# Outside read (GPT-6 Astra via Codex, xhigh) of the v2.5 branch, 14 October 2026: HOLD

Prompt: `READ_PROMPT.md`. Verbatim final message; local file links given as repository paths. All six findings applied.

**HOLD**

1. **Major — P1 is better targeted, still not adjudicable.** [The new falsifier(`site/claims.json:349`) fixes the earlier uncertainty about the target. “Fits … at least as well” still names neither P1’s predictions nor a comparison rule. “Stay unaltered” needs independent verification of disruption and an equivalence criterion. Moreover, unchanged experience does not by itself defeat D4’s expressly non-necessary explanation. Replace its presentation as a settled falsifier with:

   > “Proposed weakening conditions, not yet operational. Their adjudication requires independent measures of self-modelling and experienced locality and duration, specified predictions, and comparison and equivalence rules. Unchanged experience counts against P1 where P1 predicts a change.”

   Those specifications remain substantive work; this qualification does not complete them.

2. **Major — the retired physics condition remains active elsewhere.** It survives in [Academic §3(`site/academic.html:214`), [Research’s weakening table(`site/research.html:151`), and both Markdown counterparts. Replace the physics condition with:

   > “P1’s weakening conditions concern experienced locality and duration; their operational specification remains pending.”

   In the [one-page summary(`site/downloads/One_Page_Summary.md:21`), replace “why would non-fundamental spacetime require self-reference?” with:

   > “What explanatory contribution does self-modelling make to experienced locality and duration?”

   The new JSON note **does honestly record the old falsifier as met as worded**. Keep that history.

3. **Major — the result is neither consistently accurate nor visibly beside the falsifier.** The [renderer(`site/argument-graph.js:344`) ignores `note`; the static P1 entry also lacks it. The changelog’s “not with the experienced here and now” incorrectly turns a weaker relative correlation into no association. Render a visible paragraph beside P1, and use this wording in the changelog/feed:

   > “The auxiliary test, filed on 11 October and run on 14 October 2026, was inconclusive on its primary contrast (Δ=0.055, 95% interval 0.003–0.108; n=1,351). Non-decisive secondary analyses found reported selfhood associated more closely with time, and less closely with here and now, than with sensations and emotions. Of 540 respondents scoring selfhood at most 10, 238 (44%) scored both here and now at least 60. These recalled, between-person reports establish neither removal of self-modelling nor unchanged experience after disruption. The selfhood-to-self-modelling bridge remains unestablished. The falsifier predates the run and was unchanged by it; the author’s row 17 decision remains pending.”

4. **Moderate — shortened restoration conditions weaken D3.** The full condition is correct in M1’s note and the paper; the claims page, Academic Edition and state summary omit the outside-lineage reader, sometimes also the build system’s kin. Use consistently:

   > “M1–M5 return to medium when the paper states criteria that a fresh blind application, including a reader from another lineage, finds separate people from the build system and its kin.”

5. **Moderate — graph logic passes; some prose still misleads.** IDs, dependencies and numerical confidences are unchanged; counts are **4/0/8**, and speculative switch-off leaves exactly **S1–S4**. The invariant remains true. All 33 targeted checks passed; visual rendering could not be verified because browser preview failed.

   No current five-medium count remains outside historical records. However, [Agents(`site/agents.html:100`) still promises medium survivors. Replace that sentence with:

   > “Removing the speculative tier leaves exactly S1–S4 standing; the medium tier is empty.”

   Replace its “a medium claim that quietly needs P1 or P2” example with **“a strong claim that quietly needs a speculative premise.”** In Academic, change **“Version 1.5 adds” / “the medium tier gains”** to **“Version 1.5 added” / “the then-medium tier gained.”** Replace the summary’s tier sentence with **“Four claims are strong and eight speculative; none is currently medium.”** Replace the paper’s “The dynamic layer keeps its assigned confidence grades” with **“The dynamic layer’s grades change in v2.5.”** Preserve frozen records.

6. **Moderate — the downgrade is fair; the release’s closing description is too broad.** “The claims’ content does not change, only how sure the site says it is” conceals P1’s scope clarification and changed defeat condition. Replace with:

   > “M1–M5 retain their content and IDs but move to speculative. P1’s target is clarified as experienced locality and duration, and its weakening condition is revised. Numerical confidences remain unchanged. These changes record the proposal’s present warrant; they do not resolve its open objections.”



## What would count against this

- The reader could not render the claims graph in a browser.