privacy · proofs · compute
v3.2 · checksummed
2026-09-13 · v3.1

Peer review of v3.1 — what v3.2 fixes

Published unedited, as the v3.0 review was. Line numbers refer to the v3.1 source at commit 5e6f242. The P0 and P1 lists are applied in v3.2 (commits 9f84fbc, 94d3eab); the P2 surface items are tracked in FOLLOWUPS.md.

Date: 2026-09-13 Scope: latex/ at tag v3.1 (commit 54d8343; source tree at 5e6f242) as ground truth: front matter, Parts I–VI, back matter — 512 pp in the pdflatex build. README.md, PROJECT_REFERENCE.md, FOLLOWUPS.md and outreach/ read as the surface. The frozen brief-content-v3.1.tex and essay-content-v1.3.tex were grepped for the same defects, not re-reviewed. Purpose: the same six criteria as REVIEW-v3.0.md — originality, accuracy, consistency, cogency, soundness, utility for Kardashev Labs — read fresh, then a fix plan for v3.2. Method: every LaTeX file was read in full by one reviewer in one pass (not refereed by Part, unlike v3.0), with every numerical claim re-derived where the inputs are in the text, every cross-Part threshold grepped against every other statement of it, and the Sources file checked by script (key integrity, URL/DOI presence, scope notes). The build gate was run (check.sh: PASS, 512 pp, reference integrity clean). One external fact (Monero FCMP++ activation) was web-checked and returned conflicting sources; it is flagged, not asserted. Line numbers are exact at 5e6f242.

Status of this document: review. Remediation is v3.2; nothing here has been applied to latex/.


0. Verdict

v3.1 did what it said. Every P0–P2 item in REVIEW-v3.0.md is present in the source and, where it was a mechanism, it is coherent: the MATMUL bound is componentwise with an adversarial suite; refund safety is a bounded-time no-loss exit; viewing keys are a disclosure-scope model; the workload registry is one table; Tornado Cash runs to Storm; Kardashev Labs exists and is stack-first/coin-optional; RL16–18 read on link 9 with maturity gates and a power calculation; the hedge-vs-Work-Credits question is answered once. The monetary core of Part II (fee incidence → what the five conditions establish → pledgeability → state-contingency → denomination) is the strongest sustained argument in the document and survives an adversarial read intact.

What v3.1 did not do is propagate. The remediation was applied at the site of each finding and the document’s summaries and definitions were left as they were. The result is a thesis whose body is at v3.1 while its Thesis-in-Plain-Language chapter, its Chain-at-a-Glance, its Key Definitions box, its glossary, its KPI instrument table, and its formal-definitions appendix are at v2.x on exactly the points v3.0 and v3.1 changed: refund_safe = 1.0 (four sites), “every participant has already produced it” (glossary — the formulation the v3.0 late refinement retracted), τ = “fees plus burns” (glossary, against RL14’s gross-fees rule), “credit-weighted voting” for Work Credits (Key Definitions, against Layer 6’s “never Work Credits”), “three bridges are Medium” (front matter, against a table with six), the v3.0 tier multipliers still argued from in Part II, and a fee partition stated four incompatible ways across Parts II–V. None of these is a new error of reasoning. All of them are the kind a hostile reader finds in the first twenty pages and uses to stop reading.

Two defects rise above stale text. part-ii.tex:2111 says “no red line covers” the collateral–capacity spiral; Red Line 15 is that spiral. And part-v.tex:220 caps governance weight “at the Red Line 8 concentration threshold” when RL8 carries no number — while the v3.1 changelog states that every red line now carries a numeric trigger, which RL4, RL8 and RL12 do not.

The remaining structural gap is the one v3.0 opened and v3.1 closed only in the body: the three link-9 mechanisms have no rows in the instrument table (Appendix B), no entry for the modality × grade haircut table the body twice promises, and no red line for the scope-drift failure mode Part II names. v3.2 is a consistency and propagation release. It should not add mechanisms.


1. Grades

CriterionParts I–IIPart IIIParts IV–VPart VIBack matterOverall (v3.0 → v3.1)
OriginalityA−B+A−A−A− (A−)
AccuracyB+A−B+B+BB+ (C)
ConsistencyCBB−BC−C+ (C−)
CogencyB+B+B+B+B+ (B−)
SoundnessBB+BBBB (C)
Utility (Kardashev Labs)B−BBB+C+B− (D+)

Originality — A−

Unchanged, and v3.1 added two things that count: the deterrence reading of symmetric knowledge with the Grossman–Stiglitz answer from prover/verifier cost separation (part-ii.tex:1720-1945), and the stack-first/coin-optional investment statement (p6-kardashev-labs.tex:14-20), which is the first time the document has said what an allocator is buying in a form an allocator could sign.

Accuracy — B+

The factual spine holds and was not re-litigated where v3.0 verified it. New checks: the Tornado Cash sequence, the two 2025 SEC staff statements, the 1951 Accord, Reinhart–Sbrancia, the seventeen-year Bitcoin record, Australia 1959 and EO 6102 all hold. Arithmetic re-derived and correct throughout (§3). Residual accuracy items are small and local: gold lease rates stated as always negligible (part-ii.tex:2060; they spiked in early 2025); Fedimint guardians called “bonded” (p6-adoption.tex:181; they post no slashable bond, the same correction v3.1 made for Liquid); the stablecoin-flight claim carries no source (part-ii.tex:2193); and FCMP++ is described as “pending” (part-iv.tex:1045) where web sources conflict on whether it has activated. Nothing load-bearing.

Consistency — C+

The grade the document earns on its own standard. Appendix B opens: “a metric that appears in the body but not here, or here with a different definition, is a defect to report.” Reported: VerifySettle, VerifyReach (two incompatible definitions of succ₂), the take rate, the phase-gate durations, the fee partition, the tier multipliers, the L1 falsifier denominator, and every link-9 instrument. The front-matter summary chapters are in pre-v3.0 voice on the headline change. Two changelog claims describe states the text does not have (every instance of fee-plus-burn replaced; every red line numeric). The count of stale-after-own-correction sites is large (§2, ~25 Major), but every one is mechanical to fix and none requires an argument.

Cogency — B+

The monetary argument reads cleanly end to end. The chain table and chain-at-a-glance still locate the premium in “link 9 plus state-contingency” alone, so a reader of the summaries does not learn that the document argues three mechanisms until Part II. The flagship vignette (p1-intro.tex:203-291) denominates in USD, quotes in Work Credits and settles in XMR/ZEC; the base asset does not appear in it, which is the RL18-A configuration presented as the reference flow.

Soundness — B

The v3.0 soundness repairs are real. Open items are now honestly labelled as open rather than hidden: the usefulness–unpredictability dilemma, DVC as a generalized flow, sampling sensitivity σ, and the two “later pass owes” items (modality × grade haircut table, part-iv.tex:196; scope-drift instrument, part-ii.tex:1957). The refund_safe definition at part-iv.tex:1035 is worded so that its denominator is the set it should exclude, which is a defect in a load-bearing metric (RL2, Phase I gate) even though the intent is recoverable from the surrounding text.

Utility (for the Kardashev Labs goal) — B−

Up three grades because the chapter exists and is disciplined. It is held back by the surface: the README’s red-line section is still headed “Fifteen red lines” with “fee+burn coverage” in RL5 (README.md:84,92), PROJECT_REFERENCE.md still routes readers to /v/2.6/read/ and “15 red lines”, the outreach folder carries three documents pinned to v1.9/v2.6 counts, the Zenodo DOI still resolves to v3.0, and the banknote front door is 404 on the live site. An allocator who checks the repository against the thesis finds the repository behind it.


2. Findings

Severity: Critical = blocks peer review or breaks a core definition; Major = a referee would require the change; Minor = should fix. Line numbers exact at 5e6f242. IDs match the working notes so the remediation commit can cite them.

2.1 Front matter and Part I

IDLocationFindingSevFix
FM-3front-matter.tex:240-278 (Plain Language), 285-318 (Chain at a Glance)Both summary chapters are in pre-v3.0 voice: “the monetary bridge adds two requirements” (stress-deliverable service, holder constituency). No mention of the three mechanisms, RL16–18, or dispersion. The reading path sends allocators here first.CriticalRewrite the “monetary bridge” paragraph and the “where it is weakest” note to name pledgeability, denomination, holder-side flow and RL16–18.
FM-1front-matter.tex:316”Three bridges are explicitly Medium” vs tab:chain-strength (part-ii.tex:137-148) rating six links Medium.Major”Six links”.
FM-2front-matter.tex:221-222; bm-glossary.tex:61; bm-app-verifyprice.tex:213,223; bm-app-kpis.tex:60; p1-first-principles.tex:76refund_safe = 1.0 / “100% safe refunds” survives in the Reader’s Map, glossary, Definition 5, the KPI table and a Part I stress test after v3.1 redefined refund safety (part-iv.tex:967-980) and RL2 reads 99.5% overall.MajorOne definition (Def. 5) matching sec:l5-safety; every other site cites it.
FM-4p1-intro.tex:91; p1-stack-outline.tex:66; p1-threat-model.tex:233,257; p1-first-principles.tex:76; part-ii.tex:378,394,2678; part-iii.tex:1422; bm-app-comms.tex:31; p6-objections.tex:36”BTC↔ZEC/XMR corridors” presented as a deployed pair in ten places after v3.1 gated shielded-ZEC as non-existent (part-iv.tex:425,945).Major”BTC↔XMR (and BTC↔shielded-ZEC when admissible)”.
P5-1front-matter.tex:462 vs part-v.tex:70,220Key Definitions: Work Credits carry “governance weight: credit-weighted voting”. Layer 6: governance weight is time-locked base asset, “never Work Credits”, excluded “deliberately”.MajorDelete the governance bullet from the WC box; point to sec:l6-enforcement.
FM-6p1-intro.tex:203-291Vignette: batch “142,000 USD-equivalent in Work Credits”, settled in XMR and shielded ZEC; base asset absent; fees shown only for proof generation. This is RL18-A as the reference flow.MajorEither state the vignette exercises the stack and not the monetary claim, or route the proof fee and one settlement leg through the base asset and say which.
FM-7p1-intro.tex:304”Platforms that strip metadata cannot strip the on-chain PIDL receipt. The proof persists.” Part III (part-iii.tex:1390) already states the honest version (soft binding by perceptual hash; residual analog hole).MinorDefer to sec:l3-patterns.
P3-3front-matter.tex:198; bm-app-kpis.tex:51L1 falsifier “>20% ASNs unreachable >7 days” lacks the v3.1 “in uncensored regions” denominator (part-iii.tex:921,1476).MinorAdd the qualifier.
FM-8front-matter.tex:468; part-ii.tex:701WC “marginal cost bounded below by energy and hardware required to pass verification” — production is the expensive side.Minor”to produce work that passes verification”.
FM-10front-matter.tex:524; p1-threat-model.tex:347 vs part-iv.tex:1092”No master keys” vs “that root is a master key for its own tree”. Reconcilable.Minor”no protocol-level master key; the holder’s root is one and is never disclosed”.
FM-13p1-threat-model.tex:237Playbook counter “auto-acquire-and-burn” in burn-as-scarcity voice.Minor”auto-acquire; fee retired per routing rule”.
FM-11front-matter.tex:50”trademarks of the authors” — one author.Trivial”the author”.

2.2 Part II

IDLocationFindingSevFix
P2-1part-ii.tex:2107-2111”no red line covers it … belongs on the Value Capture Board in a later pass, and this document should not pretend it is there yet” — about the collateral–capacity loop that is Red Line 15 (p6-risk.tex:272-286).CriticalReplace with a pointer to RL15 and the collateral-architecture comparison (sec:collateral-architectures).
P2-3part-ii.tex:117-119,145,152Chain link 9 and tab:chain-strength attribute the premium to “link 9 plus state-contingency” only; §economic-linkage channels 3–4 and RL17/18 are the other two mechanisms.MajorName all three in the link-9 entry, the table row and its caption.
P4-1part-ii.tex:1296§fee-incidence-bound argues from “1.5× Silver / 2.5× Gold … 2× / 3× verification plus audit” — the v3.0 single-axis table v3.1 replaced with latency 1.0/1.3/1.6 and assurance 1.0/1.4/1.6/2.5 (part-iv.tex:336-365). Argument survives; every number is stale.MajorRestate against the assurance-tier column (Replicated 1.0 → Succinct 2.5) and the M3 redundancy cost.
P4-2part-ii.tex:286-292,318-320; part-iii.tex:309,334,387; part-iv.tex:843-851; part-v.tex:296-313,416Fee partition stated four ways: Part II band burn 30–50 / staker 20–40 / operator 20–40 “must sum to 100%”; Design A 40/30/30; Layer 4 70 prover / 20 burn / 10 “protocol treasury (security budget)”; Layer 6 70 providers / 20 burn / 10 “assurance budget”; Part III a “2% of fee revenue” Layer 0 Assurance Fund in no partition. Stakers receive 0% in the worked examples that Q4 says pay them.MajorOne canonical partition in sec:l6-monetary (with named carve-outs: assurance, treasury), every other site quoting it; fix the 2% vs 10% figure.
P5-3part-ii.tex:1275; p6-risk.tex:243,620”fee-plus-burn turnover”. Under part-v.tex:381 the burn is inside gross fees, so the phrase double-counts or means gross fees. v3.1 changelog: “every instance now reads retained-fee coverage”.Major”gross native fee turnover” (τ’s convention) at all three; correct the changelog sentence.
P2-5part-ii.tex:2089Self-quote “‘proportional to their capacity’ in §economic-linkage” — phrase no longer exists after the v3.0 rewrite.MinorDrop the third quotation.
P2-7part-ii.tex:2060Gold “lease rates are negligible and its forward curve is essentially full cost of carry” — true in normal regimes; Jan–Feb 2025 saw 1-month lease rates spike and spot backwardation on tariff-driven physical flows.Minor”in normal regimes”.
P2-8part-ii.tex:2193Stablecoin volumes “exceed most national payment systems”; stablecoins “the observed instrument of capital flight” — unsourced in a section that holds itself to base rates.MinorSource (e.g., Castle Island/Artemis 2024; Chainalysis Geography of Crypto) with the house scope note.
P2-9part-ii.tex:857”published estimates of irrecoverable bitcoin” — no key.MinorAdd (Chainalysis 2020 / Glassnode).
P2-10part-ii.tex:1667,1725,2405Parenthetical author-year citations mixed with \srcKey.MinorKeys only.
P2-11part-ii.tex:1296Tier-multiplier paragraph reads as two drafts merged (fix with P4-1).MinorRewrite.
P2-12part-ii.tex:2017-2024”Two mechanisms, and they are complements … argues both, sizes neither” — three after §denomination.Minor”three”.
P2-13part-ii.tex:1145Lemma Argument ¬2 broadens condition 2 (“burned or permanently retired”) to “retired or distributed to holders”.MinorAlign condition 2 with staker yield or narrow the argument.
P2-4part-ii.tex:1957-1965Scope drift between what is verified and what is exposed — named as “not yet covered by any red line … later pass”. Honest; still open.Openv3.2: a Value Capture Board series (off-ledger exposure disclosure ratio) or fold into RL10.

2.3 Part III

IDLocationFindingSevFix
P3-2part-iii.tex:186,201 vs part-ii.tex:800L0-A issuance cap 60%: “excess work … earns no additional Work Credits”. WC are operators’ revenue in kind; the incentive consequence (operators refuse L0-A work at the cap) is not stated.MinorState it, or pay the excess from the assurance budget.
P3-4part-iii.tex:951; bm-app-kpis.tex:47VerifyReach probes “incentivized via small WC rewards” — WC mint only against registry SKUs under the DVC bound (part-v.tex:285-290).MinorPay from the assurance budget, or register a measurement SKU.
P3-5part-iii.tex:773 vs :171,617S3 sunset keeps credits “eligible at 75%” while the compromised profile’s capacity leaves DVC; solvency bound could be breached by rule.Minor”subject to the DVC bound”.
P3-7part-iii.tex:1228L3 goal 6 “time-bounded visibility via viewing keys” — bounded forward by epoch rotation only.Minor”epoch-scoped”.

2.4 Parts IV–V

IDLocationFindingSevFix
P6-2part-iv.tex:1035refund_safe(C) = ”% of protocol-attributable failures in which every party took or could have taken a no-loss exit” — a protocol-attributable failure is by definition one where a party lacked the exit; the percentage is over the wrong set. Load-bearing for RL2 and the Phase I gate. Two thresholds (99.9% L5 SLO/admission; 99.5% RL2/adoption “overall”) use different denominators without saying so.MajorDefine over off-happy-path transactions (user-liveness class excluded); state the two denominators in one sentence in p6-adoption.tex:126.
P6-1part-v.tex:220 vs p6-risk.tex:159-163Governance weight “capped per address family at the Red Line 8 concentration threshold” — RL8 has no number. RL4 (“controlled by a single party”) and RL12 (“becomes systemic”) are also non-numeric; v3.1 changelog claims every red line is.MajorGive RL8 a concentration number (e.g., ≤ 20% of governance weight per address family, matching P5) and RL12 a forced-disclosure incidence threshold; or reword part-v:220 and the changelog.
BM-cpart-iii.tex:967-968 vs bm-app-verifyprice.tex:184-191; bm-app-kpis.tex:51Two incompatible VerifyReach specs. Layer 1: succ₁ ≥ 95% (primary transport), succ₂ ≥ 99% (any transport, ⊇ succ₁). Definition 4: succ₁ ≥ 0.98, succ₂ ≥ 0.90 with succ₂ = “multi-path reachability (resilience to single-path blocking)” — a different quantity, and one that cannot exceed succ₁. Appendix B copies Definition 4.MajorOne definition and one threshold set; Layer 1’s is the operational one.
P4-7part-iv.tex:671,752,760-763; part-v.tex:371,801; front-matter.tex:566; bm-glossary.tex:79”Physical VerifyPrice” carries two admitted senses; the document disambiguates it four times and still costs a re-read at every use.MinorRename the Layer-4 sense “Real-Resource VerifyPrice SLO” everywhere (≈8 sites).
P4-4part-iv.tex:188-197”The haircut table should be indexed on modality × grade … a later pass owes the joint table.” Open since v3.0.Openv3.2: a 3×4 table in sec:l0-ladder (Tier A/B/C × L0-A/B/C/D).
P6-5part-iv.tex:212-216; p6-risk.tex:117-125RL3’s “second reading” (M1→M3 migration) is now instrumented by Invariant 1(a) (part-v.tex:132) but Part IV still says “only the first is currently instrumented” and RL3 does not cite the M1+M2 share.MinorUpdate both.
P4-6part-iv.tex:44Sidebar “do not create money unless Part II’s value-capture conditions hold” — conditions ⇒ accrual, not money.MinorReword.
P4-8part-iv.tex:1045,1048FCMP++ “pending”; web sources conflict (a second testnet dated May 2026 vs claims of Q1 2026 activation). The text itself says the XMR-leg metric must switch to effective set size on activation.CheckVerify against getmonero.org before release; update the metric if activated.
P4-5part-iv.tex:1035,1103 vs p6-adoption.tex:19499.9% vs 99.5% (see P6-2).

2.5 Part VI

IDLocationFindingSevFix
P6-3p6-adoption.tex:139,141 vs p6-risk.tex:102,118Failure-gate table gives the RL1 and RL3 conditions verbatim with consequence “thesis weakens” / “decentralization claim fails”; the red lines say “kills”. The RL5 row already reads “Red Line 5; SoV thesis falsified”.MinorMake rows 1 and 3 match.
P6-4p6-adoption.tex:19 vs part-iv.tex:1036,1105”deployed BTC↔XMR swaps take 30–90 minutes end to end” vs VerifySettle ttf p95 target ≤ 30 min. The constitutional target sits below the deployed corridor’s stated floor.MinorTarget ≤ 90 min, or correct the 30–90 figure.
P6-6p6-adoption.tex:171,192; p6-implementation.tex:37; bm-app-verifyprice.tex:6,157MATMUL_4096 without suffix; part-iv.tex:400 says an identifier not in the registry is not a canonical workload.MinorMATMUL_4096_FP32_FREIVALDS.
P6-7p6-adoption.tex:181Fedimint “a known, bonded quorum” — guardians post no slashable bond (same correction v3.1 applied to Liquid).Minor”known federation”.
P6-8p6-kardashev-labs.tex:10”the three mechanisms of \cref{sec:state-contingency}” — they span three sections.MinorThree crefs.
P6-9p6-kardashev-labs.tex:50T₀+1 acceptance test reproduces VerifyPrice for the SNARK SKU, which has no prover.MinorScope to Freivalds SKU + SNARK verifier stub.

2.6 Back matter

IDLocationFindingSevFix
BM-jbm-glossary.tex:51Symmetric knowledge: “so every participant has already produced it and none holds an informational edge” — the exact formulation the v3.0 late refinement retracted (“everyone has already looked”) and replaced with deterrence; the Sources entry for green2026f records the retraction.CriticalCopy the deterrence wording from part-ii.tex:1730-1744.
BM-ibm-glossary.tex:39Realized take rate τ: “Native fees plus burns in USD” — RL14 (p6-risk.tex:232) pins τ to gross fees and says adding burns “would count destruction as payment”.MajorMatch RL14.
BM-hbm-app-kpis.tex:6-152Instrument table has no rows for the convenience-yield vector (RL16), haircut dispersion (RL17), free-choice denomination share / quote-settle matrix (RL18), the M1+M2 share (Invariant 1a), acceptance breadth (fork row), forced-disclosure incidence (RL12), or the Native Monetary Buyer Map fields. The appendix’s own rule calls each absence a defect.MajorAdd seven rows with unit, source, cadence, board, red line, threshold.
BM-abm-app-verifyprice.tex:213,223Definition 5: r_safe = “fraction of failed attempts where all funds return safely” and SLO ”= 1.0 over adversarial mixes” — the pre-v3.1 symmetric-refund definition.MajorRestate as sec:l5-safety defines it (see FM-2, P6-2).
BM-kbm-glossary.tex:61; bm-app-kpis.tex:96 vs p6-adoption.tex:108,229; part-v.tex:39030% fee-coverage gate: “sustained six months” (glossary, KPI table), “sustained 12–24 months” (adoption table and Phase II trigger), “Year 1 / within four quarters of viability” (coverage table).MajorOne duration; 12–24 months is the operative one.
BM-dbm-app-verifyprice.tex:32-33Views (ii) and (iii) described as “(t50,t95,e,m,f) in §l4-verifyprice” and “(t50,t95,m,b,e,f) in §telemetry-cost”; both sections now record per-run (t,e,m,b,outcome).MinorUpdate the view descriptions.
BM-fbm-app-kpis.tex:141RL10 threshold “exceeds published threshold … material share” — RL10 now says 2× frozen baseline and > 20% of capacity.MinorNumbers.
BM-mbm-glossary.tex:95; bm-app-energy.tex:44-56; part-iv.tex:1228; bm-app-sdk.tex:44; bm-app-verifyprice.tex:159”Bronze/Silver/Gold” now names three different things: latency tiers (Part IV), power-firmness tiers (App. D), and the glossary’s pre-split “latency, redundancy, interruptibility” tiers; “Gold” also names the L0 assurance tier. The PIDL receipt and SDK carry one sla_tier field where the two-axis quote needs latency + assurance.MinorGlossary entry rewritten for two axes; App. D tiers renamed (Interruptible/Curtailable/Firm) or explicitly mapped; receipt schema gains assurance_tier (or modality).
BM-nbm-sources.tex33 of 118 entries carry no trailing scope note (primary/legal/protocol sources mostly). House convention says every entry does. 19 entries defined and uncited (labelled landscape, per changelog).MinorScope notes on the 33, one line each.
BM-OKbm-sources.tex118 entries, 99 keys used, 0 undefined, 0 duplicates, every entry has a URL or DOI. Green attribution block correct: six entries, green2026f credits the observation and two objections and lists what is not drawn from it.

2.7 Outreach and repository surface

IDLocationFindingSevFix
OUT-1README.md:84,92Heading ”### Fifteen red lines”; RL5 “fee+burn coverage”.Major”Eighteen”; “retained-fee coverage”; add RL16–18 to the list.
OUT-2PROJECT_REFERENCE.md:32,34Surface table routes to /v/2.6/read/ and “Trip-wire feed (15 red lines)”.Major/v/3.1/…, 18.
OUT-3outreach/corey-petty-tldr.md:15, campaign-v1.9.md:18,86, green-conversation-prep.mdPinned to v1.9/v2.6 counts and links.MinorMark as historical or regenerate.
OUT-4RELEASE-RUNBOOK.mdTitled and paginated for v3.0 (438/303) — no v3.1 runbook was written; FOLLOWUPS.md carried the release.MinorRegenerate for v3.2 from the v3.1 commit.
OUT-5live site/nobody-examines-a-banknote/ 404 (N1b); DOI resolves to v3.0 (N2); COMPANION.substackPost empty (E2); next-gen-sov.pdf v1.0 slot byte-identical to v2.6.MajorDeploy; deposit; publish; fix the slot.

3. Verification log

ItemCheckResult
Build gatecheck.sh: 512 pp, reference integrity clean; committed PDF 3b57487b… = served = press kitHeld
Sources118 entries; 99 cited keys, 0 undefined, 0 duplicate; URL/DOI on 118/118; 33 without scope noteHeld (note BM-n)
Macro arithmetic (Part I)1.025/1.06 − 1 = −3.30%; 3 y = −9.6%; −24% → 0.904 × 0.76 = 0.687 → −31%Held
Sampling (Part III)1 − 0.95⁵⁰ = 0.923; 1 − 0.975⁵⁰ = 0.718; 3/(0.5·50) = 12%; max(50, 5·⌈0.8·12⌉) = 50; max(50, 5·32) = 160Held
𝒪ₛ rangeΣ_{i≠C} w_i = 1, components ∈ [0,1] ⇒ 𝒪ₛ ∈ [−w_C, 1]Held
Freivalds (Part IV)3/4096 = 7.3×10⁻⁴; k = 13 → 9.5×10⁻³ (inside M2); k = 20 → 1.46×10⁻² (outside)Held
MATMUL ∞-norm budgetε_r n² Ea
FP32 rounding caption√k·u = 64 · 5.96×10⁻⁸ = 3.8×10⁻⁶; k·u = 2.44×10⁻⁴Held
Reference-app coverageRLs stressed {1,2,3,5,6,7,9,10,11,12,13} = 11; unreached {4,8,14,15,16,17,18} = 7Held
Fee arithmetic (Part V)30% retained ⇒ 37.5% gross ⇒ 20% burn = 7.5% of budget; 70% issuanceHeld
tanh modulatortanh(0.02) = 0.02 → 0.2%; tanh(0.4) = 0.38 → 3.8%; ±10% at ≥ 3σHeld
Public boardsTwelve enumeratedHeld
Recycling boundaryε* = −(L−1)/L: −0.5, −0.667, −1.5, −1.333Held
Tornado CashOFAC Aug 2022; Van Loon 5th Cir. Nov 2024; delisted 21 Mar 2025; Pertsev May 2024; Storm 6 Aug 2025 §1960, hung on other countsHeld
SEC staff statementsPoW mining Mar 2025; protocol staking May 2025Held
Red-line numeric triggersRL1,2,3,5,6,7,9,10,11,13,14,15,16,17,18 numeric; RL4, RL8, RL12 notChangelog overclaims
”fee-plus-burn” instances3 remain (part-ii:1275, p6-risk:243,620)Changelog overclaims
Frozen brief/essay contentgrepped for every stale item aboveClean
FCMP++ statusWeb: conflicting (testnet May 2026 vs “activated Q1 2026”)Unresolved — check

4. Fix plan for v3.2

v3.2 is a consistency release. No threshold moves, no red line relaxes, no mechanism is added; two promised tables are supplied.

P0 — definitions and summaries (one session; mechanical)

  1. Rewrite Thesis in Plain Language and Chain at a Glance for the three mechanisms and RL16–18 (FM-3); “six links Medium” (FM-1); link-9 row and caption in tab:chain-strength (P2-3).
  2. Delete “no red line covers it” at part-ii.tex:2111; point to RL15 (P2-1).
  3. Glossary: symmetric knowledge → deterrence wording (BM-j); τ → gross fees (BM-i); phase-gate entry → 99.5% overall / 12–24 months (BM-k, FM-2); Bronze/Silver/Gold → two axes (BM-m).
  4. refund_safe: one definition in sec:l5-safety, mirrored verbatim in Definition 5, Appendix B, Reader’s Map, glossary, Part I stress test; one sentence stating the two denominators (P6-2, FM-2, BM-a).
  5. Key Definitions: remove WC governance weight (P5-1); fix “pass verification” (FM-8); “no protocol-level master key” (FM-10).
  6. Fee partition: canonical split in sec:l6-monetary with named carve-outs; Parts II, III, IV quote it; 2% vs 10% assurance resolved (P4-2).
  7. “fee-plus-burn” → “gross native fee turnover” ×3; correct the v3.1 changelog sentence (P5-3).
  8. RL8 concentration number (and RL12 incidence threshold), or reword part-v:220 and the changelog (P6-1).
  9. BTC↔ZEC/XMR → BTC↔XMR (+ gated ZEC) at ten sites (FM-4).
  10. VerifyReach: one succ₁/succ₂ definition and threshold set (BM-c); L1 falsifier denominator (P3-3).
  11. Tier multipliers in sec:fee-incidence-bound restated against the assurance table (P4-1, P2-11).
  12. README and PROJECT_REFERENCE surface (OUT-1, OUT-2).

P1 — the two promised tables and the instrument reconciliation

  1. Appendix B rows for RL16/17/18 instruments, M1+M2 share, acceptance breadth, forced-disclosure incidence, Buyer Map (BM-h); RL10 numbers (BM-f); views (ii)/(iii) (BM-d).
  2. Modality × grade haircut table in sec:l0-ladder (P4-4); delete “later pass owes” at part-iv:196.
  3. Scope-drift series on the Value Capture Board or folded into RL10 (P2-4); delete “later pass” at part-ii:1964.
  4. Vignette: state its scope or route a leg through the base asset (FM-6); defer the stripping claim to Part III (FM-7).
  5. Minor accuracy: gold lease rates (P2-7); Fedimint “bonded” (P6-7); stablecoin sources (P2-8); lost-coin source (P2-9); FCMP++ status (P4-8); corridor ttf target vs deployed timing (P6-4).
  6. Remaining minors: P2-5, P2-10, P2-12, P2-13, P3-2, P3-4, P3-5, P3-7, P4-6, P4-7 (rename), P6-3, P6-5, P6-6, P6-8, P6-9, BM-n, FM-11, FM-13.

P2 — surface and release

  1. Deploy the banknote build (N1b); Zenodo v3.1 deposit (N2); Substack companion (E2); v1.0 PDF slot.
  2. Regenerate RELEASE-RUNBOOK.md for v3.2 from the v3.1 commit; mark v1.9/v2.6 outreach documents historical (OUT-3, OUT-4).
  3. Publish this review on the site beside REVIEW-v3.0.md (N4).

5. What to keep

The Part II sequence from fee incidence to denomination is the document’s spine and should not be touched in v3.2 except at the four line-anchored sites above. The deterrence reading of symmetric knowledge, the Grossman–Stiglitz answer, the Bitcoin control case, and the “does anybody actually pay for the truth?” gloss on RL14 are the strongest new material since v2.6. The Kardashev Labs chapter is the first version of this thesis an allocator could underwrite; the surface around it is what still says otherwise.