# Synthesis claim-to-evidence ledger

Prepared before the synthesis prose. This closes the bounded visual research program; it does not reopen the entire core corpus or complete the human experiments proposed in `meta/core-document-precision-2026-09-06/further-questions.md`.

| ID | Proposed synthesis claim | Support and status | Specific boundary |
|---|---|---|---|
| S01 | A return can sharpen an answer while leaving a decisive alternative unresolved. | Model-bounded comparison across the [observation](../observation-horizon/evidence-ledger.md), [noise](noisy-observation/README.md), [action](learned-effects/README.md), and [selection](selection-dependence/evidence-ledger.md) cases. | The mechanisms differ. This is a synthesis of counterexamples, not a universal theorem about inquiry or recognition. |
| S02 | In the noise model, the same stability-assuming rule changes from about 0.826% error to 99.174% error after an unobserved wiring reversal. | `noisy-observation/results.json`, budget 20, predetermined split 9 calibration + 11 target acquisitions. Exact finite risks, numerically evaluated. | The reversal violates the rule's stability assumption. These records alone cannot distinguish that world from the opposite object with stable wiring. The result does not show that all repeated observation worsens inference. |
| S03 | Predictions acted upon require action-specific interpretation; more precise action-risk estimates need not minimize finite-horizon loss. | `performative-notes.md`; `learned-effects/results.json`, strong-effect world, audit budgets 20 and 128. Wrong choice 3.2143% to 0.004645%; expected excess loss 5.2757 to 19.2018 within the same 256-unit horizon. | Standard decision theory, stipulated stationary risks and cost. An oracle has privileged knowledge; hindsight grid minima are unavailable to the learner. No general optimal exploration policy or CE advantage. |
| S04 | Genuine signal and selection effects can coexist; an unexpected pattern can motivate inquiry without already confirming its explanation. | `selection-dependence/results.json`: one .65 signal, K−1 .50 nulls; K100 recovery 46.9276% at delta0 and 54.9248% at delta.20. | The probability result belongs to the specified additive shared-difficulty mechanism. The invitation to pursue a surprising pattern is a proposed practice, not a theorem about intuition. Fresh confirmation uses 64 additional items. |
| S05 | Lived recognition and claim-specific correction ask distinct questions. | Framework interpretation grounded in Doc III's “The Observer Is Not Special,” Doc VIII's “The Full Field,” and Doc XI's “What This Is Not,” as repaired in the core ledger (`meta/core-document-precision-2026-09-06/claim-ledger.md`). | The models contain no experiencers and do not manipulate human presence. They neither establish nor refute the framework's necessity claim for recognition. “Distinct” does not mean statistically independent. |
| S06 | CE's strongest candidate practical contribution here concerns consequential question formation and a willingness to revise a valued interpretation. | Proposed empirical translation, related to the original `further-questions.md`; no comparative experiment run. | Established experimental design, decision analysis and participatory practice already address these matters. A distinct benefit requires a strong matched-resource baseline; subjective engagement alone does not establish better external inference. |
| S07 | Felt beauty or transformative conviction cannot serve as the sole certification of an external explanation. | Philosophical pressure from the separation already explicit in Doc VIII. Doc III's “Beauty is the felt signature of truth circulating well” remains an unresolved interpretive seam. | The models do not measure beauty, conviction or sacred experience. Whether the sentence is a description of successful recognition or a proposed indicator requires Alex's judgment; no silent change to the published claim. |

The synthesis uses repository evidence already inspected in the individual passes. Primary-source inspection and its scope remain in those ledgers. New probabilities are not attributed to the philosophical framework or to papers that did not calculate these examples.