MS408

The record · iterations i01–i11

Research timeline

Every iteration, experiment, gate, and milestone — and, most tellingly, the 41 self-corrections where the refutation pass overturned the program's own conclusions before they stood. The honest record is the method.

Dots are colored by kind — ● experiment, ● milestone, ● gate, ● retraction. Entries marked ↺ self-correction are where a first-pass claim was narrowed, walked back, or retracted. Grades are A–D (glossary). A community research track is planned next — see the repo to contribute.

2026-07-05 · G0 · gate · gate-passed

Decision locks signed off

Initial program decisions D1–D8, D10, D11 locked as L11–L20; firewall, harness, and stratification rules set.

methodology · governance
2026-07-05 · i01 · iteration · gate-passed

Foundational program (Phase 0–3)

Built the corpus pipeline, validation harness, and annotation dataset; ran baseline morphology/topic/anchor studies and the first living synthesis.

methodology · pipeline · morphology · entropy
2026-07-06 · G1 · gate · gate-passed

Replication gate passed

35/38 scored published-baseline statistics reproduced (h2, MZ peak, glyph rules within tolerance); the replication firewall gate passed.

methodology · replication · entropy
2026-07-06 · G2 · gate · gate-passed

Annotation gate passed

227-page bulk annotation at schema v0.2-fine accepted; QA passed retuned thresholds; root_type flagged low-confidence.

methodology · annotation
2026-07-07 · G3 · gate · gate-passed

Synthesis gate passed

First living-synthesis flagship accepted with graded claims.

methodology · synthesis
2026-07-07 · G4 · gate · gate-passed

Final i01 gate passed

Post-adversarial-review honest revision accepted; i02 agenda approved; i01 complete.

methodology · synthesis · self-correction
2026-07-07 · i02 · iteration · refuted-and-corrected

Sharpening the envelope by subtraction (E1–E5)

grade C

Word-order/meaning-detector claims tested and corrected; the i01 'conlang best fit' was withdrawn under fair encoding-bracket tuning.

entropy · cipher · methodology · self-correction
2026-07-07 · E1 · experiment · refuted-and-corrected

Is ΔI a meaning detector?

grade C

ΔI value is not a meaning certificate; a meaningless drift-null reaches the same ΔI at the wrong scale, so ΔI cannot settle meaningful-vs-meaningless.

entropy · methodology · self-correction
2026-07-07 · E2 · experiment · refuted-and-corrected

Word-order signal confound

grade B

Reordering signal is intrinsic (block scale 812), not an artifact; the anti-cipher point was corrected — a deterministic type-preserving verbose cipher retains ΔI, reopening that cipher family.

cipher · entropy · self-correction
2026-07-07 · E3 · experiment · refuted-and-corrected

Anchor-hunt power curve

grade B

The 'null informative' claim was oversold; the null excludes only strong balanced anchors, weakening the flagship anchor leg.

anchor · methodology · self-correction
2026-07-07 · E4 · experiment · refuted-and-corrected

root↔leaf masked positive

grade C

An apparent cross-organ visual bundle; the coded 'confirmed' verdict was overclaimed and corrected to suggestive-but-unresolved after a cross-model test.

visual · annotation · self-correction
2026-07-07 · E4b · experiment · refuted-and-corrected

root↔leaf same-source confound test

grade C

Opus-4.8 re-annotation showed the association rides on Sonnet's root label; graded down B→C — leaned toward a labelling artifact.

visual · annotation · self-correction
2026-07-07 · E5 · experiment · refuted-and-corrected

Fair encoding bracket

grade C

Under equal tuning and held-out scoring, no encoding family is robustly distinguished; i01's 'conlang best fit' withdrawn, two bugs caught en route.

cipher · conlang · methodology · self-correction
2026-07-08 · i03 · iteration · refuted-and-corrected

Reconstruction + variable-introduction agenda (E6–E10)

grade C

Hardened negatives (no family distinguished, no statistic localizes meaning) while reopening the root↔leaf candidate; excluded verbose cipher but revived the abjad class.

cipher · morphology · visual · self-correction
2026-07-08 · E6 · experiment · refuted-and-corrected

Verbose/nomenclator cipher reconstruction

grade C

Verbose cipher deductively excluded on the ED1 morphology network; an abjad of real Latin reaches it, refuting the first-pass 'needs constructed morphology' and reviving the abjad class.

cipher · morphology · self-correction
2026-07-08 · E7 · experiment · confirmed

Fine-granularity anchor hunt

grade C

18 raw anchor 'discoveries' were caught by the permutation null (p=0.48); no anchor above chance at higher power. Anti-labelled-herbal null holds.

anchor · methodology · self-correction
2026-07-08 · E8 · experiment · refuted-and-corrected

Whitened encoding bracket

grade C

The whitened ranking is a regularisation artifact (ill-conditioned covariance); no family stably distinguished. The 'whitening confirms E5' claim was withdrawn.

cipher · methodology · self-correction
2026-07-08 · E9 · experiment · refuted-and-corrected

VMS-coordinate meaning localization

grade D

After two refutations of trivial endpoints, no structural axis localizes meaning — the structured-meaningless adversary out-scores meaningful text and the VMS sits below both; a pre-registered commitment was retracted.

entropy · morphology · methodology · self-correction
2026-07-08 · E10 · experiment · refuted-and-corrected

root↔leaf third vision-model rater

grade C

Haiku 4.5 also reproduced the leaf association, overturning E4b's 'Sonnet-specific artifact'; root↔leaf reopened as strongest referential candidate but unconfirmed.

visual · annotation · self-correction
2026-07-09 · G5 · gate · gate-passed

Post-i02/i03 synthesis accepted

Tim signed off the honest post-review synthesis (E1–E10 folded in); L37 locked and forward direction set.

methodology · synthesis
2026-07-09 · i04 · iteration · inconclusive

E10 confirm-or-kill (E11–E12)

grade C

The root↔leaf bundle survived a style control but the independence test showed model annotation is underpowered/untestable; E10's positive candidate withdrawn to UNRESOLVED.

visual · annotation · self-correction
2026-07-09 · E11 · experiment · confirmed

Illustration-style control for root↔leaf

grade C

The bundle survives conditioning on palette-richness style for both models; not a palette-style artifact (partial control only).

visual · annotation
2026-07-09 · E12 · experiment · refuted-and-corrected

Independent-lineage (GPT-5.1) root rater

grade C

A non-Anthropic rater does not reproduce the association; a first-pass 'KILLED' was corrected to UNRESOLVED-underpowered — model annotation cannot adjudicate, only a human panel can.

visual · annotation · self-correction
2026-07-09 · i05 · iteration · confirmed

Mid-level linguistic program (E13–E17)

grade C

Built a null-correction framework; found the VMS lacks the surface content>function collocation gap and has weak word-class structure, with A/B differing lexically not grammatically.

morphology · syntax · self-correction
2026-07-09 · E13 · experiment · inconclusive

Function/content bimodality probe

grade D

Three global operationalisations failed calibration; no VMS verdict issued (harness-first). Prompted the E13b redesign.

syntax · methodology · self-correction
2026-07-09 · E13c · experiment · refuted-and-corrected

Function/content, order-shuffle null

grade C

Calibrated; narrowed to a surface finding — the VMS has near-chance collocation and its most-frequent words are template-like, the opposite of flat function words. First substantive i05 result.

syntax · self-correction
2026-07-09 · E14 · experiment · confirmed

Distributional word-class induction

grade C

Null-corrected POS induction shows the VMS has only weak word-class structure (~0.13–0.19× real language), calibrated on real languages.

syntax · morphology
2026-07-09 · E17 · experiment · refuted-and-corrected

Content-controlled A-vs-B contrast

grade C

The apparent A/B word-class difference vanishes within the herbal section — it is a section/content confound, not dialect; corrects E14b as E12 corrected E10.

syntax · dialect · self-correction
2026-07-09 · paper-v1 · milestone · released

Paper v1 preprint

First firewall-sourced MS408 preprint packaged via the package-paper skill (through E12).

publication
2026-07-14 · paper-v2 · milestone · released

Paper v2 (folds i05)

Preprint updated to fold in the i05 mid-level linguistic program.

publication
2026-07-15 · i06 · iteration · refuted-and-corrected

Cryptanalytic direction (E18–E20)

grade B

Concluded the cipher-of-real-prose class is EXCLUDED via the joint low-entropy + retained-ΔI + weak-syntax signature; this universal headline was later retracted by i11.

cipher · entropy · syntax · self-correction
2026-07-15 · E19 · experiment · refuted-and-corrected

Joint-signature cipher test

grade B

No cipher of real prose matches the VMS's low-h2 + weak-syntax combination; the negative is grade B and the circular 'favours generation' positive was dropped.

cipher · syntax · self-correction
2026-07-15 · E19b · experiment · confirmed

Language-universality control

grade B

The exclusion holds across typologically diverse languages including Hebrew (a native abjad); an order-scrambling transposition cipher is the only surviving weak-syntax lead.

cipher · syntax
2026-07-15 · E20 · experiment · refuted-and-corrected

Transposition closure

grade B

Retained-ΔI and weak-syntax are mutually exclusive under any word-reordering cipher, yet the VMS has both — closing the cipher-of-real-prose class (headline later retracted).

cipher · entropy · self-correction
2026-07-15 · paper-v3 · milestone · released

Paper v3 (folds i06)

Preprint updated to fold in the i06 cryptanalytic direction (carried the later-retracted cipher exclusion headline).

publication
2026-07-15 · i07 · iteration · refuted-and-corrected

Characterising the generative class (E21–E22)

grade C

A context-free positional/template generator matches entropy + block-ΔI + weak-positive syntax but never morphology connectivity, lexical reuse, or frequency slope — a partial account only.

generative · morphology · self-correction
2026-07-15 · E21 · experiment · refuted-and-corrected

Minimal positional generator + ablation

grade C

A first-pass 'class sufficiency [B]' was overturned by the refutation pass — constants were grid-selected to the VMS bands and the weak-syntax leg was shuffle-passable — narrowed to an informative negative.

generative · methodology · self-correction
2026-07-15 · E22 · experiment · confirmed

Genericity/coupling sweep

grade C

A broad a-priori grid never reaches the VMS's ED1, TTR, or Zipf; the context-free bag-of-slots is INSUFFICIENT and demands an added reuse/smaller-lexicon mechanism.

generative · methodology
2026-07-15 · i08 · iteration · refuted-and-corrected

Adding frequency concentration (E23–E24)

grade C

Across three generative families, none reproduces the full 8-axis signature over the swept ranges; the summary statistics are mutually coupled. This hard constraint was later walked back by i09.

generative · self-correction
2026-07-15 · paper-v4 · milestone · released

Paper v4 (folds i07–i08)

Preprint folds the generative-class characterisation with refutation-corrected wording.

publication
2026-07-16 · i10 · iteration · refuted-and-corrected

Symbols-as-values direction (E27–E28)

grade D

The quantitative-register hypothesis finds no support and closes at grade D — a positional-numeral sub-type is shape-excluded and no robust ordinal anchor appears in the zodiac rings.

register · numeral · self-correction
2026-07-16 · E27 · experiment · refuted-and-corrected

Symbol quantification

grade D

The VMS's strong positional specialisation (0.74) shape-excludes a positional-numeral sub-type; a non-positional value scheme remains untouched (→ E28).

register · numeral · self-correction
2026-07-16 · E28 · experiment · refuted-and-corrected

Angular/ordinal anchor in circular diagrams

grade D

A sensitivity-confirmed Mantel test on all 12 zodiac rings finds no robust ordinal structure; the lone f73r hit is not reproduced — register stays D.

register · anchor · self-correction
2026-07-16 · i09 · iteration · refuted-and-corrected

Is the i08 coupling real? (E25–E26)

grade C

With ED1 made an independent knob, i08's gross incompatibility collapses to a shallow ~0.03 near-miss frontier — a retraction of the i08 hard constraint, not a promotion to 'no constraint'.

generative · morphology · self-correction
2026-07-16 · E25 · experiment · refuted-and-corrected

Decoupled-ED1 type-lexicon generator

grade C

Making ED1 an independent knob shows E24's saturation was a small-char-space artifact; the coupling largely dissolves to a shallow h2↔ED1 near-miss (initial over-read later curbed).

generative · morphology · self-correction
2026-07-16 · E26 · experiment · refuted-and-corrected

Word-length variance vs the h2↔ED1 frontier

grade C

Word-length variance lands ED1 in-band jointly with ΔI/TTR/Zipf; the residual is a ~0.03 h2 near-miss plus a fixable length artifact — the frontier is all-but-crossed.

generative · morphology · self-correction
2026-07-16 · paper-v5 · milestone · refuted-and-corrected

Paper v5 then v5b

Preprint folds i09; v5's stronger 'signature doesn't constrain mechanism' phrasing was corrected by a refutation of the walk-back itself to v5b.

publication · self-correction
2026-07-16 · methods-v1 · milestone · refuted-and-corrected

Methods paper v1 then v2

Packaged the transferable adversarial-self-correction methodology paper; v1's inferential over-claims were corrected to v2 by its own refutation pass, and the refutation archive was created.

publication · methodology · self-correction
2026-07-17 · E29 · experiment · refuted-and-corrected

Naibbe cipher vs the i06 discriminators

grade C

A first pass read 'i06 confirmed [B]'; the refutation killed it — word-boundary Latin sits in the VMS ΔI band and ~82% of Naibbe's ΔI loss is respacing, exposing that ΔI is a homophony detector and weakening i06's ΔI leg.

cipher · entropy · self-correction
2026-07-17 · E30 · experiment · retracted

Cipher exclusion re-examination (multi-seed)

grade C

On word-boundary Latin: order-preserving ciphers robustly EXCLUDED (strong syntax), but verbose+homophonic (≈ Naibbe) NOT excluded — i06's universal cipher headline RETRACTED, robust core kept.

cipher · syntax · self-correction
2026-07-17 · E31 · experiment · refuted-and-corrected

Harden the fc_z/wc_z syntax discriminators

grade B

Deconfounded syntax measures firm the order-preserving exclusion (6.8σ/8.0σ after v6b correction of a fabricated ~30σ); the homophonic class stays inconclusive. A recalled-number firewall slip was caught and corrected.

cipher · syntax · methodology · self-correction
2026-07-17 · i06-retraction · retraction · retracted

i06 universal cipher headline retracted

The 'cipher-of-real-prose EXCLUDED' headline (papers v3–v5b) is retracted to 'order-preserving ciphers excluded; verbose+homophonic (Naibbe-class) not excluded' — the VMS-as-cipher hypothesis is viable on the program's own analysis.

cipher · self-correction
2026-07-17 · paper-v6 · milestone · refuted-and-corrected

Paper v6 then v6b + methods v3

Paper v6 retracts the i06 cipher exclusion; its own refutation caught a fabricated ~30σ and over-correction → v6b, and methods v3 adds the i06/Naibbe self-correction example.

publication · cipher · self-correction
2026-07-17 · release-tier0 · milestone · released

Open-source release Tier 0

Repo packaged as a public evaluator: ms408.evaluate() API, firewall-clean reference bands, Apache-2.0 license, docs; 174 tests pass. A subsample-without-replacement bug fix put the VMS back inside its own bands.

release · methodology
2026-07-22 · i11 · iteration · refuted-and-corrected

Engaging concurrent work (Naibbe/Parisel) + release

grade B

The make-or-break Naibbe engagement exposed and corrected i06's ΔI-leg confound and retracted the universal cipher headline; also shipped the open-source release tiers and arXiv bundles.

cipher · methodology · release · self-correction
2026-07-22 · E33 · experiment · refuted-and-corrected

Block-scale like-for-like ΔI

grade C

A first pass wrongly found a homophonic cipher reaching the VMS corner ('ΔI leg dead'); the refutation showed the corner was a homophony-marker h2 artifact — under a fair model no config reaches it, so block-scale ΔI weakly separates verbose+homophonic ciphers.

cipher · entropy · self-correction
2026-07-22 · release-tier1 · milestone · released

Open-source release Tier 1

Value-pinning tests, a Naibbe worked example, a reproduce-the-paper --verify path, and CI added to the public evaluator.

release
2026-07-22 · release-tier2 · milestone · released

Open-source release Tier 2

Hardened CI, CONTRIBUTING, tutorial, and wired docs; Tier-2 packaging marked done.

release
2026-07-22 · arxiv-bundle · milestone · prepared

arXiv submission bundles prepared

Build/check arXiv submission bundles prepared for paper v6b and methods v3 (no submission yet).

publication · release
2026-07-25 · paper-v7 · milestone · released

Paper v7

Preprint folds the E33 block-scale ΔI result and announces the open-source release.

publication · release
2026-07-25 · public-site · milestone · released

Public GitHub Pages site

Scoped and shipped the Astro v1 public site (docs + papers), later extended with a library/education portal and media feed.

release · outreach
2026-07-26 · publication-h2-2026 · milestone · planned

H2-2026 publication planning package

Venue landscape, scoring rubric, and four pathways for placing the constraint-envelope (Paper A, v7) and methods (Paper B, v3) papers; every venue rejects perceived 'solutions' and welcomes the hypothesis-shrinking posture.

publication · planning