Status: Phase 0 executed 2026-08-03 — findings, deviations. Verdict GO; priors broken or corrected on anchors 1, 2, 3, 5, 7. All 11 workstreams (§2–§11) executed as first passes 2026-08-03 — see drugs/research/ws0N-findings.md for each. Red team applied; four confidence-calibration amendments logged, no conclusions reversed. Whitepaper published: www.gubment.com/drugs (Whitepaper No. 8), live on the site index. An independent second scorer for §10's scorecard remains a standing requirement before the ranking is presented as fully settled — stated in the whitepaper's own honesty box, matching GBMT-1's precedent. Headline finding: contingency-management safe-harbor expansion (A3) is the filing's clearest first-move recommendation — rulemaking-executable (OIG authority already in active use), near-zero cost, and top-ranked or tied-top under 3 of 4 objective weightings. Superseded 2026-08-10 by both verification passes. Phase 1 (verification-log.md) found the anti-kickback mechanism misdescribed and the rulemaking not underway (RIN 0936-AA13, NPRM target July 2027). Phase 2 (steelman-log.md) found the executable path is state Medicaid §1115 coverage — five states approved, California delivering since March 2023 — not a federal safe harbor; that "near-zero cost" prices the rule change rather than the staffed service; that the "3 of 4 weightings" claim is not reproducible because no weight vector was ever committed; and that A3 holds first place in only one of three defensible readings of its own board. Current headline: contingency management is worth doing and is already being done under existing law by states — the filing named the wrong lever, the wrong venue and the wrong clock, and its own scorecard no longer returns a single first move. Scope: United States; psychoactive drug use and its governance — overdose mortality, addiction and the treatment system, enforcement and criminal justice, and legal-market design. Alcohol and tobacco are in scope as the incumbent legal drugs and the standing comparators; excluding them is itself a policy choice the analysis must not silently repeat. Method: Imports method/gubment-method.md (M1–M9) in full. This document contains only domain content. Artifact: A feasibility assessment of the leading "fix drugs" theories — enforcement, treatment expansion, harm reduction, decriminalization, legalization — naming binding constraints in priority order and scoring candidate policy architectures. Whitepaper No. 8. Site framing already shipped: "Overdoses, treatment you can't get into, and what legalizing actually changed."
1. Definitional decisions to settle before fieldwork
- Which drug problem? At least four travel under "the drug crisis," with different populations and remedies: (a) overdose mortality — concentrated, fentanyl-era, and recently falling; (b) addiction — SUD prevalence and the treatment gap; (c) enforcement harms — arrests, incarceration, records, and their distribution; (d) legal-market design — cannabis now, alcohol and tobacco always. Conflating them recommends the wrong instrument somewhere. Treat as linked but distinct systems, the way GBMT-2 treated market types.
- Which drugs? Opioids/fentanyl, stimulants (meth, cocaine — the polysubstance half of the death toll nobody's plan addresses), cannabis, and the legal pair. "Drugs" as a single category is mostly an artifact of the Controlled Substances Act, not of pharmacology or of harm distributions. Decide per workstream and say so.
- What counts as "treatment." MOUD (methadone/buprenorphine — the evidence-backed core for opioids), abstinence/12-step, residential rehab (a large, lightly regulated industry with mostly unmeasured outcomes), and contingency management (for stimulants). Treatment-gap statistics conflate these; the gap number depends entirely on which definition and whose assessment of "need."
- "Worked" for whom, on what. Use prevalence, death, disease transmission, crime, neighborhood disorder, and liberty move independently and sometimes oppositely (a policy can cut deaths while raising visible use). Pick the measured quantity per claim; never let "worked" float.
- Objective function (per M6 and the GBMT-1 review lesson): deaths averted, use/addiction prevalence, criminal-justice footprint, public order, and individual liberty rank architectures differently — a mortality-first weighting favors harm reduction and MOUD; an order-first weighting favors enforcement; a liberty-first weighting favors legalization. Score under explicit weightings; report rank stability. Where the ranking flips is the finding.
2. Workstreams
§2 Baseline — use prevalence and SUD prevalence (NSDUH, with its known break-in-series and coverage problems named), overdose deaths (CDC WONDER final + provisional, by drug class, age, geography), treatment capacity and utilization (N-SSATS/N-SUMHSS, SAMHSA), spending inventory: federal drug control budget by function (supply vs. demand side), Medicaid as payer, state/local enforcement spending (likely unmeasured — the H2.3 rhyme), and the alcohol/tobacco comparator columns. Seed hypotheses: H2.1 the "~1 in 10 with SUD get treatment" gap number is definitionally underdetermined — it moves by multiples across need-definitions, and self-perceived need is a small fraction of imputed need; H2.2 no instrument measures treatment outcomes at national scale — capacity and admissions exist, results don't; H2.3 state/local enforcement spending on drugs has no unified accounting, making every "enforcement vs. treatment spending" claim a partial count.
§3 The overdose system — the fentanyl supply transition and its geography (why the West died later), polysubstance mortality (stimulant+opioid now the modal death), naloxone access and OTC status, drug checking, and the 2023–2025 national overdose-death decline — the domain's biggest live phenomenon, cause genuinely contested (supply-side changes, naloxone saturation, susceptible-population depletion, treatment expansion). Seed: H3.1 the recent decline is not attributable to any single policy instrument, and every faction is currently claiming it; H3.2 the decline's timing and geography fit supply-side explanations at least as well as any demand-side program's rollout.
§4 Treatment economics & the delivery system — the workforce question again (addiction counselors and the pay floor — GBMT-1's educator analog), methadone's regulatory cage (OTP-only dispensing, DEA/SAMHSA dual regulation, the 2024 42 CFR Part 8 liberalization), buprenorphine after X-waiver repeal (the 2023 natural experiment: did removing the legal barrier move prescribing, or does prescriber willingness bind?), the rehab industry's quality/fraud problem (Florida shuffle, body brokering), insurance parity enforcement (MHPAEA), and contingency management's anti-kickback ceiling. Seed: H4.1 methadone regulation is a binding legal constraint without medical justification — comparator countries dispense via pharmacy at population scale; H4.2 X-waiver repeal moved prescribing only modestly → the binding constraint on MOUD is provider willingness and payment, not statute; H4.3 the interventions with the strongest evidence (MOUD, contingency management) are the ones most constrained by law, while the intervention with the weakest outcome evidence (unregulated residential rehab) is the one the market freely supplies — the filing's likely headline shape.
§5 Enforcement & criminal justice — arrest volumes and composition (possession vs. sale), the actual share of incarceration attributable to drug offenses (state vs. federal — the discourse-vs-data anchor), supply-side interdiction against the price/purity record (five decades of falling real prices), the iron law of prohibition (enforcement pressure selects for potency — fentanyl as endogenous outcome), and the steelman: what enforcement does measurably buy (market disruption effects, drug-market violence suppression, leverage into treatment via courts). Per M3, this workstream owes the enforcement case equal effort, built from its best evidence (focused deterrence, drug courts' RCT record), not its worst. Seed: H5.1 marginal supply-side spending has no detectable effect on retail price or availability at current margins; H5.2 "mass incarceration is a drug-war artifact" is largely false for state prisons — the population share is low-double-digit percent and falling; H5.3 (steelman) swift-certain-fair supervision models (HOPE-style) and focused deterrence outperform both mass enforcement and non-enforcement in their trial record.
§6 Federalism & legal mechanics — the CSA and scheduling (including the cannabis Schedule III rescheduling's actual legal effect — 280E relief, not legalization), the state-federal cannabis contradiction as a standing federalism experiment (banking, tax, interstate commerce), methadone/OTP statutes vs. regulation (what SAMHSA can change alone vs. what needs Congress), the Medicaid IMD exclusion, anti-kickback law as the contingency-management ceiling (what OIG safe-harbor discretion has and hasn't done), and Byrd analysis for any federal instrument. Seed: H6.1 unlike housing, the binding rules here are mostly federal (CSA, DEA, IMD, anti-kickback) — the inverse of GBMT-2's finding, making Congress and agency rulemaking the load-bearing venue; H6.2 a material fraction of the treatment agenda is executable by rulemaking alone, no statute required.
§7 Market actors & the money — opioid settlement funds (~$50B+ over ~18 years): governance, reporting quality, and whether disbursement follows evidence; the legal-cannabis industry's economics (price collapse, tax-driven illicit persistence, CA vs. WA as design contrast); private equity in treatment and the rehab industry's unit economics; pharma conduct on naloxone and buprenorphine pricing; alcohol industry as the incumbent-lobby template any legalization design inherits. Seed: H7.1 settlement funds are being spent with weaker evidence requirements than the litigation's own theory of harm demanded; H7.2 legal cannabis markets under-displace illicit supply where tax+regulatory cost exceeds the illicit price premium — legalization is a market-design problem, not a switch.
§8 Civil society & coalitions — the fault-line map: harm-reduction orgs vs. the recovery/abstinence community vs. law-enforcement associations vs. parent/prevention groups; the post-2020 realignment (Oregon backlash, SF discourse shift); where each camp's data comes from. The GBMT-1/2 lesson: name each actor's funder and whether their number is data or advocacy.
§9 Precedents — the most citogenesis-prone workstream in the whole gubment project; every case below is routinely mangled in public discourse and gets the root-tracing treatment. Domestic: Oregon Measure 110 (passed 2020, recriminalized 2024 — separate the policy's measured effect from the fentanyl-arrival confound; the synthetic-control literature exists), cannabis legalization states a decade in (CO/WA: use, harms, revenue, illicit share, youth use), drug courts (large RCT/meta record), alcohol Prohibition (the canonical case, usually mis-summarized in both directions), tobacco control (the demand-side success story — 50% smoking decline with no possession arrests — conspicuously absent from drug-policy debates). International: Portugal 2001 (the most-cited, least-read precedent — what decriminalization actually changed vs. what the treatment investment did, and what the post-2010 austerity period shows), France 1995 buprenorphine liberalization (office-based prescribing, ~80% overdose-death decline — the strongest MOUD natural experiment), Switzerland heroin-assisted treatment + four-pillars (Zurich), British Columbia decrim + safe supply 2023–24 (the live partial-reversal case, the cautionary pair to Portugal), Netherlands coffee-shop separation, Sweden's enforcement-forward model (the restrictionist steelman with its own contested mortality record). Each with transferability assessment per the GBMT-1 template.
§10 Candidate architectures (seed list, to be scored per M6): methadone deregulation (pharmacy dispensing, office-based prescribing — the France/Australia import); MOUD-everywhere (ED-initiation, jail/prison MOUD mandates, telehealth permanence); contingency-management legalization (safe-harbor expansion); treatment-industry quality regime (outcome measurement + conditional funding — the "rehab that reports results" architecture); harm-reduction scale-up (naloxone saturation, drug checking, safe consumption sites at the measured-evidence scale, not the advocacy scale); Portugal-style decriminalization with dissuasion apparatus (vs. Oregon-style without — the design difference is the finding); cannabis federal designs (descheduling + state opt-in vs. Schedule III status quo); supply-side modernization (precursor diplomacy, parcel interdiction — scored on its actual record, not its budget); prevention/tobacco-playbook transplant (taxation, age-gating, marketing limits applied to legal drugs including alcohol); enforcement steelman package (focused deterrence + swift-certain-fair supervision); and the comparator row per the GBMT-1 review lesson — status quo drift (no new policy; current trend continuation) — because several trend lines are currently improving on their own and every architecture must beat the counterfactual, not zero. Seed: H10.1 the strongest-evidence architectures are legal-constraint-bound, not funding-bound — federal rulemaking and safe-harbor discretion are the cheap, fast levers (the rhyme with GBMT-1's Byrd finding and GBMT-2's legal-capacity finding).
§11 Sequencing — rulemaking-executable items first (they're cheap and test H6.2), settlement-fund governance second (the money is already appropriated and moving), statutory architectures last; the cannabis track runs parallel (different system, different politics). To be tested, not assumed.
3. Anchor Table (seed — verify per M4; ★ = Phase 0)
All values below are unverified priors stated from memory, recorded here precisely so they can be broken. Nothing in this table may be cited until the "Verified" column is filled.
| # | Anchor (unverified prior) | Verify in | Verified value & source | Delta |
|---|---|---|---|---|
| 1 ★ | US overdose deaths peaked ~110–112k/yr (2022–23), then fell ~25% through 2024–25 — the largest decline on record | §3 | Peak was higher and later; decline is bigger. True rolling 12-month peak ≈112–114k, in 2023 (not "2022-23" as a pair — 2022 alone was not the peak year; CY2022 final = 107,941, CDC/NCHS Data Brief 491). CY2024 = 80,391–81,313 (−27% YoY); CY2025 = 69,973 (−14% YoY) — cumulative peak→2025 decline ≈37–38%, not ~25% (CDC NCHS press releases 2025-05-14, 2026-05-13). "Largest on record" confirmed, independently, via peer-reviewed analysis (Post et al., JAMA Network Open 2025) describing it as the longest sustained decline in 40+ years — the only prior comparable decline (2018) was −4 to −5% YoY | Both peak timing/magnitude and decline magnitude were understated. A June 2025 apparent "spike" was later shown to be a CDC nowcasting-model artifact, corrected Aug 2025 — CONFIRMED at primary tier 2026-08-10 (verification-log.md, headline finding 3). Citation corrected: Post et al., Am J Public Health 2026;116(5):591–593, not AJPH 2025. Correction to a correction: an earlier verdict this same day withdrew this claim, reasoning from the title alone ("The 2025 Drug Overdose Spike That Wasn't: Neither Politics nor Data Errors Explain the Anomaly") plus a wrong assumption that the body was paywalled. The body is open access at PMC13066679 and says "Instead, the anomaly was a model artifact"; that a "second revision released in August 2025" clarified "that the January 2025 'spike' was an artifact"; and that "the anomaly resulted from applying growth-era algorithms to a period of decline." The title rules out political manipulation and data error as causes, leaving the model artifact. The withdrawal is reversed; see deviation #22. The caution stands too: real-time provisional data produced one nationally publicized false reversal. Peak/decline figures independently re-derived this pass from CDC VSRR provisional counts (data.cdc.gov xkb8-kh2a): 12-month-ending peak 111,466 (June 2023); 12 months ending Dec 2025 69,939 (−37.3%); 12 months ending Feb 2026 67,531 (−39.4% — the as-of-check-date figure) |
| 2 ★ | "Only ~1 in 10 Americans with SUD receive treatment" — and self-perceived need for treatment is a small fraction (single-digit %) of survey-imputed need | §2 | Treatment-receipt half of the claim is stale; perceived-need half replicates. 2024 NSDUH: 48.4M (16.8%) with past-year SUD; receipt = 19.3% of the "needing treatment" (SUD-or-treated) denominator, or 12.4% of the SUD-only denominator (SAMHSA 2024 NSDUH Highlights). The "1 in 10" figure traces to a single root — the 2015 NSDUH specialty-facility-only, DSM-IV, single-mode-collection figure (10.8%), relayed via the 2016 Surgeon General report — which SAMHSA itself says is not comparable to 2021+ data (DSM-5 transition + 2020 multimode redesign, both documented methodology breaks). Perceived need: 3.3–6.6% across 2015/2023/2024, independently corroborated by a peer-reviewed reanalysis (PMC12026904) — this part of the anchor is robust and not an artifact of definition | Prior partly broken. The exact "1 in 10" ratio is a decade-stale, definition-narrow figure being cited as current; under any 2023/24 definition the rate is 12–24%, not ~10%. The perceived-need claim, the more surprising half, holds up strongly |
| 3 ★ | Federal drug control budget ≈ $40–45B/yr, split roughly half supply-side/half demand-side — historically ~2:1 supply-side; state/local enforcement spending has no unified accounting | §2 | Total and historical-ratio claims confirmed; current split corrected. FY2023–25 total = **44.2–44.5B * *([ONDCPFY2025FundingHighlights](https : //bidenwhitehouse.archives.gov/wp − content/uploads/2024/03/FY − 2025 − Budget − Highlights.pdf)), split * * ≈ 55–5615.9B), published 1993, never repeated | Current split corrected from ~50/50 to ~55/45 demand-majority; the "2012 accounting artifact" hypothesis is only partly right — it explains the total's near-doubling, not the later crossover to demand-majority, which reflects real opioid-crisis-era appropriations growth |
| 4 | Drug offenses account for ~13–15% of state prisoners but ~45% of federal prisoners; state systems hold ~7× the federal population | §5 | Blank — not attempted in Phase 0; queued for §5 execution | |
| 5 ★ | Portugal 2001: drug deaths and HIV transmission fell post-decriminalization; the causal attribution to decriminalization (vs. the simultaneous treatment expansion) is weaker than cited | §9 | Confirmed, and the citogenesis is worse than the seed anchor implied. HIV decline is the one strongly corroborated win (new IDU-attributed diagnoses ~1,287→16, 2001–2019, Transform Drugs 2021, cross-verified against EMCDDA-derived figures). Deaths are genuinely treacherous: three sources give three incompatible 2001 baselines (131 / 76 / ~400) from incompatible measurement systems (INML toxicology vs. INE cause-of-death), and Portugal later doubled its autopsy rate, muddying any 2010s "deaths rising" read. Both "use doubled" (lifetime prevalence) and "use fell" (past-year/youth) are true, from the same underlying survey, spun in opposite directions by advocates on both sides. Greenwald/Cato 2009 — the root of most US "Portugal miracle" claims — is confirmed as a Marijuana Policy Project–funded advocacy report. Hughes & Stevens' own papers are explicitly framed as a corrective to both the "success" and "failure" readings; attribution to decriminalization vs. the simultaneous treatment-investment expansion "cannot be firmly established" | Confirmed as stated, with the death-data mess and the advocacy-root identification adding precision the seed anchor didn't have. Recent (2022–24) Portuguese data could not be verified — flag as an open gap, not settled either direction |
| 6 | Oregon M110: overdose deaths rose sharply post-2021, but synthetic-control estimates attribute little-to-none of the rise to the policy vs. fentanyl arrival timing | §9 | Largely confirmed, with a genuine (not advocacy-driven) scholarly dissent. Three independent peer-reviewed studies (Joshi et al., JAMA Psychiatry 2023; Zoorob et al., JAMA Network Open 2024; an April 2026 changepoint study) find little-to-no detectable M110 effect on overdose deaths and attribute the rise to regional fentanyl-arrival timing (Oregon's mortality inflection point precedes M110's passage by ~11 months and its implementation by 14). One peer-reviewed dissent (Spencer, J. Health Econ. 2023) estimates 181–182 additional 2021 deaths (~23%) — but Spencer's own paper attributes this partly to decriminalization occurring without the promised treatment-capacity buildout, which (per §9 findings) didn't start disbursing funds until June 2022, over a year after decriminalization took effect | Confirmed as stated. Unlike Portugal, this is a genuine multi-study peer-reviewed dispute, not a single-advocacy-root problem — see ws09-oregon-findings.md |
| 7 ★ | France 1995: office-based buprenorphine cut overdose deaths ~79–80% within four years | §9 | Single-root and imprecisely sourced — do not cite the magnitude without this caveat. The "79%, 1995–1999" figure traces to one paper (Auriacombe et al. 2004), whose underlying mortality dataset could not be independently verified (paywalled). A second, incompatible figure ("80% between 1994–2002") also circulates and gets conflated with it. No independent dataset re-derives either figure. The qualitative direction (France's opioid mortality is now dramatically lower than the US's) is well corroborated; the precise magnitude and its root computation are not. Durability and transferability are both actively contested: French GPs initiating buprenorphine fell 43.8% 2009–2015 (the documented "end of the French Model"), and French national surveillance (DRAMES, 2011–2021) shows methadone — not buprenorphine — is now the leading cause of French substitution-drug deaths. A 2023 peer-reviewed paper (Sud et al.) directly critiques the naive US-transferability extrapolation (Weimer et al.'s "37,000 lives saved" claim) as ignoring the 22.5× scale gap and fentanyl-vs-heroin supply differences | Prior needs a confidence downgrade, not a number change. Treat as a plausible but single-root figure; the durability reversal (2009–2015) and the methadone-mortality finding are new information the seed anchor lacked entirely |
| 8 | Youth cannabis use did not rise in legalization states (NSDUH/YRBS/Monitoring the Future) | §9 | Blank — not attempted in Phase 0; queued for full §9 execution | |
| 9 | Contingency management is capped in practice at trivial incentive levels (~$75–599/yr) by anti-kickback rules despite the strongest evidence base for stimulant UD; recent OIG/HHS action raised the ceiling only modestly | §4/§6 | Prior broken on mechanism, roughly right on magnitude (filled 2026-08-10 from primary text; see verification-log.md). The strong-evidence half is confirmed — OIG's own rule calls CM "the most effective currently available treatment for stimulant use disorders" (85 FR 77684, 77791). But there is no anti-kickback cap: OIG states "There is no OIG-imposed $75 limitation on contingency management program incentives," and the $75 nominal-value guidance runs to the Beneficiary Inducements CMP, not the anti-kickback statute (85 FR 77791–92). The $75 that bound CM was SAMHSA's own grant condition (PEP24-06-001). The real regulatory ceiling is the patient-engagement safe harbor, 42 CFR 1001.952(hh)(5): $500 base, CPI-indexed to 591/605/$623 for 2024/2025/2026, value-based-enterprise only, in-kind only | The "~$75–599/yr" range was accidentally close and structurally wrong. $599 brackets the real indexed safe-harbor figure; $75 was a grant condition. Neither is an anti-kickback cap, and the "one office can raise it" inference does not follow — 42 U.S.C. 1320a-7d(a)(1)(B) requires action by the Secretary *in consultation with the Attorney General*, by notice-and-comment. The CM safe-harbor NPRM (RIN 0936-AA13) sits in Long-Term Actions with a July 2027 target. **Extended 2026-08-10 by the Phase 2 pass** ([steelman-log.md](../research/steelman-log.md)): the prior's premise that CM "is capped in practice at trivial incentive levels" is now testable against two things it lacked. First, a **data-driven effective magnitude** — [Rash et al., *JAMA Psychiatry* 2025;82(9):940–945](https://pubmed.ncbi.nlm.nih.gov/40601332/), from 112 protocols: **$128/week voucher, $55/week prize; $1,536 and $660 over 12 weeks.** The prize benchmark, which is the design real programmes run, sits 5.6% above the 2026 safe harbor and below SAMHSA's 750.Second, * * whatisactuallybeingpaid * * : CMShasapprovedCMcoverageunder§1115in * *fivestates * *atmaximaof * *596–$1,092** (Kaufman et al. 2025, PMC12486863) — Washington's 1.80× the 2025 safe-harbor cap. The prior is broken in a third way: the ceiling is not what constrains CM. The VA has run a national CM programme since 2011 with no such ceiling and delivers it to 1.2% of eligible patients (1,698 of 138,280 with stimulant use disorder, July 2018–December 2020; Coughlin et al., Am J Psychiatry 2025) — the same cohort in which CM is associated with 41% lower one-year mortality (aHR 0.59, 95% CI 0.36–0.95) |
| 10 | Alcohol-attributable deaths ≈ 178k/yr (CDC ARDI) — exceeding peak all-drug overdose deaths; tobacco ≈ 480k/yr | §2 | Confirmed at agency-primary tier, both legs (filled 2026-08-10 by the Phase 2 pass; see steelman-log.md and method/sources/cdc-alcohol-and-tobacco-mortality-via-ncbi.md). Alcohol: 178,307 average annual deaths, 2020–2021, from CDC's Alcohol-Related Disease Impact application across 58 alcohol-related causes — Esser et al., MMWR 2024;73(8):154–161, read in full from PMC (CC0). Up 29.3% from 137,927 in 2016–2017. Tobacco: "at least 480,000 premature deaths annually" — The Health Consequences of Smoking—50 Years of Progress, 2014 Surgeon General's Report, ch. 12, read from NCBI Bookshelf. Against the filing's own verified overdose peak (111,466, 12 months ending June 2023), alcohol alone is 1.60× | Anchor holds; one vintage caveat the filing does not carry. Phase 1 recorded this row's claims as UNVERIFIABLE because cdc.gov 403s every automated fetch — the block is on the route, not the source: MMWR is deposited in PMC and Surgeon General's Reports are on NCBI Bookshelf, and both served cleanly over curl. The caveat: the 480,000 figure is the average annual estimate for the 2005–2009 exposure period, published 2014. It is CDC's standing number, but "kills 4–7 times as many Americans annually" rests on a twenty-year-old base over which smoking prevalence has fallen substantially — an as-of note, not a retraction. Separately, the MMWR's own policy implication is evidence the §10 board never had: CDC recommends "policies that reduce the number and concentration of places selling alcohol and increase alcohol taxes" — the A9 instrument, scored 2 on evidence because §9's tobacco arm was never run |
| 11 | US methadone is dispensable only via ~2,000 OTP clinics with (historically) daily observed dosing; peer countries dispense via community pharmacies; the 2024 federal rule relaxed take-homes but not the OTP monopoly | §4 | Confirmed as to effect; the source of the restriction is narrower than the filing said (filled 2026-08-10). The 2024 rule is verified at primary tier (89 FR 7528): individualized take-home criteria finalized, first-day dose raised to 50 mg, and scope expressly limited — "the final 42 CFR part 8 rule only applies to dispensing of methadone in OTPs." But 21 U.S.C. §823(h) only requires a separate annual registration for practitioners dispensing narcotics for maintenance/detox, granted on "standards established by the Secretary." It never names OTPs, never confines dispensing to them, and sets no clinic count | "Written directly into statute" is overstated. The separate-registration requirement is statutory; the OTP-exclusive structure is 42 CFR Part 8 plus DEA registration standards issued under it. Asked to allow pharmacy dispensing, SAMHSA claimed no statutory bar — it said it "continues to work with Federal partners to explore ways through which access to MOUD might be expanded." Consequence: §11's Tier-3 "requires new statute" placement for methadone deregulation is owed a re-examination |
| 12 | Opioid settlements total ~$50–57B over ~18 years; only a minority of disbursed funds have public, itemized reporting | §7 | Blank — not attempted in Phase 0; queued for §7 execution | |
| 13 | Retail heroin/cocaine real prices fell for decades while purity rose, through the interdiction-budget peak — the price/purity paradox | §5 | Blank — not attempted in Phase 0; queued for §5 execution |
4. Execution notes
Phase 0 first (M6): verify ★ anchors, pull the §2/§3 baseline skeleton (CDC WONDER final + provisional overdose series by drug class and state — committed pipeline; NSDUH treatment-gap tables with the definitional decomposition), and scan three precedents at depth: Portugal (root-trace the canonical claims), France 1995 (root-trace the 80% figure), Oregon M110 (collect the synthetic-control studies). Kill conditions: (a) if §2 shows the treatment-gap number is definitionally underdetermined the way GBMT-2's shortage number was, that is a headline, not a footnote, and every architecture premised on "closing the gap" must restate its target in an identified quantity; (b) if §3 shows the overdose decline is supply-side driven and policy-invariant, demand-side architectures must be scored against that counterfactual, and any architecture selling itself as the cause of the decline gets the M8 stamp it earns; (c) if §9 root-tracing shows Portugal's causal story does not survive, the decriminalization architecture loses its flagship precedent and must stand on BC/Oregon evidence — which currently points the other way. Hypothesis-tilt check (M3): the seed set above leans skeptical of enforcement and of the discourse's favorite numbers; §5's steelman obligation and the Sweden/enforcement-package rows exist to force equal effort in the unfashionable direction — Phase 0 should try to break H5.1 and H5.2, not decorate them. Effort shape on current priors: §3, §4, §9 carry the most information per hour — the overdose decline is live and unclaimed, the treatment system's legal cage is under-reported, and the precedent record is the most citogenesis-contaminated in the project. Per M3, those priors are exactly what Phase 0 should try to break.