Errata
This page logs substantive corrections, clarifications, and post-publication fixes to Next Generation Stores of Value: Privacy, Proofs, Compute.
Minor typographical changes may be folded into subsequent versions without individual notice. Substantive changes that affect interpretation, citations, figures, formulas, claims, or implementation guidance will be logged here and reflected in Updates.
Erratum E-001
Date: 2026-07-28 · Affected version: v1.2
Affected section: §23. Extended Telemetry
Issue: The introduction to the public boards states that "the original five have grown to seven," but eight boards are enumerated in the list that follows (Proof & Compute, Settlement & Privacy, Neutrality & Admission, Layer 0 Capacity, Economic Coverage, Value Capture, Wrapper Dominance, and Agency Preservation).
Correction: The count should read "eight." The enumerated list is correct; only the summary count is wrong. v1.2 is left unmodified so that the published web edition and the fixed v1.2 PDF continue to agree. The count is corrected in v1.3, which enumerates nine boards after adding the Market Realization & Wrapper Board.
Impact: Presentational only. No metric, threshold, claim, or board definition is affected.
Errata E-002 – E-006
Date: 2026-08-26 · Affected version: v1.9 (as first shipped)
Affected sections: §14. Layer 0, §19. Layer 4, §21. The Modular Stack, §22. Layer 6, §27. Risk Analysis
Issue: A systematic assessment pass (logged in Updates; shipped as v2.0) found five internal contradictions in the v1.9 text as first published: (E-002) Gold-tier sampling coverage was specified as "80% of deployed units," which no destructive lot-sampling program can satisfy and which contradicted the tier-minimum sample-rate table; (E-003) Tier A "pristine" collateral eligibility was granted to workloads whose proof components the same document labels Experimental or conjecture-grade; (E-004) the MATMUL workload carried an absolute ε = 10-5 FP32 bound that honest provers running standard BLAS kernels would fail, alongside a "transcript determinism" requirement that bit-identical FP32 reproducibility cannot meet; (E-005) the emergency governance path appeared to grant 3-of-5 signers value-affecting powers that Red Line 8 declares a thesis-ending condition; (E-006) naive demand-to-price tokenomics passages were stated as findings without staging or refutation cross-references.
Correction: All five are corrected in v2.0, which supersedes v1.9. The v1.9 artifacts are left unmodified so its citation record remains stable. Coverage is counted in production lots at tier minimum rates; a maturity gate blocks Tier A issuance for non-Production components; the FP32 bound is scaled and relative with a pinned reference kernel; the emergency path is constitutionalized as subtractive-only with routine-use escalation to Red Line 8; the tokenomics passages are explicitly staged as routing rules with the refutation cross-referenced where it lands.
Impact: Substantive but claim-neutral. No red line was relaxed and no falsification threshold was moved; the corrections resolve internal inconsistencies so the published conditions are the ones actually meant. Readers holding the v1.9 PDF should cite v2.0 for the corrected text.
Erratum E-007
Date: 2026-08-26 · Affected version: v1.9 (as first shipped)
Affected section: §24. Legal, Policy, and Jurisdictional Posture
Issue: The lawful-privacy table headed its rightmost column "Cannot Require," implying the protocol holds a boundary regulators must respect. Under the Travel Rule (FinCEN; FATF Recommendation 16; the EU Transfer of Funds Regulation), obligated entities must transmit and retain actual originator and beneficiary identity data. Predicate-based compliance is not a lawful substitute under current law in most major jurisdictions.
Correction: The column is retitled "What Current Law Actually Requires" and the surrounding text states the accurate position: predicate compliance is an engineering option that is not yet a legal option, requiring legislative change before regulated counterparties can rely on it; outside regulated perimeters the predicates are simply private; and who curates the cleared set is an unresolved hard question. Corrected in v2.0.
Impact: Substantive. The earlier framing overstated the legal position of the design; the correction narrows the claim to what the law supports.
Erratum E-008
Date: 2026-08-26 · Affected version: v1.9 (as first shipped)
Affected sections: §26. Adoption Curve, §31. Why These Could Become Money
Issue: Phase I expectations and constitutional VerifyPrice SLOs included INFER_LM_70B with a ≤30s laptop-class verification target, though no production proof system verifies a 70B-parameter forward pass on laptop hardware within seconds. The final chapter title, "Why These Become Money," asserted as settled what the chapter's own conditional text argues.
Correction: The 70B row is marked aspirational with no Sev-1 threshold until ZKML maturity lifts the gate; Phase I expects INFER_LM_7B instead. The chapter is retitled "Why These Could Become Money" and the base-rate problem is stated: no asset in monetary history is known to have crossed from utility cash-flow claim to monetary premium. Corrected in v2.0.
Impact: Interpretive. The corrections align stated expectations with engineering reality and align the title with the conditional claim the chapter defends.
Erratum E-010
Date: 2026-08-30 · Affected versions: v1.4–v2.2 (price model); v2.2 (WNG aggregation)
Affected sections: §19. Layer 4 (price model), Appendix A, §23. Extended Telemetry (WNG)
Issue: The coherence audit (shipped as v2.3) found two mathematical errors. (i) The published USD cost derivation multiplied energy in Joules directly by a USD/kWh price — a dimensional error overstating the energy term by a factor of 3.6×106, present since the disruption-adjusted VerifyPrice work and made load-bearing by v2.2, which pinned cost percentiles to the model. (ii) The v2.2 WNG aggregation spec stated the revaluation artifact backwards: it claimed an unadjusted Wrapper–Native Growth Gap rises when the asset's price falls, when the artifact in fact arises on a price rise (wrapper exposure is a stock revalued with every tick; native use accumulates at transaction prices).
Correction: Corrected in v2.3. The price model is derived per run with the explicit J→kWh conversion (e / 3.6×106 × penergy), and cost percentiles are percentiles of per-run cost, never recomputed from percentile inputs; the WNG specification unit-denominates both sides so revaluation is stripped by construction, with the USD-denominated series published alongside and the deflator distinguished from Red Line 14-B's workload unit-price index. One hundred twenty-three further coherence repairs shipped in the same version.
Impact: Substantive for instruments, neutral for claims. The energy-term error overstated a cost input by six orders of magnitude but fed only summary cost fields (no red line reads on them); the WNG direction error appeared in a rationale sentence, not the formula. No red line or falsification threshold moved. Readers citing the price model or WNG aggregation should cite v2.3.
Erratum E-009
Date: 2026-08-29 · Affected versions: v1.3–v2.1
Affected sections: §23. Extended Telemetry (VerifyFlow), Appendix H, Appendix A, §19. Layer 4
Issue: An instrument audit (shipped as v2.2) found two definitional defects in the flagship measurement families, present since their introduction. (i) The asset-level flow vector defined total exposure demand as QRC + QRD, but QRD consumed raw net creation — whose elasticity-driven component is already inside κ in QRC — so the same dollars were counted twice and the overstatement propagated into MPR and every reading built on it. (ii) VerifyPower was written as the reciprocal of the VerifyPrice vector, which has no mathematical meaning; a vector has no reciprocal.
Correction: Corrected in v2.2 and refined in v2.3 (where QRD was tightened further to consume the intercept α only, with the regression residual published as its own series and absorbed by no aggregate); undecomposed creations publish flagged; VerifyPower is now a scalar on a pinned reference workload. Eight further hardening fixes (aggregation windows, degenerate-domain rules, measurement contracts, identification discipline) shipped in v2.2.
Impact: Substantive for instruments, neutral for claims. Every defect inflated or fabricated a measurement; none moved a red line or a falsification threshold. v2.1 artifacts remain stable; readers citing the flow-vector definition should cite v2.2.