Status: Phase 0 executed 2026-08-03 — findings, baseline, deviations. Verdict GO, with the §3 news-desert single-root finding fired as a soft headline, the §7 constitutional screen resolved to "directionally confirmed, not yet operative," and anchor 3's ad-duopoly half broken. Full execution pending. Scope: United States; the supply of civic information — who still covers your town, who owns the feed, and who decides what you get to see. Local and accountability journalism (print, digital, broadcast, public media), the ownership structures above it, and the platform distribution layer that now sits between newsrooms and readers. Excluded by prior decision (PR #2): copyright/DMCA/piracy — a different binding constraint (legal/industry-economics vs. constitutional/editorial-discretion) that would muddy a single scorecard; it is queued as its own future filing, not folded in here. Method: Imports method/gubment-method.md (M1–M9) in full. This document contains only domain content. Artifact: A feasibility assessment of the leading "save the news" theories, naming binding constraints in priority order and scoring candidate policy architectures — Whitepaper No. 9.
1. Definitional decisions to settle before fieldwork
- Which media problem? At least five travel under "the death of the news," with different mechanisms and different fixes: (a) local news collapse — the supply of town/county-level civic information (deserts, ghost papers); (b) business-model collapse — the advertising revenue base of journalism migrating to platforms and to non-news uses; (c) ownership concentration — chains, hedge funds, and broadcast groups consolidating what survives; (d) distribution gatekeeping — algorithmic feeds, moderation, and search deciding what reaches whom; (e) trust and demand — declining willingness to consume or pay for news at all. Treat as linked but distinct, the way GBMT-1 treated age bands and GBMT-2 treated the four housing problems. An instrument aimed at (b) can succeed completely while (a) continues — revenue can recover at national outlets while county coverage stays dead.
- The First Amendment inversion. Every prior filing asked which level of government can act and will it. This domain adds a constraint no prior filing had: the Constitution affirmatively forecloses entire instrument families. Content-based subsidy conditions, compelled carriage, and "must-cover" mandates run into Miami Herald v. Tornillo; the 2024 NetChoice decisions treat platform feed curation as protected editorial speech, which reaches instruments aimed at "who decides what you see" directly. Historically the government's largest media interventions were structural and viewpoint-neutral — founding-era postal rate subsidies, spectrum licensing — and that is probably not an accident. Every candidate architecture in §11 passes a constitutional screen before it is scored on anything else. This is the domain's analog of GBMT-2's Byrd analysis, and it may be the binding constraint.
- What counts as a "local news outlet" — and who counts them. The news-desert map is the domain's headline statistic, and it is definition-driven: whether digital-only startups, public radio, county-seat TV, ethnic media, and "ghost papers" (a masthead with no reporters) count as coverage moves counties in and out of the desert column wholesale. There is exactly one primary census (Medill's State of Local News, inherited from UNC); most public numbers are restatements of it. This is the H2.3-of-media, and the housing-shortage lesson applies with full force: a count quoted without its inclusion criteria is not a finding. Same problem one level down for "journalist": BLS occupational counts, newsroom self-reports, and Medill's per-outlet estimates measure different things.
- Beneficiary — the citizen's information supply, or the incumbent outlet? Mirrors childcare's child-vs.-parent question and elder care's older-adult-vs.-caregiver question. A subsidy can keep a masthead alive — including one owned by the hedge fund that stripped its newsroom — without producing a single additional hour of accountability coverage. Ghost papers are the proof this failure mode is real, not hypothetical. Each candidate architecture states whether its unit of support is the outlet, the journalist, or the coverage itself (e.g., beat-level grants), and the scorecard treats those as different targets.
- Platforms are simultaneously market actors and speech actors — and the law treats each half differently. Antitrust sees Google's ad stack as a market to be restructured (the live US v. Google remedies); speech law sees Meta's feed as an editor to be protected (NetChoice). The same firm is both, in the same product. Workstreams keep the two frames separate: §6 analyzes platforms as markets (ad revenue, referral traffic, bargaining power); §7 analyzes them as governors of distribution (moderation, ranking, access). Instruments that confuse the frames — regulating the market by commanding the speech — are where bargaining codes get into constitutional trouble, and the distinction is load-bearing for §11.
- Objective function (per M6/GBMT-1 lesson): supply of local/accountability coverage, independence from state influence, viewpoint diversity, market competitiveness, and public trust rank candidate architectures differently. A direct-subsidy design scores well on coverage supply and worst on independence-from-state; a structural-antitrust design is near the reverse. Score under explicit weightings; report rank stability, as GBMT-1/2/7 did.
2. Workstreams
§2 Baseline — newsroom employment by sector and occupation (BLS OEWS/CES; QCEW for industry payrolls); outlet counts and closures (Medill census, with inclusion criteria made explicit); revenue by stream — newspaper advertising's collapse from peak, circulation/subscription revenue, digital ad market shares; the public-spending inventory (CPB's appropriation history ending in the 2025 rescission, state programs: NJ Civic Information Consortium, NY/IL employment tax credits, CA's Google deal) vs. philanthropic flows (Press Forward, American Journalism Project, INN member revenue). Committed pipeline: BLS QCEW flat-file API for NAICS 511110 (newspaper publishers) and adjacent industries; OEWS for occupation 27-3023. Seed: H2.1 all public support for journalism at every level of government combined is at least an order of magnitude smaller than the advertising revenue the sector lost; H2.2 no consistent national census of local outlets exists independent of the single Medill lineage — the count is definition-driven (kill-condition test, §3); H2.3 the "platforms took our ad revenue" causal story is substantially wrong in composition — the largest losses (classifieds, retail inserts) migrated to non-news uses (Craigslist, direct e-commerce) that no bargaining instrument can claw back.
§3 The desert map & what fills the void — county-level coverage: true deserts, single-outlet counties, ghost papers with mastheads but no reporting staff; who backfills — TV duopolies, public radio, digital startups, and "pink slime" networks (partisan operations wearing local-news mastheads). The desert-metric lesson from GBMT-1 applies: measure coverage capacity (reporters on the ground, stories filed about local government), not outlet counts, wherever the data permits. Seed: H3.1 the desert count moves by >2× under defensible alternative inclusion criteria (this is the kill-condition test); H3.2 pink-slime outlets concentrate where legitimate local outlets died — the void gets filled, just not with journalism; H3.3 ghost papers make outlet-count metrics systematically optimistic: the masthead survives the newsroom by years.
§4 The civic-harms literature (audit before you build on it) — the empirical case for intervening at all: local news loss → lower turnout and split-ticket voting (nationalization of politics), higher municipal borrowing costs, less corporate and governmental accountability, higher polarization. This literature is the foundation every candidate architecture stands on, and the Auckland lesson (GBMT-2 Phase 0 §4) says foundations get audited: independence structure (how many findings trace to the same authors/data), replication status, and identification quality — searched per M2 (OpenAlex), not recalled. Seed: H4.1 at least two civic-harm results (municipal borrowing costs; turnout/nationalization) survive an independence audit with multiple arm's-length lineages; H4.2 the effect sizes commonly quoted in advocacy for journalism subsidies are the literature's upper bounds, not its central estimates.
§5 Ownership & consolidation — chain ownership (Gannett post-merger), hedge-fund/private-equity owners (Alden Global's model), broadcast groups (Nexstar, Sinclair; the FCC national cap, the UHF discount, and the live 2025–26 deregulation/merger wave); what ownership type measurably does to newsroom staffing and coverage output — the domain's analog of elder care's PE-nursing-home literature (§6 of GBMT-7), located by search rather than recalled headline. Seed: H5.1 hedge-fund acquisition produces staffing/coverage cuts measurably beyond the secular industry trend (size it against matched non-hedge-fund papers); H5.2 broadcast consolidation caps are the one media-ownership lever where the federal government retains clearly constitutional authority (spectrum licensure), and it is moving toward less restraint exactly as print ownership concentrates.
§6 Platform economics & the bargaining question — platforms as markets: US digital ad revenue shares (the "duopoly" claim, tested against current data — Amazon's rise), the US v. Google search and ad-tech remedies as live structural interventions; referral-traffic dependence and its collapse (Meta's news deprioritization, Google AI Overviews and zero-click search); the bargaining-code experiments — Australia's News Media Bargaining Code, Canada's C-18 and Meta's news ban in response (the domain's cleanest natural experiment in asymmetric dependence), California's negotiated settlement. Seed: H6.1 bargaining-code payments flow disproportionately to the largest incumbent publishers, not to local outlets — the instrument's distributive reality diverges from its local-news marketing; H6.2 Canada's Meta news ban measurably damaged Canadian news outlets' reach while leaving Meta's Canadian usage essentially untouched — establishing which side needs the other, which is the fact any bargaining instrument's leverage theory depends on; H6.3 platform referral traffic to news is in secular decline independent of any regulation — an instrument that taxes the referral relationship is taxing a melting asset.
§7 Distribution gatekeeping & the constitutional screen — platforms as speech governors: feed ranking, moderation, deplatforming; Section 230's actual mechanics vs. its public reputation; the 2024–26 case law (Moody v. NetChoice — feed curation as protected editorial judgment; Murthy v. Missouri — jawboning; the TikTok divestiture-or-ban statute as the structural exception that survived). The workstream's output is the constitutional screen §11 applies: which instrument families are foreclosed, which are contested, which are safe. Seed: H7.1 after NetChoice, instruments that operate on content or ranking (must-carry, anti-bias mandates, algorithmic requirements) are foreclosed or heavily contested, while instruments that operate on structure or money (antitrust, taxes, subsidies, transparency-with-limits, data portability) survive — pushing every viable architecture toward the viewpoint-neutral/structural family; H7.2 Section 230 repeal/reform proposals are largely orthogonal to the local-news problem despite dominating the policy conversation — a salience/relevance mismatch worth documenting.
§8 Public & nonprofit media after the defunding — CPB's 2025 rescission ($1.1B, the first successful rescission of an appropriated CPB budget) and wind-down: which stations closed or cut, and where — testing whether the loss concentrates in small/rural markets where the public station was the last local newsroom standing; the international comparison (US federal public-media spending per capita vs. peer democracies — an order-of-magnitude gap even before the rescission, verify the actual multiples); the nonprofit sector as the would-be replacement — Press Forward's committed capital, AJP, INN member growth and revenue, and whether the philanthropic pipeline is within an order of magnitude of replacing lost commercial newsroom capacity. Seed: H8.1 the CPB cut binds hardest precisely in the deserts — rural and small-market stations with the highest CPB revenue share and no commercial substitute; H8.2 total philanthropic commitment to local news is under 10% of the sector's lost annual revenue — a supplement, not a replacement, and any architecture that assumes "philanthropy scales" must price that gap.
§9 Civil society & political economy — the fault-line map, per GBMT-1's §13: press-freedom organizations that oppose state subsidy on capture grounds; subsidy advocates (Rebuild Local News, Free Press); platform-accountability advocates whose instruments the constitutional screen may kill; incumbent trade groups (News/Media Alliance, NAB) whose preferred designs favor incumbents; the platforms themselves. The political-economy asymmetry to map: unlike housing's satisfied-incumbent-homeowner majority, media policy's blocking constituency is bipartisan distrust — each side expects the other's government to weaponize any instrument it builds. That symmetry-of-fear is a design input (it argues for automatic, formula-based, insulated instruments), not just an obstacle. Public trust in media (near record lows) is the demand-side constraint on any architecture that assumes people want the product back.
§10 Precedents — Domestic: the founding-era postal subsidy (the original structural media subsidy — quantify its scale in modern terms; it is the strongest historical evidence that viewpoint-neutral infrastructure subsidy is compatible with a free press); CPB 1967–2025 (the full arc: creation, insulation design, half a century of survival, then defunding by rescission — the durability lesson for any new institution); the NJ Civic Information Consortium (small, live, state-level); NY's 2024 first-in-nation payroll tax credit for local news employment (the live state natural experiment — uptake and early results); Illinois' 2024 credit; California's 2024 Google settlement in lieu of legislation. International: Australia's NMBC (deals struck, then Meta's 2024 walk-away — what leverage actually existed); Canada's C-18 (the cautionary case: legislation → news ban → measurable outlet damage); the Nordic direct press subsidies (Norway/Sweden/Denmark — decades of direct state subsidy to newspapers with no evident state capture: the steelman for subsidy, and the design details that made it survivable); the BBC licence-fee model under strain. Each with transferability assessment per the GBMT-1/2/7 template.
§11 Candidate architectures (seed list, to be scored per M6, each passing §7's constitutional screen first): federal refundable payroll tax credit for local journalists (the Local Journalism Sustainability Act shape, NY model federalized — support tied to employing reporters, not to outlets' opinions); platform bargaining code / final-offer arbitration (Australia model, priced against §6's evidence on who gets paid and what leverage exists); digital advertising tax earmarked to a civic-information fund (Maryland's litigated precedent as the cautionary tale); public media re-funding with redesigned insulation (endowment or multi-year automatic formula vs. annual appropriation — CPB's failure mode); demand-side news vouchers / tax credit for subscriptions and donations (viewpoint-neutral by construction, citizen-directed); state/municipal civic-information consortia (NJ model scaled); philanthropy-matching fund (public match on Press Forward-style commitments); structural platform remedies (ad-tech divestiture outcomes from US v. Google, data portability/interoperability) rebadged as media policy; "managed transition" — no rescue, let the nonprofit/digital ecology finish emerging — as the do-nothing comparator; and cash (the universal comparator, per M7). Seed: H11.1 (the GBMT rhyme): demand-side money without distribution reform inflates incumbent revenue without adding county-level coverage — the voucher flows to the New York Times, not the Nebraska weekly; H11.2 the constitutional screen plus the political symmetry-of-fear (§9) jointly select for automatic, formula-based, viewpoint-neutral designs and against discretionary grant-making — the postal-subsidy shape, rediscovered.
§12 Sequencing — measurement first (the desert map's definitional problem must be fixed before any geographic targeting can work); the constitutional screen second (it is cheap and prunes the tree); state natural experiments (NY, NJ, IL, CA) read before federal design; platform-side structural outcomes (US v. Google remedies) treated as exogenous events to design around, not levers this policy area controls. To be tested, not assumed.
3. Anchor Table (seed — verify per M4; ★ = candidate for Phase 0)
Figures below are approximate priors stated to orient the search, not findings — the same discipline as GBMT-1/2/7. Expect several to be wrong; a wrong anchor is a finding about this protocol, not an error to quietly correct.
| # | Anchor (unverified prior) | Verify in | Verified value & source | Delta |
|---|---|---|---|---|
| 1 ★ | Newspaper newsroom employment has fallen by more than half since the mid-2000s; total newsroom employment (all sectors) fell far less because digital gained | §2 | Confirmed. Newspaper newsroom employment 2008→2020: ~71,000→~31,000 (-57%, Pew/BLS-OEWS). All-5-sector newsroom employment: ~114,000→~85,000 (-26%). Digital-native newsroom jobs +144%. BLS CES total newspaper-industry employment (all jobs): 2001 peak ~412,000 → 2024 ~90,800 (-78%) | Confirmed on both halves. Single-root: all three Pew figures are BLS-OEWS-derived; ASNE's discontinued (~2015) census is the one independent corroborating source |
| 2 ★ (KILL CONDITION TEST) | Roughly 2.5 newspapers close per week; 200+ US counties have no local news outlet; over half have one or none (Medill-lineage figures — test whether the count survives alternative defensible inclusion criteria) | §3 | Confirmed arithmetically (2025: 212 zero-outlet counties, 1,525 one-outlet, ~55% combined of 3,144 — [updated 2026-08-10: Medill's report and dataset say 212/1,525; 213 appears only in its landing-page essay. The combined total, 1,737, is unchanged either way]; ~3,500 papers lost since 2005; 136 closures in the past year per the report itself) | Does not move 2x, but is single-root (all figures trace to one UNC/Abernathy→Medill lineage) with two undocumented, opposite-direction biases: commercial TV/radio entirely excluded (undercounts coverage) vs. "ghost papers" counted as served (overcounts it). Neither sized by any publication found. Soft-fires the kill condition — see phase0-findings §1. [Updated 2026-08-10 (Phase 2): still neither sized. The Mississippi DMA-overlay check does not size the TV bias — Nielsen assigns every US county outside parts of Alaska to one of 210 markets, so its condition holds for well-served counties too. On the capacity measure the protocol's own §1.3 prefers, and which counts TV and radio journalists, the deficit is larger, not smaller: the Local Journalist Index 2026 finds 70% of counties below the national average of 7.8 journalist-equivalents per 100,000, and 77% producing zero local education articles in Q1 2026. See media/research/steelman-log.md, target A.] |
| 3 ★ | Newspaper advertising revenue fell on the order of 80% from its mid-2000s peak; Google+Meta together take a majority of US digital ad revenue (verify both halves — the second may be stale as Amazon rises) | §2 §6 | Ad revenue: ~49B(2005peak)→ 9.6-9.76B (2020/2022) = **~80% decline**, confirmed. Google+Meta combined digital-ad share: fell below 50% in 2024; ~47.1% (2025) trending to ~44.8% (2026) as Amazon rises to ~17.3% by 2026 (not 2024) [updated 2026-08-10: every figure here traces to one estimator, eMarketer, whose vintages disagree about the crossing year — single-root by this filing's own standard] | Prior broken on the ad-duopoly half — independently re-verified, not just subagent-sourced. First half confirmed |
| 4 ★ | CPB's ~535M/yrfederalappropriationwasrescindedin2025( 1.1B over two years) and CPB is winding down; station-level damage concentrates in small/rural markets (verify current status and the incidence claim) | §8 | Rescissions Act of 2025 (Public Law 119-28, signed Jul 24 2025) rescinded CPB's FY2026+FY2027 appropriations, 535M/yreach, 1.1B total — independently re-verified. CPB's board voted to dissolve Dec 10 2025 (announced Jan 5 2026) — independently re-verified. Rural grantees averaged 17% of revenue from CPB vs. 9% non-rural; 33 (many tribal) relied on it for ≥50% | Sharper than stated, not broken — CPB didn't just lose funding, it ceased to exist after 58 years. See phase0-findings §3 |
| 5 | Local news loss causes measurable civic harms: higher municipal borrowing costs (single-digit-to-low-teens basis points in the canonical study), lower turnout, more straight-ticket voting (verify the specific estimates and the literature's independence structure) | §4 | Borrowing costs: +5.5-6.4bps (Gao/Lee/Murphy, JFE 2019) — single-lineage, no independent replication; the widely-cited "$1.1B/yr" 2026 update is co-authored by the same lead author via an advocacy nonprofit. Turnout/straight-ticket voting: confirmed directionally by 3 independent teams (US, different data/methods, no shared authors). Corporate accountability: partially corroborated across 3 teams/domains, shared-closure-dataset risk not ruled out. Polarization: confirmed by 2 independent international teams (US, Germany); no causal trust evidence found | The literature is uneven, not uniformly weak. The single most-quoted number (municipal borrowing costs) is the one that fails the independence test; turnout/polarization pass it. See ws03-04-findings §4 |
| 6 ★ | Canada's C-18 led Meta to block news for Canadian users (Aug 2023); Canadian news outlets lost a large share of social referral traffic/engagement while Meta's Canadian usage was essentially unaffected | §6 §10 | Ban began Aug 1 2023, still in effect mid-2026. One-year effects (Media Ecosystem Observatory): national outlets -64%, local outlets -85% FB/IG engagement [Updated 2026-08-10 — withdrawn: the cited MEO report contains no local-versus-national split and no "64%" figure; its 85% is an all-outlets Facebook+Instagram number. See media/research/verification-log.md, finding 1.]; ~30% of local outlets went dormant. Meta's own Canadian usage stayed essentially flat (Reuters/Similarweb + MEO survey, two source types) |
Confirmed, with the local/national asymmetry as the sharper unstated finding — corrected 2026-08-10: the asymmetry now rests on the report's own local-specific figures (30% of the 713 local outlets dormant; 212 of the 217 that went dark, 98%, were local). Single-institution flag: nearly all effect-size data traces to one research body (MEO), not yet peer-reviewed |
| 7 | Australia's bargaining code produced on the order of AU$200M/yr in platform-publisher deals; Meta declined to renew its deals in 2024 | §6 §10 | AU$200M+/yr and Meta's 2024 walk-away confirmed, but the dollar figure traces to one person's estimate (Rod Sims/JNI), not an audited total | New finding: ~90% of the money went to three large incumbents (Brevini & Ward, peer-reviewed) — clearest evidence yet against "bargaining codes fund local news." Australia's legislative response remains unenacted as of mid-2026. [Update 2026-08-04: bill finalized Aug 3 2026 — a 2.5% levy on platforms without news deals — pending introduction to Parliament; still not enacted. See moving-targets-log.md] [Updated 2026-08-10 — withdrawn and replaced: the ~90% citation names a paper that does not exist, and deal values are confidential, so no distributional split is auditable (Australian Treasury's own statutory review says it could not obtain them). Treasury's review instead concludes the Code "has been a success to date," records 30+ agreements across large and small, metropolitan and regional businesses, and documents the ABC placing reporters in 19 regional locations, 10 previously unserved. The Aug 3 2026 NBI doubles the regional/small-publisher distribution loading from 10% to 20% and adds a grants program for publishers under $150k. See media/research/steelman-log.md, target B.] |
| 8 | The largest broadcast groups reach a large share of US TV households under the 39% national cap as modified by the UHF discount; a 2025–26 deregulation and merger wave (e.g., Nexstar–Tegna) is testing the cap (verify current status) | §5 | Confirmed, and imminent. Nexstar-Tegna closed Mar 2026 above the cap (~80% raw / ~54.5% UHF-discounted reach) via an ad hoc FCC waiver. The FCC has a vote scheduled Aug 6 2026 — 3 days after this research — to formalize case-by-case waivers in place of the hard cap, independently re-verified | Confirmed and time-critical: this is a live regulatory inflection, not a stable baseline. See ws05-findings |
| 9 ★ | Moody v. NetChoice (2024) held that platform feed curation is protected editorial speech, without finally resolving the Texas/Florida statutes (verify the holding's actual scope — it is the hinge of §7's screen) | §7 | The editorial-speech holding commanded a real 6-justice majority (stronger than "dicta"). [Updated 2026-08-10 — corrected: five justices joined Part III–B, the platform-curation reasoning; Jackson joined only Parts I, II and III–A and declined to reach the merits; Alito, Thomas and Gorsuch called the discussion "nonbinding dicta." See media/research/verification-log.md.] Both TX HB 20 and FL SB 7072 remain unresolved in active litigation as of mid-2026 |
Directionally confirmed; precision added on vote composition and litigation status. The screen is directional, not yet final — see phase0-findings §4 |
| 10 ★ | New York enacted the first-in-nation payroll tax credit for local news employment (2024, ~$30M/yr for 3 years); early uptake exists but is modest (verify design, uptake, and any results) | §10 | Design confirmed: $30M/yr, 50% of salary up to $50k, capped $300-320k/business, authorized through Dec 2027. First application cycle only closed April 25 2026 — no results published | "Modest uptake" not verifiable — too early to tell, same shape as elder care's WA Cares finding. Bonus: CA's parallel $175M/5yr Google deal was collapsing at filing time (Newsom's January 2026 proposal zeroed the state match). [Update 2026-08-04: the enacted FY2026-27 budget restored $10M/yr × 2 years, matched dollar-for-dollar by Google — survived at ~8× reduced scale, not zeroed. See moving-targets-log.md] |
| 11 | US federal public-media spending per capita was on the order of 1.50beforetherescission, vs.tens − to−>100 in Northern European democracies (verify the comparative figures and their vintage) | §8 | Neff & Pickard: two distinct point estimates, not a range — $1.40/capita (federal-only) vs. $3.16/capita (broader total-public-funding, the figure used in their peer-reviewed paper; US ranks 25th of 33 countries) | Both real, from the same team, different scopes. Comparator figures (Norway ~111, Germany 142) not independently pinned to a single vintage — flagged, not resolved |
| 12 | Roughly 3 in 10 Americans express trust in mass media, at or near the series' record low (Gallup-style figure — verify current) | §9 | Confirmed and worse than stated. Gallup Sept 2025: 28%, first time below 30%. Partisan split is the real story: Republicans 8%, Independents 27%, Democrats 51% | Confirmed; the partisan gap (43 points) is more load-bearing than the topline. See ws09-findings |
| 13 | Philanthropic commitments to local news (Press Forward et al.) total on the order of $500M cumulative, against sector revenue losses measured in tens of billions annually (verify both sides of the ratio) | §8 | Press Forward + AJP combined ≈ 700M * *deployed/raisedacross3 − 7yearsvs. * *38.7B single-year (2022 vs. 2005 peak) ad+circulation shortfall. Ratio ≈ 55× | Prior undersold the gap. Not "$500M vs. tens of billions annually" (implying roughly one order of magnitude) but ~1.7 orders of magnitude short of one year's loss alone, ~2-3 orders of magnitude short cumulatively. See ws02-08-findings |
4. Execution notes
Phase 0 first (M6): verify ★ anchors, pull the §2 baseline skeleton (BLS QCEW NAICS 511110 + OEWS 27-3023 as a committed pipeline), scan the C-18/Meta-ban natural experiment + the CPB rescission aftermath + the NY tax credit as the three live precedents. Kill conditions: (1) if §3 shows the news-desert count is not derivable within a defensible range under stated inclusion criteria, that is the headline — the domain's version of housing's shortage finding, and it lands on the statistic every advocacy document leads with; (2) if §4's civic-harms literature fails the independence audit (single lineages, no arm's-length replication), the case for intervening — not just the instrument choice — is weaker than advertised, and that is a headline, not a footnote; (3) if §7's screen forecloses the most-discussed instrument families, the deliverable's frame becomes "what remains constitutionally available," not "what polls best." Effort shape: §2, §3, §6, §7, and §10 carry the most information per hour on current priors — but per M3, those priors are exactly what Phase 0 should try to break. Watch the tilt: the hypothesis set above leans toward "subsidy skepticism + constitutional pessimism"; the steelman owed in the unfashionable direction is the Nordic case — direct state subsidy that worked for decades without capture — built with equal effort.