Protocol & scope
United States; the regulation, market structure, and actual-use economics of cryptocurrencies and blockchain-based financial assets — consumer protection, systemic risk, illicit finance, the stablecoin payment layer, and the question underneath them all: what verifiable, non-speculative use has emerged after fifteen-plus years. CBDC design and mining's energy footprint excluded by scope. Imports the gubment method (M1–M9) in full: two-source rule, root-tracing, pre-registered hypotheses, anchor quarantine, and a deviations log that stays visible — plus this domain's own addition, an explicit as-of date on every figure, because nothing here ages slower than the news cycle.
Executed 2026-08-04 across Phase 0 and full-execution passes: a committed federal-microdata baseline (Fed SHED 2025; the FDIC 2023 unbanked-household survey via its CPS supplement), the wash-adjustment reconciliation that resolved the kill-condition question, the FTX/Celsius/Voyager failure-mechanism autopsy at primary sources, systemic risk against the regulators' own 2023 post-mortems, illicit finance with the cash comparator stated honestly, market structure through the litigation record, the political-economy asymmetry quantified from FEC filings, five precedents with transferability assessments, an eight-architecture scorecard, and sequencing. A red-team pass moved one scorecard cell, withdrew two headline claims, and caught seven facts cited ahead of their sources — all before publication.
Anchor table — priors, stated before evidence, verified after
Twelve anchors written down as unverified guesses before research began, so they could be broken. One broke upward, one was conflating three different questions, one turned out to be unmeasured by anyone, and the kill-condition anchor resolved better than its own protocol predicted.
| # | Anchor (unverified prior) | Verified value & delta |
|---|---|---|
| 1 ★ | 15–20% of US adults have ever held crypto; regular use single digits | Confirmed at the top line (Pew: 16→19%, 2021–2026, flat) — but the anchor conflated three question types. Own committed pipelines: 8.8% held past-year (SHED 2025), 4.83% of households (FDIC/CPS 2023), transactional use 1–3% in every year measured. |
| 2 ★ | Non-speculative use under 10% of volume under any defensible wash-adjustment (KILL CONDITION) | Partial fire, then reconciled: real payments $350–550B = 0.6–0.9% of gross (two methods converge); BCG and Visa filter for the same thing by different methods, and the measurement window explains about a third of their gap (the "$21.5T annualized" figure was one hot month × 12; the live-pulled trailing year is $14.6T); only Chainalysis's $28T stays fully unreconcilable — methodology confirmed unpublished. Corrected 2026-08-10: this row said the two share "an identical bot filter" and that the gap "is a time-window artifact." BIS WP 1359 attributes the specific threshold to Visa alone and describes BCG's method as a separate three-step behavioural approach; BCG's own paper was never readable. At the growth rate this filing cites, the window carries $4.2T to about $7.4T, not $14.6T — which is ($7.4 − $4.2) ÷ ($14.6 − $4.2) ≈ 31% of the gap, leaving roughly two-thirds unexplained. Corrected again 2026-08-10: this row first said the window explained "about half," an arithmetic error inside the correction itself. |
| 3 ★ | FTX repaid customers at or near 100% of petition-date value | Confirmed and needs the two-number statement: 103–120% of Nov-2022 dollar value, in cash (primary-fetched) — which is ~27–32% of in-kind value after BTC's ~3.7× rise. Second finding: no federal instrument separates exchange-failure from price-decline losses. Kill condition 2 fired. |
| 4 ★ | Stablecoin issuers hold ~$100–150B in Treasuries — "mid-sized sovereign" | Broken upward: ~$175–190B combined (Tether ~$115B, Jun 2026 attestation; Circle ~$66B), ranking ~15th among foreign holders, near Norway. Tether has never completed a full audit; GENIUS disclosure rules not yet in force. |
| 5 | Illicit share low single digits per Chainalysis; academics higher | Directionally right, incomplete: ~1–1.2% (Chainalysis/TRM — shared incentives, not independent), revised sharply upward every year by the vendor's own history; 2025's growth is state actors (+694% sanctioned-entity flows). No clean cash comparator exists — the closest (UNODC) measures a different thing. |
| 6 ★ | 2024 election spending ~$130–200M via super-PACs | $133.0M in independent expenditures, summed directly from FEC filings — the "$200M+" figures are double-counted fundraising. Second-largest single-sector PAC operation on record, in one cycle. Corrected 2026-08-10: this row said "2026 already exceeds it." On independent expenditures it does not — $42.7M across two of the three committees through 2026-06-30, roughly $77M crediting the third at its 2024 level, against 2024's $133.0M. 2026 receipts ($166M across two committees) do exceed it. The $133,006,906.56 sum was re-verified to the cent against the FEC's own data. |
| 7 | MiCA fully in force since end-2024; evidence thin | Precision fix: legally effective 2024, but national grandfathering only completed EU-wide 2026-07-01. Issuer-relocation evidence now rich (Tether exited; Circle complied); consumer-outcome statistics in circulation trace to no regulator — flagged unverifiable. |
| 8 ★ | A US stablecoin statute was enacted 2025–26 — verify scope | Enacted 2025-07-18, not operative: all five agencies missed the one-year rulemaking deadline. Excludes algorithmic stablecoins (the Terra class) and does not touch token taxonomy — the CLARITY Act that would remains stalled without a Senate floor vote. |
| 9 | El Salvador: minimal sustained adoption; rolled back for the IMF | Confirmed and sharpened: 25.7%→8.1% usage 2021–2024 (independent surveys); legal tender ended 2025-01-29 as an IMF prior action. Live addition: the IMF's own review contradicts the government's "still buying" claims. |
| 10 | Japan's custody regime produced materially lower loss rates | Directionally consistent, honestly underpowered: n=2 post-reform failures, both ~100% recovery via ad hoc parent backstops (one FTX-funded). Cleanest data point: FTX Japan repaid customers ~2 years before FTX global. Mechanism: loss-reallocation, not breach prevention. |
| 11 | A majority of US-person volume runs through offshore venues | Unmeasured — by anyone. No regulator or researcher has isolated the ratio; the CFTC's Binance findings measure the wrong denominator. Plausible, empirically unestablished, and load-bearing for every enforcement argument. The gap is the finding. |
| 12 | SEC enforcement fell sharply after the 2025 transition | Confirmed exactly: 33→13 cases (−60%), penalties $4.98B→$142M (<3%). Framing corrected: the CFTC retreated in parallel, not in offset — "the ratchet turns both ways" is about time, not agencies. |
Workstream findings
§2 · BaselineFederal microdata, tabulated directly — and the question types untangled
Committed stdlib pipelines for Fed SHED 2025 (n=12,934; parse verified cell-exact against the codebook) and the FDIC 2023 survey via its CPS supplement (the FDIC hosts no microdata; the survey rides the June 2023 CPS). Neither instrument asks "ever held" — the circulating 15–20% figure is a different question, answered by Pew alone.
§3 · Actual useThe wash-adjustment divergence, reconciled — and one number that can't be
Three vendors' "adjusted volume" figures decomposed: BCG and Visa filter for the same thing by different published methods, and roughly a third of their gap is a measurement-window artifact against rapid growth — including a correction to this project's own Phase-0 figure, caught by pulling the live dashboard. Chainalysis's $28T is unverifiable (methodology gated) and its estimate is absent from the BIS literature, though the vendor is cited there as a data source. Two independent methods converge on $350–550B of real payments. Corrected 2026-08-10 — see anchor 2 above and verification-log.md F3, F5.
§4 · Consumer protectionAll three collapses were bank failures in crypto clothing
FTX (a written segregation promise defeated by a secret negative-balance flag), Celsius (Custody accounts commingled against their own marketing; new deposits paying old withdrawals), Voyager (one uncollateralized loan, concealed default) — traced to examiner reports and federal complaints. Zero crypto-native mechanisms. The fix the record supports: segregation plus mandatory independent audit, because rules on paper already existed and were violated in secret.
§5 · Systemic riskThe channel is old-fashioned; the asset is new; the scale is not there yet
Regulators' own 2023 post-mortems: crypto was proximate cause only at Silvergate; Signature and SVB died of uninsured-deposit concentration with crypto as reputational amplifier. Stablecoin Treasuries ≈ 1% of bank deposits today; the Fed's own modeling puts high-adoption displacement at $600B–$1.26T. SAB 121 rescinded; Basel's crypto standard unimplemented in US capital rules.
§6 · Illicit financeLow share, state-actor-driven growth, and a receding ransomware economy
~1–1.2% illicit share (flagged single-lineage); the 2025 surge is sanctions evasion ($104B, +694%), not street crime; ransomware payment rates collapsed 79%→28% (the workstream's one genuinely two-vendor figure). Cash comparator finally given teeth: $90B of Russian oil moved through ordinary shell companies — the crypto rail's equal.
§7 · Market structureFive years of litigation bought a judicial split, not an answer
Ripple final only by mutual abandonment; Coinbase dismissed expressly without merits; Terraform adjudicated — and its judge rejected Ripple's core distinction. Zero appellate resolution. H7.1 confirmed: the industry's "clarity" ask (jurisdiction) is not the fix the documented losses point to (verification). Corrected 2026-08-10: this said CLARITY's custody title is "eligibility-only, with no audit-cadence mandate found." The bill text was fetched for the first time on 2026-08-10 (govinfo, BILLS-119hr3633eh). It contains no proof-of-reserves clause, no customer-asset reconciliation and no concentration limit — all three verified by literal search, zero hits each — but it does require qualified custodians to submit audited financial statements, and the GENIUS Act requires monthly examination of stablecoin reserves by a registered public accounting firm. The gap is one of perimeter, not absence.
§8 · Political economyA 25–45× asymmetry and a constituency that doesn't exist
$133.0M in industry independent expenditures (summed from FEC primary filings) against consumer-advocacy budgets 25–45× smaller in total. No organized victim constituency — no PAC, no lobbyist, no endorsements; victims' stories entered the Senate record via the industry's own allies. Fairshake spent $13.5M against Democrats, $0 against Republicans.
§9 · PrecedentsFive cases, one procedural clock, one hard boundary
1933/34 disclosure transfers to anything with an issuer and stops cold at DeFi; the state money-transmitter patchwork survives GENIUS for everyone but stablecoin issuers; the Dodd-Frank swaps buildout took 3 years (core) to 10 (full) against CLARITY's 270–360-day deadlines; China's ban lost to VPNs with tools no democracy has; Japan, MiCA, and El Salvador covered at anchor depth.
§10 · ScorecardA leader by absence of failure; a tie that indicts the evidence base
Eight architectures × five anchored objectives × six weightings, five sensitivity readings published. Custody-with-audit-teeth leads five of the six weightings — as the only row without a documented failure — and the do-nothing comparator takes the sixth. Do-nothing ties every marquee instrument. Restriction is the only architecture actively scored down, and it finishes last under every weighting in every reading. 25 of 40 cells at 3; the flat board is the finding. (Corrected 2026-08-10: this said 26 — a count true of an earlier version of the board and never recomputed after the 2026-08-06 adjudication moved four cells. The rankings themselves were independently recomputed and are correct as published.) (Corrected again 2026-08-10, Phase 2 steelman: this said custody "leads everywhere." It does not — the scorecard's own W5 column has the do-nothing comparator ahead, 3.25 to 3.15, and has said so in print since the 2026-08-06 adjudication raised that row's enforceability score. Worse, the lead is one cell wide: read 4A the way the red team read it and custody leads no weighting outright. That downside sensitivity had never been published. It is now, along with a fourth reading in which the CLARITY row's consumer-protection cell is credited for the bill's express anti-commingling duty — which changes no ranking, and moves that row from below the do-nothing comparator to level with it.)
§11 · SequencingCustody first (relabeled honestly), measurement before extraterritoriality
Custody/audit attaches to whichever vehicle moves — a strategy judgment, labeled; GENIUS reclassified from to-do to implementation-and-perimeter problem; taxonomy second at swaps-buildout timelines; the offshore share gets measured before anyone builds architecture on it; statute over guidance, because 60% of enforcement evaporated in one administration change.
Deviations log
Forty-five entries — subagent delegation with QA cross-checks, primary-source fetch failures logged rather than smoothed over, one AI-search hallucination caught before citation, this project's own Phase-0 figure corrected by its own deeper pass, the red-team dispositions, and the two parallel blind re-scores and their adjudication. (Corrected 2026-08-10: both public pages said "thirty-seven"; the committed log had run to 39 by the 2026-08-06 adjudication and to 45 with this pass.) The full table lives in the repository; highlights:
| # | Deviation | Effect |
|---|---|---|
| 16 | Phase 0's "Visa ~$21.5T annualized" figure turned out to be a one-month run-rate extrapolation circulating in press | Caught by this filing's own §3 pass pulling the live dashboard: actual trailing-12-month figure $14.6T. Anchor corrected in the open, per the anchor-quarantine discipline — one level deeper than usual: a sourced figure, not just a prior, was the artifact |
| 19 | An AI-generated search summary attributed specific category dollar figures to BIS WP 1359 that appear nowhere in the paper | Caught by the §3 agent re-reading the primary PDF before citing — flagged as a likely search-engine hallucination and excluded; the housing filing's Katerra catch, recurring in a new medium |
| 5, 15, 21 | IC3 report PDFs, the CFTC Binance complaint, and both bankruptcy examiner reports resisted machine text-extraction | Every affected figure carries its secondary-sourced provenance explicitly; none was silently promoted to primary-verified |
| 25, 32 | DPRK/sanctions dollar figures and all China-activity figures ride a single commercial lineage (Chainalysis) | Flagged per-figure; the one genuinely two-vendor-corroborated number in §6 (ransomware payment-rate collapse) is identified as such |
| 36 | Red team: one cell moved, two headlines withdrawn, one composite-instrument disclosure forced | The same failure families housing's red team caught recurred despite the authoring pass explicitly importing housing's rules — evidence the check must stay structural, not remembered |
| 37 | Seven facts used by the scorecard had not been landed in the committed record | All were established in this filing's own verification passes; backfilled to their proper findings files with sources and marked, rather than struck or silently kept |
Red team
Fifteen attacks against the committed §10/§11 draft: four sustained, six partially sustained, five rejected — with the rankings arithmetic independently recomputed and verified. The full log lives in the repository; the ones that drew blood:
The leading architecture's score mischaracterized its own evidence
Sustained — cell corrected. The draft's A=4 called Japan's recoveries "the mandated mechanism"; the record says ad hoc parent bailouts, one of them funded by FTX itself. Corrected to A=3; custody still led on most weightings, but as the only row without a documented failure, not as a proven fix. Update: the 2026-08-06 adjudication between the two blind re-scores reversed this downgrade and restored A=4 — the red team's reasoning rested on the regime reallocating rather than preventing loss, while the axis anchor reads "prevent or reallocate." The published board is v4, with A=4. Phase 2 update, 2026-08-10: this cell is the whole of the leader's margin, and the board carried no published sensitivity for the red team's own reading of it. Computed now: at A=3 custody ties the GENIUS-shape row across five weightings and loses the sixth, leading nothing outright. Both readings are published in the scorecard. Two passes have now scored this cell 4 and one has scored it 3; the honest description is contested, not settled.
The "Terra sits outside every architecture" headline
Sustained — headline withdrawn. The SEC won the Terraform case; Terra had a compellable issuer and disclosure law reached it (post hoc, at ~$4.47B against a much larger loss). The unowned territory is narrower and scarier: the no-issuer autonomous protocol, which nothing reaches even after the fact — and which hasn't had its mega-loss yet.
Seven facts cited ahead of their sources
Sustained — all backfilled. Every one had been verified in this filing's own research passes but never landed in the committed findings files. Each is now in the record with its source and a backfill marker — the every-cell-cited rule enforced by adversarial grep.
The broadened enforceability axis stacked the deck for the filing's own thesis
Partially sustained. The narrow-reading sensitivity was computed and published: the board's middle tiers merge into a five-way tie, enforcement-only becomes numerically identical to doing nothing, and the top doesn't change. One perverse artifact named in print: under the broad axis, the null instrument outranks the only architecture with a demonstrated $4.3B offshore action.
The recommended instrument is a composite with no as-implemented exemplar
Sustained — disclosed in both documents. The segregation half is evidenced; the audit-teeth half — the part the filing calls load-bearing — has never been implemented anywhere. The recommendation stands as the best-reasoned mechanism chain in the filing, labeled as derived rather than demonstrated.
Withdrawn 2026-08-10, Phase 2 steelman. "Never been implemented anywhere" is false. The futures commission merchant customer-funds regime is the exemplar: 7 U.S.C. §6d(a)(2) has required customer assets to be "separately accounted for" and un-commingled since 1936, and 78 FR 68506 (Nov. 14, 2013, effective 2014-01-13) added the verification half — 17 CFR §1.20(d) permits customer funds only at a depository that gives the CFTC "direct, read-only electronic access to transaction and account balance information" and sends the acknowledgment letter to the Commission and the SRO "without further notice to or consent from the futures commission merchant." It was adopted after MF Global's reported $900 million shortfall and after Peregrine Financial Group "falsely represented that it held in excess of $220 million of customer funds when in fact it held approximately $5.1 million" — concealed in the reports PFGI filed with the Commission, underneath a segregation duty added to the statute in 1936. (The release does not state how long it ran; no duration is claimed here.) That is this filing's Part 4 thesis demonstrated rather than derived, in the agency and the self-regulatory body the CLARITY Act would put in charge. Limit: no post-2014 outcome study was located, so this is design evidence, not demonstrated effect — no cell was raised on it. Both documents are archived in method/sources/.