Assessment 232: producer-checked release envelopes

13 October 2026. Written native construction 80 and an exact restricted scheduling criterion. No root admission or external priority claim.

Result and contribution

The complete argument replaces construction 79’s fixed output lag with a producer-checkable common-prefix envelope. Each target prefix must determine enough ordered output for the remaining actual input to pay all unreleased traversal. A token chain created during ordinary records generates bounded chunks, avoiding longest-common-prefix discovery, supplied annotations and endpoint scans.

The envelope is necessary and sufficient for the specified beta-only non-speculative reader at rate C. The composed native reader also pays the earlier alpha traversal and setup, using rate max(C,4). Necessity is not asserted for this larger service budget or for arbitrary representations. An arbitrary learned first-half replacement, actual support masks and one captured current activation dictionary remain in the native construction.

Every fixed-delay contract is included by choosing C=d+1. Reversing the first half of the suffix and leaving the second half unchanged satisfies C=2, while failing every fixed-delay contract as widths grow. A separate midpoint-prefix witness gives unbounded prefix ambiguity. This is a strict extension of those sufficient contracts, not a separation from every prior native construction. The contribution is an extension of the Lab’s release-tail criterion and charged capsule/queue primitives. The previous primary comparison supplies no automatic native compiler or claim of external priority.

Directly dependent bounded-width extension

The same envelope now admits up to W speculative output prefixes, for a fixed W. Each program node stores append-only finite slots (parent slot,chunk), built and checked during ordinary records. Stable local indices avoid comparing whole output prefixes or repairing old child pointers. Query workers fork through immutable queue/cursor roots; all current branches receive their own bounded service. Terminal functionality leaves one owner. This is sufficient, with constants depending on C,W; the exact necessity theorem remains the one-path statement.

A last-bit-controlled first output bit meets W=2,C=1 but fails every one-path fixed-C envelope across growing widths. Its suffix-only bounded branching is already an earlier control; the extension here combines charged program construction, growing-delay envelopes and arbitrary learned alpha replacement.

Independent adversarial review

The reviewer independently derived the envelope equivalence and token producer, then read the complete manuscript against constructions 78 and 79. Review covered sparse support, functionality, prefix payload updates before descent, suspended capsule paths, token allocation during paid alpha transfer, captured activation versions, r=1, C>r, final chunks and setup interference.

Review distinguished beta-only rate-C necessity from full rate-max(C,4) sufficiency. It also corrected the ambiguity witness: a prefix of length n-1 leaves only one first-block bit unknown; the exponential ambiguity example uses j=floor(n/2), with j unread bits yielding 2^j possible output prefixes. The fixed-delay failure still uses j=n-1. The parent incorporated this correction and made explicit that unsupported short/long query targets can abandon their workers. The complete corrected manuscript was approved. The reviewer then audited the bounded-width extension: append-only slot stability, terminal uniqueness, persistent queue forks and a full K-service batch for every child. Total C*W service alone does not establish a converse for arbitrary speculative readers.

Finite computational evidence

The readable checker adapts construction 79’s relation-wrapper model. Its independent oracles compare pairwise prefix conditions with incremental chunk insertion, direct relation witnesses with represented answers, maximal safe common-prefix emission with the prescribed envelope, and tick-by-tick service with the ordered-job recurrence. Explicit token lists check the producer schedule separately from its closed formula.

All pass. The additional branching audit compares semantic width/functionality with append-only transition slots, then follows separately serviced immutable worker copies to the unique endpoint. The maximal-prefix oracle deliberately recomputes semantic information without charging its cost; the native construction uses the token producer instead. Python containers and snapshot copies are semantic models, not native storage operations. Empty tables test components; the complete language requires a nonempty right block. The checker does not execute the native compiler, prove an infinite bound by enumeration, or establish external novelty.

Failure boundary and next decisive test

Whole-suffix reversal on the full domain fails every fixed-rate envelope at sufficiently large width. This rejects only the non-speculative ordered-output pipeline. Prepared aggregate predicates, speculative path traversal and other representations remain outside that obstruction.

Replace fixed simultaneous width by an input-derived speculative frontier with paid local service. Charge discovery, traversal and elimination at their release times, including stored but temporarily unserved candidates; a total-work bound alone does not suffice. Seek a useful variable-width subclass without frontier scans or supplied decompositions. Arbitrary whole-suffix permutations, arbitrary-name routing, cyclic state sharing and heterogeneous service remain open. Preserve this target through the next editorial pass.

Integration and accounting

The maintained proof, checker and this review remain in GitLab; styled HTML/KaTeX readings and an exact-download highlighted source reader are generated projections. This extends the program overview and its existing homepage entry, without creating a paper or changing root claims. Custody, claim scope, proof/evidence, failed route, surviving frontier, reader synthesis and Site representation are explicit. Deployment and exact source-parity checks belong to the canonical integration receipt.

Scheduled completion 137, mathematical count 143, window three and native construction 80 advance once after the reviewed packet. The last six-result reassessment remains 140. Next is editorial completion 138; the next dedicated primary-proof review is 140. Prior SAT/scaffold results, imported entry-129 state/port bounds and installed-writer qualifications are preserved.