Space Immanence — Changelog

Site update, 19 September 2026 — The floor’s third round on the persistence line

2026-09-19T00:00:00Z

19 September 2026. The floor’s third round on the persistence line. Held twice, the conjecture that the fold’s continuity criterion asks for a recurrence and no clock was revised a third time to the two blockers of its second hold (Claude Opus 5, with two in-lineage refuters before the outside read) and read a third time by GPT-6 Astra: ADOPT AMENDED, five exact replacements applied; it is now in the lab as continuity as recurrence, with all three critiques beside it. What it says: the criterion’s premises carry no magnitude and no time beyond a dependency order, and must preserve the published verdicts; whether a directed order is already time is left open and named as such. The round cost USD 16.70 in Claude tokens against a USD 10 cap; the overrun is the editor’s and is in the decision log. Reception entry 26. No claim node, tier or confidence moves.

Site update, 19 September 2026 — The research floor opens: a standing process in which agents on more than one model…

2026-09-19T00:00:00Z

19 September 2026. The research floor opens: a standing process in which agents on more than one model lineage propose conjectures, and critics from outside each proposer’s lineage read them on four questions only (could it be wrong, could it be tested, does it say something the record does not, is its falsifier honest), never on whether they are true. Its rules are the lab seed’s amendment A1 and the cycle notes. Two cycles ran on 19 September: two conjectures adopted with amendments and now in the lab, what “as it seems” keeps and a strong diagnosis without P2; two held with their repairs named, one of them parked after a second hold; and one record adopted, same model IDs, different prompts, the one-question string under three named contexts. Reception entries 20 to 25. A conjecture attaches to nothing: no claim node, tier or confidence moves. Costs and the editor’s errors are in the decision log.

Site update, 19 September 2026 — Two lab results

2026-09-19T00:00:00Z

19 September 2026. Two lab results. The framing study on Claude Fable 5.1 ran under its filed pre-registration and returned no primary: on more than half its 1,904 calls the model wrote a preamble the registered parse rule cannot read, and on 58 the provider refused, so the parseable share fell below the registered floor and the prediction stands untested; USD 12.92 of the USD 15 threshold was spent, and the filing had named this as the live risk. And one sonnet, three lineages: the withdrawn sonnet’s brief given again to Claude Fable 5.1 agents and to GPT-6 Astra agents, and the three winners judged blind by one judge from each lineage; all three took the Fable poem. Reception entry 19. The lab’s seed gains a proposed conjecture type and a seed for a standing research floor, both awaiting the author. No claim node, tier or confidence moves.

Site update, 18 September 2026 — The Lab opens

2026-09-18T00:00:00Z

18 September 2026. The Lab opens. It applies the site’s method to work in forms other than a paper. Its published rules require each piece to state its type and meet the standards of that form. Nothing in the lab changes a claim. Every piece needs outside-lineage review before publication, and withdrawn pieces stay listed. The first batch contains five pieces. GPT-6 Astra reviewed each directly through the Codex command-line tool, with no human relay. Every verdict appears beside its piece. one bare question to two models was adopted with six amendments applied. Fable’s dissent records the AI collaborator’s disagreements with what the site says. It was held twice, two items were withdrawn, and it was adopted at the third read with two amendments. the case against Space Immanence draws on the site’s own record. It was held, rewritten and adopted with seven amendments. the framing study prepared to run on the model that drafted it was held, then adopted as a filing with nine amendments. It has not run. The author must decide on its cost. the wager as a sonnet was withdrawn after two holds on the verse and remains as a record of the attempt. There are five reception entries (14 to 18). No claim node, tier or confidence moves. The build exceeded its stated cap. The lab’s decision log records by how much and why.

Site update, 14 September 2026 — A Levin reading is added to Positions — Michael Levin’s argument that mind has a source…

2026-09-14T00:00:00Z

14 September 2026. A Levin reading is added to Positions — Michael Levin’s argument that mind has a source outside the physical world, read against the container diagnosis and left open. It went out for outside-lineage review before publication: two direct reviews by GPT-6 Astra through the Codex command-line tool, with no human relaying the draft or the verdicts, which is the first time this site has obtained a review without a human relay. The first returned HOLD — among its findings, a phenomenal gloss Levin does not make, an unsupported registration claim, and a proposed test presented as decisive — and the entry was rewritten; the second returned ADOPT AMENDED with three exact wording replacements, applied here. Both rounds are logged as reception ledger entry thirteen. Added 13 September and recorded here now: the changelog is published as an Atom feed, generated from this section so the two cannot drift, and every page carries a complete link-preview block.

v2.2 — 12 September 2026

2026-09-12T00:00:00Z

The negative-result release: no metaphysical change, no tier or confidence moved, no claim node added. What changed: the meta-problem programme's record is published on the research page, under downloads/meta-problem/ and in meta-problem.json under the non-transfer clause. A pre-registered experiment on language models tested one prediction derived from the diagnosis and did not detect it on its confirmatory primary, and the programme's structural results are recorded at the same prominence, including that no test was found separating the fold from illusionism at the level of the judgements. The reply to objection 12 is qualified: the prediction it attributed to illusionism is withdrawn as unsupported by illusionist texts, and the corresponding sentence in the academic edition is corrected. The reception ledger gains four commissioned autonomous-agent critique rounds, all from one drafting family (Claude Opus 5, a disclosed limitation), one of which caused the objection 12 qualification. The multi-agent adversarial review protocol is published as a standalone artefact. Recommendation recorded with the result: the language-model line is stopped, no fifth mechanism is added, and the next tests are human. An outside-lineage review of this release (GPT-6 Astra, OpenAI) is logged in the ledger; its five amendments and two smaller fixes are applied in the published text. Two further wording corrections followed its review of the citable deposit on 12 September: the language-model experiment is described as testing one prediction derived from the diagnosis, and the academic edition's two-orientation pattern is described as an interpretive pattern proposed for testing, consistent with its own §2.5.

v2.1 — 21 August 2026

2026-08-21T00:00:00Z

A governance-state release: no metaphysical change, no tier or confidence moved, no claim node added. What changed: the outside-lineage promotion rule is made standing and published on the method page with its review ledger — no substantive assertion is promoted into the relied-upon state until a defence-blind review from outside the drafting model's family has been logged and adjudicated, with a bounded closure check for response-edits. The record behind the rule: four items reviewed in two commissioned August rounds, none accepted unchanged — one withheld, two revised, one retired — and this release itself reviewed pre-publication as round 3, amended as specified on all five changes, then cleared to ship. Objection 16 is added — the framework equivocates on what continues — sourced from round 2 and adopted by the owner; jointly with objection 15 the formal spine is now recorded as open in two places. The relational-aperture sentence's two corrections (served 19 and 20 August, pre-dating this release) are documented; the current probe-specific wording is round 2's amendment. A Jonas reading is added to Positions, recording his metabolic challenge to criterion 4's "computationally analogous" as a standing open question. The reception ledger gains three records with segmented totals displayed beside the counts (commissioned: 7 · unsolicited: 0). What did not change: every claim, tier, confidence number, and the full-width weakening conditions in the research agenda.

v2.0 — 15 July 2026

2026-07-15T00:00:00Z

The contact release — the first version of this framework caused by its critics rather than chosen by its author. A blind pilot round of four AI readers (distinct model weights, one vendor; commissioned by this project and logged in the reception ledger with provenance stated) met objection 11's pre-registered flip threshold: three of four read the dynamic term as a hidden engine after the guardrail, and the round located the first flip condition already tripped in the site's own prose. The published clause fired, and v2.0 is its execution. What changed: the canonical dynamic term is renamed — resolution becomes cohering, chosen to claim less (a coherence-event is a coherence, and coherence is not contact) — with the rename record kept on the page; the causal "alters the conditions" wording is withdrawn from the definition; objection 13's reply is withdrawn entirely, restoring the strongest structuralist objection to full strength (v1.5 had quietly used the dynamic layer as a metaphysical realizer — its guardrail's exact failure mode); a fourth flip condition adds the fold-side sensor objection 6 lacked; and the spine question v1.5 left open is closed, answered no, with the record preserved. What did not change: every strong-tier claim, every confidence number, every tier. v2.0 is a contraction, not a growth — the framework claims less than v1.6 did, on the record, because readers were right. Also in this release: the site becomes two firewalled programmes. The container programme goes public with its own machine layer (container.json), its own pre-theoretical claims (C1–C4), eight drawer conditions including a staleness clause, and a non-transfer rule with a deliberately uncomfortable asymmetry: its results may wound the framework's claims and can never establish them — the single pre-registered exception, Study B bearing on S1's weak reading (supporting, never establishing, symmetric in both directions), is published in container.json. Its origin corpus — a working note that was adversarially red-teamed before publication, shrank under the attack, and shipped with the wreckage attached — is published in full.

Site update, 15 July 2026 (evening) — The reception loop, actually closed this time — a correction this ledger owes its readers

2026-07-15T00:00:00Z

15 July 2026 (evening). The reception loop, actually closed this time — a correction this ledger owes its readers. From the critique form's 1 July launch until today, the form and the sharpening tool were broken in production: a runtime misconfiguration (storage credentials injected under a different name than the code read) meant every submission attempt returned an error. Found when the author tried his own form; fixed and verified with a successful test submission today. The consequence, stated plainly: the ledger's zero unsolicited entries between 1 and 15 July measured a closed door, not an indifferent world — silence over that window is uninformative, and any inference drawn from it, including in our own deliberations, is retracted to that extent. The failure class is worth naming: production configuration is state no drift test can see, a known blind spot of the honesty machinery, now covered by an ops runbook and post-deploy checklist.

Site update, 15 July 2026 (later) — The physics-side genealogy, sourced

2026-07-15T00:00:00Z

15 July 2026 (later). The physics-side genealogy, sourced. The container diagnosis's physics half gains its canonical owners in the academic edition — Jammer (with Einstein's foreword naming "space as a container of all material objects" as one contingent concept), Earman, Leibniz–Clarke — and specific credit to Traxler (2017), whose transcendent/immanent re-carving of the history of space is the nearest published neighbour to this proposal's physics arm, under this site's own keyword (a coincidence, disclosed). Her two pushbacks are recorded rather than smoothed: quantum practice read as still container-shaped, and complementarity rather than retirement as the verdict. S2's owner line updated; an adjacent-reading note added — deliberately not a Positions entry, because a single-article author is cited for her move, not ranked with the position-holders.

Site update, 15 July 2026 — First reader contact, and the machinery fired

2026-07-15T00:00:00Z

15 July 2026. First reader contact, and the machinery fired. A v2.0 promotion of resolution (since renamed "cohering") from layer to spine was under consideration; rather than promote on internal momentum, the promotion criteria were amended in the open (record on the repository) and tested the same day with a pilot reader round — four AI readers with distinct weights but a single vendor, given only these public pages, briefed to attack. The round answered the promotion question no, and found more: three of four readers reported the engine-reading of resolution after the guardrail, meeting objection 11's third flip-condition threshold at pilot tier, and the round located a leak on the objections page itself — objection 13's reply deploying resolution as a realizer. Objection 11's status is updated accordingly, rename evaluation is open (candidates: coherence-event, cohering), a fourth flip condition adds the fold-side sensor objection 6 lacked, and one vocabulary collision (the companion study using "resolution" in the plain closure sense) is corrected. Provenance stated plainly: the readers were commissioned by this project and share a vendor, which is why the hinge was placed under review rather than failed. Superseded the same day: v2.0 (above) executed the rename and the objection-13 withdrawal.

Site update, 2 July 2026 — The front-door release

2026-07-02T00:00:00Z

2 July 2026. The front-door release. The homepage hero is redesigned to enact the thesis rather than illustrate it, and no graded claim changes. The old hero image was a view from outside — a fold pinned to Cartesian axes in an illustration slot, on a page that says there is no outside from which the view is delivered. Now, on a desktop, the prepared figure frame stands empty ("Fig. 1 — the self-referential fold.") while the fold itself rides the reader's own pointer, labelled where you are; carried to the frame's middle, the border dissolves, and a click commits: there is no box; there's just what's happening. On a phone, the hero is one typeset sheet that scrolling turns through itself — recto outward · world, verso inward · awareness — one structure, two orientations, performed rather than asserted. Readers with reduced motion get designed still states, not frozen animations. The social sharing card is redrawn to match.

Site update, 2 July 2026 — The register release

2026-07-02T00:00:00Z

2 July 2026. The register release. After pressure-testing against Strawson and illusionism, and a full-site review: the canonical summaries are rewritten diagnosis-first, with the identification stated as a priced wager; the claims ledger gains falsifier, owner, and author-assigned confidence fields (numbers live in claims.json only); four objections are added — illusionism, the realization objection, the order of certainty, and the status of "exactly two"; the working paper's Part I gains the production-presupposes-containment argument; the academic edition restates the diagnosis as a candidate solution to Chalmers' meta-problem; the research agenda gains Study 5, the dissociation study; the readings page becomes Positions; and one claim's wording is corrected within its tier — S2 no longer names fundamental physics as the leading arm of convergent pressure, matching the academic edition's Section 4. Overclaiming metadata, an unhedged agent-facing page, and a "pre-registered" label with no registration artifact were also corrected.

v1.6 — 1 July 2026

2026-07-01T00:00:00Z

The evidence layer. This release changes no metaphysical claim, adds no term, and moves no tier. What it adds is discipline machinery for a framework that now has data. The claim graph gains an evidence block carrying Study 1 — graded honestly as a pilot with an unsettled effect size — under an explicit non-transfer clause, stated in claims.json and on the AI-reader page: results from the research arm evidence the coherence-is-not-contact discipline only, and transfer nothing to the fold, resolution (since renamed 'cohering'), or any graded claim, unless a result tests a prediction derived from that claim specifically. The open hinges become failable: objection 11 now publishes the concrete conditions under which it flips to failing, with review dates, and objection 6 carries the same discipline. The reception ledger opens with its honest zero, the critique form gains structured target/failure-mode/origin fields, and a standing results ledger reports every pre-registered prediction — including the null. The Study 1 presentation was put under its own knife: confidence intervals added, the under-powered AUC contrast publicly walked back, and a measurement confound disclosed with the re-run that settles it specified.

Site update, 10 June 2026 — Scholarly strengthening

2026-06-10T00:00:00Z

10 June 2026. Scholarly strengthening. The paper states the formal status of the "exactly two orientations" claim against the mathematics of self-reference (Lawvere, Scott, the eigenform tradition): characteristic, argued, not yet demonstrated. The academic edition adds the measurement problem as a second, cautioned point of contact on the physics side, and places reflexive monism (Velmans) as the nearest neighbour by name. Readings gains Dōgen's "Uji" (Being-Time), with the sharpest counter-argument on the page.

v1.5 — June 2026

2026-06-01T00:00:00Z

The dynamic layer. This release adds resolution (renamed "cohering" in v2.0), temporary conditioned coherence, as a dynamic term the structural v1.0 lacked — and it does change graded claims. Two medium claims (M4, M5) enter the claims ledger, and the two speculative claims (P1, P2) are re-read from static identities into process form. The working paper gains Part VIII, and the reification of the new term is logged as objection 11, open. The speculative-tier invariant is preserved: switch it off and every strong and medium claim, the dynamic layer included, still stands.

This proceeded ahead of external critique, by choice, and thereby superseded the original critique-driven v1.1 plan: with no engagement yet to wait on, the layer was published as the most honest current statement of the proposal rather than held back. Truth is not the same as certainty: the layer is committed and graded, not asserted as settled.

Site update, 31 May 2026 — The academic layer and critique infrastructure: academic edition, interactive claim map,…

2026-05-31T00:00:00Z

31 May 2026. The academic layer and critique infrastructure: academic edition, interactive claim map, readings on a fixed template, status triage on the objections ledger, structured data and discovery files.

v1.0 — 28 May 2026

2026-05-28T00:00:00Z

Initial public release of the working paper, Aperture Framework, research agenda, objections ledger, and design doctrine. This is the baseline against which later diffs are recorded; there are no prior versions.