Protocol & scope
United States civic information supply — local and accountability journalism, the ownership structures above it, and the platform distribution layer between newsrooms and readers. Copyright/DMCA/piracy explicitly excluded as a separate future filing. Imports the gubment method (M1–M9) in full: two-source rule, root-tracing for every claim, pre-registered hypotheses, and a deviations log that stays visible rather than getting edited away. This is the first Gubment filing to publish a dedicated constitutional screen (§7) — the First Amendment forecloses whole instrument families before economics or politics enter the picture.
Twelve workstreams executed across Phase 0 and two full-execution passes on 2026-08-03: the baseline, the desert map and its hidden biases, the civic-harms literature audit, ownership and consolidation, platform economics, the constitutional screen, public spending and philanthropy, political economy and trust, historical/international precedents (including a required subsidy steelman), an architecture scorecard, and a sequencing plan. A red team pass followed, catching and correcting two overclaimed findings before publication. Since publication the filing has been through both phases of the Verification Protocol: a Phase 1 fact-check (556 claims, all verdicts public) and a Phase 2 steelman built against the filing's own conclusions, which withdrew one of them.
Anchor table — priors, stated before evidence, verified after
Every anchor below was written down as an unverified guess before research began, so it could be broken. ★ rows were verified in Phase 0; the rest were verified during full execution.
| # | Anchor (unverified prior) | Verified value & delta |
|---|---|---|
| 1 ★ | Newspaper newsroom employment fell by more than half since the mid-2000s; total newsroom employment fell far less | Confirmed on both halves. Newspaper newsroom: -57% (2008→2020). All-sector newsroom: -26%, cushioned by a 144% rise in digital-native jobs. Single-root caveat: all three headline figures are BLS-OEWS-derived. |
| 2 ★ (kill condition) | ~2.5 newspapers close/week; 200+ counties have no local outlet; over half have one or none | Confirmed arithmetically (212 zero-outlet, 1,525 one-outlet, ~55% combined — corrected 2026-08-10 from 213/1,524) but single-root (one UNC/Abernathy→Medill lineage) with two undocumented, opposite-direction biases. Soft-fires the kill condition rather than a hard stop. Phase 2, 2026-08-10: still neither bias sized — the Mississippi TV-market check cannot size one, because Nielsen assigns every US county outside parts of Alaska to a market. On a reporter-count measure that includes TV and radio, the deficit is larger: 70% of counties below 7.8 journalist-equivalents per 100,000 (Local Journalist Index 2026). |
| 3 ★ | Newspaper ad revenue fell ~80% from mid-2000s peak; Google+Meta take a majority of digital ad revenue | Ad revenue: confirmed, ~80% decline. Ad-duopoly claim: broken — combined share fell below 50% in 2024 as Amazon rises. |
| 4 ★ | CPB's ~$535M/yr appropriation was rescinded in 2025 (~$1.1B/2yr) and CPB is winding down; damage concentrates in small/rural markets | Sharper than stated: CPB's board voted to dissolve the institution outright (Jan 2026), not merely wind down. Rural/tribal incidence confirmed (17% vs. 9% revenue share). |
| 5 | Local news loss causes measurable civic harms (borrowing costs, turnout, polarization) | Uneven, not uniformly weak. Municipal borrowing costs: single-lineage, the widely-cited "$1.1B/yr" update is co-authored by the original lead author via an advocacy nonprofit. Turnout/polarization: confirmed directionally by independent teams across countries. |
| 6 ★ | Canada's C-18 led Meta to block news; Canadian outlets lost referral traffic while Meta's usage was unaffected | Confirmed, then corrected 2026-08-10: the cited MEO report contains no local-versus-national split and no 64% figure — its 85% is an all-outlets Facebook/Instagram number. The asymmetry survives on the report's own local-specific figure: 30% of the 713 local outlets previously active on social went dormant, and 212 of the 217 outlets that went dark (98%) were local. Meta's own Canadian usage stayed flat by two independent measures. |
| 7 | Australia's bargaining code produced ~AU$200M/yr in deals; Meta declined to renew in 2024 | AU$200M+/yr and Meta's 2024 walk-away confirmed. Withdrawn 2026-08-10: the ~90%-to-three-incumbents finding is cited to a paper that does not exist, and the Australian Treasury's own statutory review states it could not obtain deal values from platforms or publishers at all — so no distributional split is auditable. That review instead concludes the Code "has been a success to date," records 30+ agreements across large and small, metropolitan and regional publishers, and documents the ABC placing reporters in 19 regional locations, 10 previously unserved. |
| 8 | Broadcast groups reach a large share of TV households under the 39% cap; a 2025-26 deregulation wave is testing it | Confirmed and imminent — Nexstar-Tegna closed above the cap via waiver (Mar 2026); FCC vote scheduled Aug 6 2026 to formalize case-by-case waivers. |
| 9 ★ | Moody v. NetChoice held feed curation is protected editorial speech, without finally resolving TX/FL statutes | Confirmed with precision added, then corrected 2026-08-10: five justices joined the platform-curation reasoning (Part III–B); Justice Jackson joined only Parts I, II and III–A and declined to reach the merits, and three justices called the discussion “nonbinding dicta.” Both TX HB20 and FL SB7072 remain unresolved in litigation. |
| 10 ★ | NY's payroll tax credit (2024, ~$30M/yr); early uptake exists but is modest | Design confirmed; "modest uptake" not verifiable — the first application cycle only closed Apr 2026. Bonus finding: CA's parallel Google deal shrank ~8× from its $175M vision but survived — the enacted FY2026-27 budget restored $10M/yr × 2 years with a dollar-for-dollar Google match [updated 2026-08-04; the original filing recorded it as zeroed, per the January budget proposal]. |
| 11 | US public-media spending per capita ~$1.50 vs. tens-to-$100+ in peer democracies | Confirmed as two distinct point estimates (not a range) from the same research team: $1.40/capita (federal-only) vs. $3.16/capita (broader total, the paper's own figure; US ranks 25th of 33 countries). |
| 12 | Roughly 3 in 10 Americans trust mass media, near the series' record low | Confirmed and worse: 28% (Sept 2025), first time below 30%, with a 43-point partisan gap (Republicans 8%, Democrats 51%). |
| 13 | Philanthropic commitments (~$500M cumulative) vs. sector losses (tens of billions annually) | Undersold: Press Forward + AJP combined ≈$700M vs. a $38.7B single-year shortfall — a ratio of ~55×, not the ~1 order of magnitude the seed implied. |
Workstream findings
Each workstream was executed with a full source register and two-source-rule discipline. Full writeups are in each workstream's own findings file in the repository.
§0 · Phase 0Desert count single-root, ad duopoly broken, constitutional screen directional
Verified all 7 starred anchors plus two bonus rows. Verdict GO, with the news-desert count firing as a soft kill-condition headline and the constitutional screen resolving to "directionally confirmed, not yet operative."
§3+4 · Civic-harms audit & desert bias sizingThe most-quoted advocacy number is the weakest-evidenced one
Municipal borrowing costs — the figure every "local news pays for itself" argument reaches for — fails the independence test; turnout and polarization findings pass it. Sized the TV/radio-exclusion bias directly: all 5 of Mississippi's zero-outlet counties sit in a covered TV market.
§5 · Ownership & consolidationA live regulatory inflection, not a stable baseline
Hedge-fund ownership cuts newsroom staffing 14% (peer-reviewed, independently re-verified). Broadcast concentration is accelerating on two tracks: an ad hoc merger waiver already set (Nexstar-Tegna) and a general rule change (FCC vote, Aug 6 2026) about to formalize it.
§2/§8 · Public spending & philanthropyPhilanthropy is 1.7–3 orders of magnitude short, not one
Illinois's tax credit is the one architecture in the filing with real disbursement data proving it works at small scale. Press Forward + AJP combined are ~55× short of a single year's ad-revenue shortfall vs. the 2005 peak.
§9 · Political economy & trustThe reform coalition is split on its own signature bill
Media trust hit a new low (28%, 43-point partisan gap). Incumbent trade groups champion the JCPA bargaining-code exemption; structural-antitrust reform groups (Free Press, Public Knowledge) actively oppose it as incumbent-favoring — the same critique this filing's own evidence supports independently.
§10 · Precedents, including the required subsidy steelmanThe postal subsidy is a genuine steelman; Nordic subsidies are a mixed one
The founding-era postal rate subsidy — universal, formula-based, viewpoint-neutral — is the strongest case for capture-resistant design. Nordic subsidies are credited in the policy literature with avoiding viewpoint capture, though not by the financial study we cite; they didn't prevent commercial decline, and Finland retrenched almost entirely. Phase 2, 2026-08-10: this workstream's negative claim — that no funding mechanism outside the appropriations cycle has a completed, successful durability test — is false. Germany's household broadcasting levy was blocked by one hostile state legislature in 2020 and its constitutional court ordered the increase into effect, on a duty to fund broadcasting the First Amendment does not create.
§11 · Architecture scorecardNo single fix wins, corrected via red team rather than asserted
Eleven candidate architectures scored on 5 anchored dimensions. The first draft's "structural antitrust wins every weighting" headline was withdrawn after a red-team pass found its underlying score unverified; a blind re-score then put public media outside the appropriations cycle on top. Phase 2, 2026-08-10: the bargaining code's bottom placement is withdrawn — three of its four 1-scores rest on an unauditable figure from a nonexistent paper, and it re-scores to mid-board. The leader's margin rests on a diversity score inherited from the dissolved CPB, not on the durability score the honesty box flags; corrected on both cells it falls to fifth. Four of the five weightings are still uncommitted; the fifth is now computed and does not change the leader.
§12 · SequencingTrack exogenous remedies, redesign funding mechanism before scale, pair vouchers with distribution reform
Structural antitrust remedies are already moving through the courts and cost nothing to track. Public-media funding mechanism (inside vs. outside the appropriations cycle) should be resolved before scale is negotiated politically — but per the red-team correction, this is a necessary, not solved, design question. Phase 2, 2026-08-10: the "redesign the bargaining code's distribution formula first" recommendation survives for the US, but the redesign is not hypothetical — Canada legislated a per-journalist formula with hard incumbent caps in 2023, and Australia doubled its regional/small-publisher loading on this filing's own filing date.
Deviations log
Every departure from the protocol as originally written, logged with its reason and effect on findings — per the method, this stays part of the record, not an appendix to it. 18 entries; the eleven filed with the whitepaper are below, and the fuller table lives in the repository.
| # | Deviation | Effect on findings |
|---|---|---|
| 1 | Anchor rows 5, 8, 12, 13 not attempted in Phase 0 | Left blank, not guessed; all four closed during full execution |
| 2 | Anchor verification delegated to parallel subagents, QA-reviewed by the primary session | Three of the most surprising claims (CPB's dissolution, the Rescissions Act's figures, the ad-share drop below 50%) were independently re-verified by direct search before entry |
| 3–4 | Census CBP and BLS QCEW's open API both have no pre-2014 coverage for this industry slice | The committed baseline pipeline covers 2014–2025 only; mid-2000s peak figures rest on secondary literature, stated explicitly rather than presented as pipeline-verified |
| 5 | Several primary sources (congress.gov, cpb.org, one paywalled journal) returned fetch errors | Claims resting on them rest on secondary citation, flagged per-instance; none were load-bearing for the GO verdict |
| 6 | Non-starred rows 6, 7 (Canada/Australia bargaining codes) verified together beyond minimum scope | Produced the report's strongest bonus finding on incumbent-capture ahead of full §6/§9 execution |
| 7 | Post-Phase-0 execution prioritized the civic-harms audit and desert bias-sizing over strict workstream order | Per the protocol's own re-ranking; rows 8, 12, 13 remained queued one pass longer |
| 8 | The ghost-paper bias was sized only via an existing 2018 literature estimate, not a fresh same-state audit | The two desert-count biases cannot yet be netted against each other for one state — queued for the next pass |
| 9–10 | §5, §2/8, §9, §10 executed as a single parallel batch; §11/§12 synthesized directly by the primary session | §11/§12 are integration tasks across all prior workstreams, synthesized by one author rather than fanned out, consistent with the childcare/housing/drugs pattern |
| 11 | The §5 broadcast-ownership finding is time-sensitive to the exact research date (Aug 3 2026, 3 days before the FCC's scheduled vote) | Flagged explicitly rather than presented as a stable baseline; should be re-checked against the actual vote outcome before further citation |
Red team
Seven attacks on the §11 scorecard/§12 sequencing, six absorbed outright with corrections applied, none dismissed without a documented reason. Full account in ws11-red-team-log.md.
Attack 1 — the anchored scales smuggled conclusions in
Absorbed. The original durability and trust axes named specific mechanisms ("court-driven," "citizen-directed") as their own top-tier examples, reverse-engineered from the answer. Rewrote both scales to be abstract and track-record based.
Attack 2 — three specific cells were vulnerable, most notably structural antitrust's market-competitiveness score
Absorbed, and the underlying gap was worse than flagged. No workstream in this project ever verified US v. Google's actual remedies; the citation used covered a different market entirely. Downgraded and disclosed as an open gap rather than a scored finding.
Attack 3 — the weighting scheme never tested the one objective (viewpoint diversity) most likely to flip the ranking
Absorbed. Added a fifth, pluralism-first weighting specifically to stress-test the original ranking. It does produce a different leader.
Attack 4 — the "wins every weighting" headline is close to circular given attacks 1–3
Absorbed — the headline is substantially walked back. The corrected, more modest finding: no single architecture dominates across every weighting tested.
Attack 5 — "redesign public-media funding outside the appropriations process" ignored a counter-example already in this project's own findings
Absorbed — the most consequential correction in the log. Finland's identically-designed subsidy was cut ~96% anyway. Downgraded from "solved lesson" to "necessary but unproven condition."
Attack 6 — no hybrid/combination architectures were scored, though obvious complementary pairs exist
Absorbed, qualitatively, with the gap disclosed rather than pretended-solved. Two plausible hybrids flagged for the next pass rather than invented numbers assigned to them.
Attack 7 — labeling 9 of 10 architectures "SAFE" overstates the constitutional screen's own rigor
Absorbed. The screen does real gating work on only one candidate; the table now states this plainly and distinguishes settled doctrine from an unresolved trend.