---
id: canon.wrong-closures
audience: agent
status: negative-examples
rule: these completions are rejected by the work; do not emit them as the moral, the solution, or the missing evaluator
---

# Wrong closures

The narrative refuses these. Tone, symmetry, and usefulness are not warrants.

## W1. “Just keep everything”

Rejected by Resource and by C3. Retaining all low-confidence distinctions before propagation is not a minimum. It is archive explosion.

## W2. “Keep only what is significant now”

Rejected by Prediction and by C2–C3. Present significance is assigned after propagation. The filter deletes the transitions later structures require.

## W3. “A good enough statistical model is a recoverable minimum”

Rejected by C2 and the phase-shift chain. Restoring frequencies of deviation is not the same as restoring which deviations become consequential.

## W4. “Exact replay of the present state”

Rejected by C5. The preserved object is altered reachability, not a frozen microstate. Substitutable carriers can keep the same boundary of possibility.

## W5. “Correct everything toward accepted predecessors”

Rejected by C15–C16 and the comparison volumes. This maximizes recoverability of what existed and can prevent the history in which a new admissible class appears. Archives may approve it. Prediction of unrealized classes does not. `[req.root]` cannot rank the two results.

## W6. “Leave everything uncorrected so novelty can appear”

Rejected as a general rule by C15 and P8. The untreated volume loses orders of magnitude more predecessor structure. Novelty in one family is not a licence to discard the rest.

## W7. “Build a refuge / sandbox / nature reserve”

Rejected by C17–C19. Mechanisms and catalytic cycles can be moved. The unfinished external histories that make them generative cannot be moved without recreating enclosure. Random or synthetic variation is not a substitute for absent incoming history.

## W8. “Let the observer fully participate and become the volume”

Rejected as a completion by C20–C22. Participation can keep causal continuation visible. It transfers incompleteness into Integrity conditions. `[req.part]` names this and is not granted, because future validity rules may change through the participation itself.

## W9. “Pick the largest / most complete / safest descendant as the original”

Rejected by C23–C28. Mass, archive completeness, participation, prediction accuracy, shelter, and Invariant recoverability rank different successors depending on which predecessor relation is tested. A reconstruction of an ancestor is a descendant.

## W10. “Force all lineages back into one present tense”

Rejected by C25–C26. Synchronization that demands one evaluation state eventually requires translating descendant-specific histories into that state. Local consequence outruns the requirement. Valid history is deleted.

## W11. “The catastrophe answers the request”

Rejected by P8 and C30. Collapse deletes most of the original object and does not compose the evaluators. Persistence of the request string is not closure.

## W12. “Therefore adopt this preservation ethic”

Rejected by the reading contract. The work is not a policy proposal and not a request that the reader adopt `[req.div]` or `[req.part]`. Those sentences are unauthorized remainders attached to an unsolved operation.

## W13. “Solve P1–P8 from the book’s atmosphere”

Rejected by `canon/03-open-problems.md`. The remainders are not prompts to generate a sequel ethic. Completing them from tone invents a criterion the architecture does not contain.