Assessment 212: immutable codes by residual scale

12 October 2026. Written native construction 70 removes construction 69’s creator-code funding guard. The complete checked language now requires only the greedy context producer, bounded right-row degree and a fixed alias/residual length-class span. Context count, nesting and late shorter contexts remain unbounded.

Strongest result and why it matters

Assign a separate immutable code to each pair of owning context and residual length class. Its first actual record supplies the code. The existing span check automatically bounds that code by a constant times the class’s residual lengths. No old code is shortened, and no old receipt or selected predicate is migrated.

Permanent ownership still selects one context. Inside it, one current and one ahead class reader suffice. At residual length $M=2^h$, start the next reader with its class code and a snapshot of the observed prefix. Rate $2A+4$ pays code decoding and prefix replay before length $2M$. Composing with the inherited source-index readers retains a fixed number of workers. Code-directory holes are explicit; absence in one class does not imply absence in later classes.

The producer is carried through ordinary input. Residual doubling boundaries pay the class-directory route. Context marks retain paid capsule snapshots. Existing code clocks use the same lookahead schedule; a new class uses a prospectively saved clock of length $2|v|+2+b$, incremented by each residual bit. New contexts use the rolling-prefix producer and a class directory prepared during initial fill. All delimiter actions close already reached fixed interfaces.

The long-creator/short-alias example strictly separates the new checked class from construction 69 for every fixed funding constant. This is a structural extension of serial reuse, not an independent-store composition theorem or a native SAT lower bound.

Failed approaches and phase correction

Replacing every context code by its final shortest alias would be semantically adequate, but a frozen old row receipt does not automatically expose that final code at bounded local cost. Explicit relabeling must route and update every affected receipt; lazy indirection must charge current-code lookup after the row leaf. Neither obligation is removed by the existence of a short representative.

A geometric sequence of shorter codes bounds total code length but does not itself bound repeated residual processing: a separate full query per eligible epoch can repeat a length-$n$ residual at every epoch. This is a failure of that reader schedule, not a lower bound on all epoch representations. Scale-indexed codes replace temporal epochs with disjoint query-length classes; only the current class contributes.

All right preparation precedes left selections. Consequently a claimed necessity to migrate already-selected states during right-side code creation would use the wrong phase order. The actual unresolved burden of final-code replacement is preparing or resolving old row receipts before their later use. The new construction avoids both burdens by never replacing codes.

Review and computational evidence

The parent and an independent adversarial reviewer converged on class-specific immutable codes. The reviewer independently checked the size bound, lookahead service rate, interior directory holes and resettable prospective length clock, then approved the complete native producer and capsule-closure argument relative to the stated inherited interfaces, with no blocking gap.

The independent audit, with no primary imports, passes 4,455 accepted and 2,704 rejected histories, 25,416 retained selected versions and 1,790,151 query comparisons. It exercises 281,337 context resets and 1,911,447 class promotions, up to 13 contexts and multiple populated classes per context. Another 2,016 worst-size-code deadline checks cover 32 hole patterns. Cases include decreasing code scales, exact-context queries, missing names, duplicates and the strict-extension family.

These finite semantic and virtual-symbol checks do not simulate native dictionaries. The written native proof separately inherits section 15’s capsule, queue, pending-window and compiler contracts. The additional argument is the scale refinement and fixed-width delayed decoder. It is classified as an extension of those existing constructions, with no external priority claim or root admission.

Next decisive test and continuity

Replace the fixed row receipt tuple for sibling-closed clipped-cone rows by a checked mixed-region representation after class-specific coding. Charge its ordinary-input completeness producer, contextual holes, code routing and pending union. First determine whether the transformed row family preserves the required mixed-region budget; do not infer that closure from semantic injectivity. The degree bound also bounds the code namespace: one short alias with targets $P_i00$ for arbitrarily many equal-length distinct prefixes creates arbitrarily many class-one contexts. The present fixed-length unique-code interface cannot cover that family unchanged. An aggregate or prefix-funded route must replace individual code multiplicity. This is an interface obstruction, not native impossibility. Arbitrary alias/residual scale ratios and independent-store joins remain separate.

Entry 129 remains an imported growing-alphabet matching/positive-path bound; no native lower bound or summary extraction follows. Replay, affine-fragment, persistent causal and unary coherent cubical controls keep their existing scope. Native recognition yields a PEG for reversal of the complete checked language.

The maintained proof, code, failure analysis and handoff live in GitLab. The existing Site receives generated styled readings and a highlighted source page with exact download. This is a research-program update, not a new paper. Root CLAIMS.yaml is unchanged.

Scheduled 116 to 117; mathematical 132 to 133, window four to five, reassessment 128 retained. Native construction 69 to 70. Next 118 is research; dedicated primary-proof review and editorial duties both remain at 120. Validation and exact deployment/source parity are recorded in integration state.