Date: 2026-08-03; red-teamed same day (ws11-red-team-log.md); independently re-scored and reconciled 2026-08-04 (ws11-rescore-log.md) — 24 cells materially reconciled (the B axis was still leaking design labels; five below-neutral-without-evidence violations; the comparators' cells gained verified-trajectory citations). This is the v3 matrix; per-cell bases live in the rescore log. Method: Per M6 — anchored ordinal scales (1-5) authored to be abstract and outcome/track-record based, one axis per objective in §1.6 of the protocol, every cell cited to a workstream finding. Rankings computed under five objective weightings; rank-stability reported honestly, including where it does not hold. A constitutional screen (§7) is applied as a separate gate, not folded into the numeric score. This version supersedes the pre-red-team draft (still visible in git history) — the original claimed a stronger, more rank-stable result than the corrected scoring supports; see the red-team log for the full account.
Anchored ordinal scales (revised — abstract, not naming specific mechanism types)
A — Supply of local/accountability coverage added 1=no capacity added, or added only to already-strong outlets · 3=moderate, real add but not desert-targeted · 5=large, demonstrated add specifically in undercovered areas
B — Independence from state influence (demonstrated durability against a hostile political majority, not a design label) 1=no demonstrated track record of surviving a hostile majority, or a completed real-world test that failed · 3=has meaningfully outlasted normal political cycles but with no completed test against a determined hostile majority · 5=multi-decade demonstrated record of surviving a hostile majority without being unwound
C — Viewpoint diversity / anti-incumbent-capture 1=documented to concentrate resources in incumbents · 3=neutral · 5=demonstrated to reach small/local/nonprofit outlets in practice
D — Market competitiveness (measured against project-internal verified evidence only — a plausible theoretical effect with no project-sourced verification scores no higher than 3) 1=entrenches concentration · 3=neutral or theoretically plausible but not project-verified · 5=project-verified, direct reduction in concentration
E — Public trust effect (segmented per §9's 43-point partisan gap) 1=plausibly widens the partisan gap · 3=neutral/uncertain · 5=plausibly narrows it
Scores (A, B, C, D, E) with citations
| # | Architecture | A | B | C | D | E | Basis |
|---|---|---|---|---|---|---|---|
| 1 | Federal payroll tax credit (NY/IL/NM federalized) | 3 | 1 | 4 | 3 | 3 | IL: 260+ jobs (2025), 30% to nonprofits, mostly outside Chicago metro — ws02-08. NY's ownership-neutral eligibility gap (excludes Gannett, not Alden) is a real anti-incumbent leak — phase0-findings §6 |
| 2 | Platform bargaining code (Australia model) | 1† | 1† | 1† | 1† | 3 | ~90% of AU money to 3 incumbents (Brevini & Ward, peer-reviewed); Meta's walk-away shows fragility to platform exit; opposed by the reform coalition itself (Free Press, Public Knowledge) — phase0-findings §5, ws09 |
| 3 | Digital ad tax earmarked to a civic-info fund | 3 | 2 | 3 | 3 | 3 | Untested US design; Maryland's non-earmarked analog already lost its pass-through provision on First Amendment grounds — phase0-findings §4. Plausible-if-well-designed, not evidenced |
| 4 | Public media re-funding, appropriation-funded variant | 4 | 1 | 5 | 3 | 2 | CPB funding mattered most in exactly the markets (rural/tribal, up to 96% revenue share) this would restore. But CPB is this project's one completed real-world durability test — formula-based, multi-year-advance-funded, 58 years old — and it still dissolved on one bare-majority rescission vote. B=1 reflects a completed failed test, not a design-category assumption — phase0-findings §3 |
| 4b | Public media re-funding, non-appropriated variant (dedicated fee/endowment) | 4 | 3* | 5‡ | 3 | 3 | Same coverage case as #4. B revised down from an earlier draft's 4 to 3, post-red-team: Finland's press subsidy had the same design features (formula-based, arm's-length) and was still cut ~96% from peak; the BBC's licence fee — the one concrete "outside appropriations" precedent found — is itself under an unresolved Charter review. No project-sourced evidence of a non-appropriated public-media mechanism with a completed, successful durability test exists [Verification Phase 2, 2026-08-10: this negative claim is false. Germany's Rundfunkbeitrag is one — see ‡ below and ws10.] — ws10 |
| 5 | Demand-side vouchers/tax credit for subscriptions & donations | 3 | 1 | 3 | 3 | 3 | B revised down from an earlier draft's 5 to 3, post-red-team: citizen-direction is a C-axis (distribution) feature, not a B-axis (durability) one — the authorizing program remains as repealable as any tax credit (NY's own credit needs reauthorization by 2027). Two independent randomized free-subscription offers produced minimal activation and no detectable local-news/civic effect. They do not show where payments would land, so the incumbent-versus-desert allocation claim remains unmeasured. §14 |
| 6 | State/municipal civic-info consortia (NJ model scaled) | 3 | 2 | 4 | 3 | 3 | NJCIC funds hyperlocal/ethnic/nonprofit outlets by design, but its own funding was just cut 60% year-over-year (2.5M→1M) — the same appropriation-dependence problem as CPB, at smaller stakes — ws02-08 |
| 7 | Philanthropy-matching fund | 3 | 2 | 3 | 3 | 3 | Press Forward + AJP combined are ~55x short of just one year's revenue shortfall vs. the 2005 peak — cannot reach material scale regardless of design quality — ws02-08 |
| 8 | Structural platform remedies (US v. Google outcomes, rebadged) | 3 | 3 | 3 | 3 | 3 | D revised down from an earlier draft's 5 to 3, post-red-team: no workstream in this project independently verified the actual current status or strength of the DOJ's ad-tech remedies — the original citation (ws05) covers broadcast/newspaper ownership, a different market, and does not support this claim. D=3 reflects a plausible, undocumented-within-this-project effect, not a verified one. B similarly revised down from 5 to 3: court-driven remedies are not immune to settlement modification or a change in enforcement priorities — no multi-decade track record exists for this specific remedy type. This architecture needs its own dedicated workstream before being scored with confidence |
| 9 | Managed transition (do-nothing comparator) | 1 | 5 | 2 | 1 | 1 | Current trajectory: -78% newspaper industry employment since 2001 and continuing; the FCC's Aug 6 2026 vote loosens the ownership cap — the baseline is deteriorating, not static — phase0-findings §7, ws05 |
| 10 | Cash (universal comparator, per M7) | 1 | 5 | 2 | 2 | 2 | No journalism-specific targeting mechanism; the demand-side diffusion problem applies even more strongly than to vouchers |
[Verification Phase 1, 2026-08-10 — flagged, not re-scored.] * 4b's B=3 does not survive the axis's own anchor. B reads "1 = no demonstrated track record of surviving a hostile majority"; "3 = has meaningfully outlasted normal political cycles." The blind re-score applied that rule hard — "a brand-new instrument cannot score 3" — moving #1 to 1 ("no track record exists"), #5 to 1 ("never implemented anywhere") and #7 to 2. A non-appropriated US public-media mechanism has never been implemented either, and this row's own basis text concedes it. B=3 was reached by reasoning down from a higher score via the Finland cap rather than up from the floor — the same design-label leak the re-score existed to catch, surviving on the row the filing's headline rests on. At B=1 the neutral-weight mean falls 3.6 → 3.2; 4b stays top (next best 3.0) but "clear" loses most of its margin. Phase 1 does not re-score: M6 requires a blinded scorer. Referred to the next re-score and to Phase 2's S5 sensitivity.
[Verification Phase 1, 2026-08-10 — coverage gap.] "Rankings computed under five objective weightings" cannot be checked. No weighting vector is defined anywhere in this record except pluralism-first (A15/B10/C60/D5/E10) in the red-team log, and no per-weighting rankings exist in this file or the re-score log — both report only a neutral-weight mean. The red-team log's "The recomputed rankings are in the revised scorecard" is not satisfied. Either commit the five vectors and their rankings, or stop claiming five weightings. The public pages have been changed to the neutral-weight claim in the meantime.
[Verification Phase 1, 2026-08-10 — publication gap.] Row 10 (Cash, the M7 universal comparator) is scored here but was absent from the published whitepaper table, which showed ten of these eleven rows without saying one was omitted. Fixed on the public page.
[Verification Phase 2, 2026-08-10 — flagged, not re-scored.] † Row 2's four 1-scores do not survive their own evidence base. The steelman pass (steelman-log.md, target B) reached three primary documents none of which is in this record: the Australian Treasury's statutory review of the Code (Dec 2022), Canada's Online News Act Application and Exemption Regulations SOR/2023-276, and CRTC exemption decision 2024-262.
- A=1 ("no capacity added, or added only to already-strong outlets") is contradicted by Treasury's own review: the ABC "appointed 57 regional positions, including reporters in 19 locations, 10 of which did not previously have them"; Guardian Australia grew its newsroom "by over 40 journalists." Rule-consistent score 3.
- C=1 and D=1 rest on the ~90%-to-three-incumbents figure, which Phase 1 found is cited to a paper that does not exist and was never read — and which Treasury states it could not audit at all: "the review has not been provided with the details of these agreements by either digital platforms or the relevant news businesses." Against that, SOR/2023-276 s.10(2)–(3) mandates distribution by full-time-equivalent news-producing employees with hard caps (≤30% to broadcasters other than CBC, ≤7% to CBC), s.7 requires a majority of the money to go to local/regional/national news production, and s.4(1)(a) requires an open call — under which Google "received responses from more than 600 news businesses representing 1435 news outlets" and "did not reject any news business that submitted an attestation" (CRTC 2024-262). Both cells are the below-neutral-without-evidence pattern this scorecard's re-score corrected in five other places; rule-consistent score 3 each.
- B=1 ("Meta's walk-away is a completed failed real-world test") records half a test. Google did not walk away: 23 Australian agreements, and in Canada a regulator-set, inflation-indexed $100M/yr under a five-year exemption order. Rule-consistent score 2.
Rescored 3/2/3/3/3, row 2's neutral mean is 2.80 (conservative variant A3/B1/C3/D2/E3: 2.40) against the published 1.40. No ordering above it changes, but the published claim that this is "the one architecture the record actively refutes on four of five axes" does not survive either variant. Phase 2 does not re-score: M6 requires a blinded scorer. Referred to the next re-score.
[Verification Phase 2, 2026-08-10 — the load-bearing cell is C, not B.] ‡ 4b's C=5 is inherited from the architecture that dissolved. The re-score set it on "distribution to ~1,500 stations including the highest-dependency tribal stations is anchor-5 'demonstrated in practice'." That practice is the appropriated variant's (#4). CPB's own record (9 Oct 2025) states it "has been forced to reduce its staff by 70 percent" and that "the federal partnership that sustains 1,500 local public media stations will end," with NJ PBS and WPSU ceasing operations in 2026 and pooled interconnection, alerting and royalty infrastructure ending. A dedicated-fee mechanism inherits the design, not the demonstration.
Sensitivity, computed against this matrix (neutral weights, and the pluralism-first vector A15/B10/C60/D5/E10 — the only weighting defined anywhere in this record):
| Reading of 4b | Row | Neutral | Pluralism-first | 4b rank |
|---|---|---|---|---|
| As published | 4/3/5/3/3 | 3.60 | 4.35 | 1st |
| B→1 (Phase 1) | 4/1/5/3/3 | 3.20 | 4.15 | 1st |
| B→5 (Germany counts as Finland does) | 4/5/5/3/3 | 4.00 | 4.55 | 1st |
| C→3 (record not inherited) | 4/3/3/3/3 | 3.20 | 3.15 | 1st neutral, 4th pluralism-first |
| B→1 and C→3 | 4/1/3/3/3 | 2.80 | 2.95 | 5th on both |
4b's rank is robust to every single-cell change. It loses the lead only when both of its above-neutral cells are read as inherited rather than demonstrated — at which point #4, #6 and #8 tie at 3.00 on neutral weights and #4 leads at 4.05 under pluralism-first. Phase 1 named B as the soft cell; the arithmetic says C is.
[Verification Phase 2, 2026-08-10 — one half of the weightings gap closed.] The pluralism-first ranking the red-team log said was "in the revised scorecard" and Phase 1 could not find is computed here: 4b 4.35 · 4 4.05 · 6 3.50 · 1 3.40 · 8 3.00 · 3 2.90 · 7 2.90 · 5 2.80 · 10 2.15 · 9 2.00 · 2 1.20. It does not change the leader — that claim was true of the pre-re-score board, where #8 led, and was carried forward unrevised. The other four weightings remain undefined and the "5 weightings" claim remains UNVERIFIABLE.
Rankings, re-scored (full reconciliation: ws11-rescore-log.md)
Under the reconciled matrix, the non-appropriated public-media variant (4b) is the clear most-consistent performer — neutral-weight mean 3.6, with nothing else above 3.0 — a strengthened version of the post-red-team finding, now resting on corrected cells (its C rose to 5 on the ~1,500-station tribal/rural distribution evidence; its B stays capped at 3 by the Finland counter-example, a cap both scorers reached independently). The bargaining code (2) is the record's one affirmatively multi-axis-refuted architecture — 1s on four of five axes, each resting on peer-reviewed or platform-measured evidence rather than absence of evidence: not merely last, but the only instrument the record actively argues against. The comparators are no longer filler: do-nothing's 1s on supply, competitiveness and trust are verified-trajectory cells (accelerating consolidation; the two-country polarization mechanism operating on the more-trusted local tier). 17 of 55 cells are 3-by-hard-rule — most of this scorecard's differentiation is evidence-density, the same meta-finding as the same-day GBMT-2 and GBMT-8 re-scores.
Hybrid architectures — not scored, flagged for the next pass
The single-instrument, mutually-exclusive framing above misses at least two combinations this project's own evidence suggests would outperform any single entry, but which this pass did not evidence carefully enough to assign numbers to rather than assert qualitatively:
- #4b + #6 (a non-appropriated national funding floor distributed through existing local nonprofit/ethnic-outlet channels like NJCIC's) plausibly inherits #4b's coverage/diversity strength and #6's demonstrated local-distribution capacity, while diluting neither's specific weakness (#4b's newness, #6's small scale).
- #8 + #4b (antitrust action on concentration paired with funding restoration) plausibly beats every single instrument since their weaknesses sit on different axes (#8 weak on A/coverage, #4b weak on D/market-competitiveness) — and this is also how such policy typically actually gets built (bundled packages, not either/or menus).
- #5 + an explicit desert-weighting bonus — §12 already argues for this qualitatively; it was never re-scored as its own row, which is a gap, not a resolved recommendation.
Constitutional screen (gate, applied separately from the numeric score — revised labels post-red-team)
Per phase0-findings §4, the screen is directional, not final — both marquee cases (TX HB20, FL SB7072) remain unresolved as of mid-2026. This screen does real gating work on only one candidate (#3); for the rest, §1.2's framing already excluded the content-based instrument families the screen exists to catch, before this table was built — the labels below reflect that rather than implying a uniformly rigorous independent check across all ten.
| Architecture | Screen status | Basis |
|---|---|---|
| 1 Payroll credit | SAFE (settled) | Content-neutral eligibility, closest analog to content-neutral distinctions already upheld (Leathers) |
| 2 Bargaining code | CONTESTED | Untested US design; a domestic arbitration-forcing mechanism has not been litigated |
| 3 Ad tax earmarked | CONTESTED, elevated risk | Maryland's non-earmarked analog already lost its pass-through provision on First Amendment grounds; a journalism-specific earmark invites a stronger content-based-purpose challenge — the one candidate this screen actually filters |
| 4 Public media (appropriation-funded) | SAFE (settled) | The pre-existing, 58-year-tested design; no new constitutional exposure versus reinstating what already existed |
| 4b Public media (non-appropriated) | SAFE, directional/untested | A genuinely novel funding mechanism, not covered by #4's precedent — does not automatically inherit #4's safety credential and has not itself been litigated |
| 5 Vouchers | SAFE (settled) | Citizen-directed, viewpoint-neutral by construction |
| 6 State consortia | SAFE (settled) | Already operating (NJCIC); no novel constitutional question |
| 7 Philanthropy match | SAFE (settled) | Voluntary, no compelled-speech or taxation issue |
| 8 Structural platform remedies | SAFE (settled) | Ordinary antitrust enforcement; no First Amendment content issue |
| 9 Managed transition | SAFE (trivially) | No government action, no exposure |
| 10 Cash | SAFE (settled) | Orthogonal to media policy entirely |
Independent re-score note
Completed 2026-08-04 — see ws11-rescore-log.md; the standing requirement below is retained for the record. Per M6, a genuinely independent re-score (a second pass starting from the anchored scale definitions alone, blind to this pass's cell values) is still the appropriate next step before any of these numeric ranks — even the corrected ones — are treated as final for a whitepaper. The qualitative findings (the bottom-tier stability of #9/#2; the #8-versus-project-evidence gap on ad-tech antitrust; the Finland counter-example undercutting #4b's confidence) are more robust to re-scoring noise than any specific architecture's exact numeric position, and the whitepaper should lead with those rather than with a single "winner."