic and graph-cone experiment, not native execution or timing measurement.

Controls and next decisive test

Replay pays each traversal. Affine joins retain their coordinate hypotheses; causal sharing does not backpatch immutable roots; coherent unary algebra does not supply this branching index. The native SAT controls retain ordinary late pins and final-symbol deadlines. Imported entry 129 retains its growing alphabet and conditional state/port scope. No imported theorem is newly claimed or re-reviewed here.

Build a source-filtered query route for arbitrary names and ownership at most two. Release a fixed number of candidate continuations early enough: they must already have processed the consumed query prefix, or the remaining actual suffix must pay source routing, prefix replay and ongoing readout. Compare online prefix routing and persistent multi-index access in the next primary-proof review. No supplied incidence index. External-heap import, variable widths within a store and right-group reopening remain separate.

Integration and accounting

The proof, checker, review, checkpoint and reader synthesis have one maintained GitLab source. Styled HTML/KaTeX and syntax-highlighted code with line anchors and exact raw downloads are generated projections. The homepage research date is unchanged because this is an unsuccessful extension, not a completed new advance. Build, rendered-link checks and publication receipts belong in integration state. Browser viewport inspection is unavailable in the supported workflow and is not claimed.

Scheduled completion advances 128 to 129 exactly once. Mathematical count 139, window five, reassessment 134 and native construction 76 remain unchanged. Last primary review 220 and editorial 221 remain; next scheduled 130 is a dedicated primary-proof review of at least two papers, and next editorial is 132. Installed legacy reverse-writer retirement remains separate and unverified.