Amplification & capture — claim scoreboard
What this is: BPI-style verify/debunk of the Sep 2026 Coxon weekend story and the “METR as gatekeeper” claim. Every row is a claim we scored — not a vibes essay.
Read first if you’re cold: master index + glossary · positive assets · money graph
Quick dictionary (first use)
| Term | Meaning |
|---|---|
| BPI-style | Claim → primary evidence → verdict (CONFIRMED / PARTIAL / DEBUNKED / …) |
| Layer A / Layer B | Setup/timing vs statute/invoice — always split; missing B ≠ missing A |
| SFF | Survival & Flourishing Fund — Tallinn-backed grant tables |
| AIPI vs AIPN | AIPI = SFF payee (Hack Foundation). AIPN = AI Policy Network (Wildeford’s lane). Don’t merge |
| QT | Quote-tweet |
| METR | Model Evaluation & Threat Research — eval org in Amodei’s “embedded evaluator” pitch |
| DigitalBorders | Third-party geo charts Jurvetson posted; published graphics ≠ our raw sample yet |
Epistemic lock (Mark): don’t use “not a statute yet / no invoice” to erase a confirmed setup. Score the fire first; stamp second.
Scoreboard — core claims
| Claim | Verdict | One-liner |
|---|---|---|
| J-A1 views + WSJ urgency | PARTIAL | Mega-viral real; “sparked” softer |
| J-A2/J-A3 geo % foreign | CONFIRMED as DB charts; raw rows unreproduced | Charts on Jurvetson post (not site CSV) |
| J-A4 DB methodology | PARTIAL | Documented; not case-validated |
| J-A5 foreign manufacture | PARTIAL/UNSUPPORTED | Mechanism ok; Coxon-specific no |
| P-B1 Layer A network+timing | CONFIRMED (setup) | SFF orgs + Tallinn investor + WSJ prep + early US QTs |
| P-B1 Layer B covert PR/psyop | NOT ESTABLISHED | No invoice/direction; Coxon denied 3rd parties |
| P-B2c SFF→Encode/Futures/AIPI | CONFIRMED | SFF tables |
| P-B2d Tallinn Anthropic investor | CONFIRMED | Series A |
| P-B2e Tallinn board member | DEBUNKED/PARTIAL | Observer/investor ≠ director |
| P-B2i bill caused by Coxon | DEBUNKED chronology | Bill announced Sep 3 |
| C-C1 Coxon bio | CONFIRMED | OpenAI→Anthropic→resign |
| C-C2 GV scholarship $20,159 | HIT | Good Ventures 2022; ≠ LTFF |
| C-C3 trunk org ties | PARTIAL | GV aisle; no SFF/METR/MIRI named $ to Coxon |
| R-D2 gatekeeper / meter | CONFIRMED (de facto influence) | Not “debunked” for lacking statute |
| R-D3 ARC→METR lineage | CONFIRMED | metr.org |
| R-D6 tokens/access dependency | CONFIRMED (structural) | Self-admitted; causal “bought science” open |
| R-D7 choke endpoint | CONFIRMED trajectory; NOT YET statute | Smoke→fire setup; stamp separate |
| R-D9 China-doom disclaimer | CONFIRMED | Article refuses China-funded doom as fact |
| R-D10 X vs article | CONFIRMED | Framing delta; does not kill Layer A |
Scoreboard — weekend / creator timeline
| Claim | Verdict | Layer A | Layer B |
|---|---|---|---|
| Same-brief feel | PARTIAL | CONFIRMED timed activation (Coxon→Hubinger ~83m; Marks; WSJ prep; SFF-org QT roles; shared extinction/race lexicon) | NOT ESTABLISHED (Baier denial; no boost invoice; QT order OPS_GAP) |
| Three-lab / CEO pacing lockstep | CONFIRMED (policy); PARTIAL if = same Coxon brief | Amodei Pace the Frontier → Altman “pace the frontier” + Musk “Dario is right” (+Hassabis) ~Sep 12–13 | Musk Sep 10 Coxon “setup” = messenger discount — does not take xAI out of pacing coalition (Mark recalib 2026-09-14) |
| Creator aisle activation | UNSUPPORTED | PDKU/Aella/Grimes = prior July infrastructure (Money graph), not Sep blast | NOT ESTABLISHED; secondary Aella “surprise”; archive OPS_GAP |
Weekend / creator BLUF: Layer A activation pattern for Sep 8–12 is real — do not soft-pedal. Pacing lockstep recalibrated (Mark): Musk attacked Coxon-as-person and publicly backed Amodei’s pacing ask (“Dario is right”) — messenger-attack ≠ policy divergence; three-lab (+DeepMind) pacing rhyme CONFIRMED. Directed psyop invoice still not established. Creator aisle did not light as same-window PDKU activation on primaries found.
Primary anchors: Coxon 2097476196791709843 (2026-09-09 00:04Z) → Hubinger 2097497037956891126 (~83m) → Musk 2097866303633752463 (“setup” on Coxon) → Jurvetson 9/11 → Amodei 9/12 Pace the Frontier (METR named; unilateral permanent embeds) → Musk/Altman/Hassabis pace endorse 9/12–13 → Multiplex 9/13 → Roemmele 9/14.
Scoreboard — frontier cartel / competitive pacing
| Claim | Verdict | One-liner |
|---|---|---|
| Cartel-rational pacing rhyme | CONFIRMED motive-shape (Layer A); formal cartel NOT ESTABLISHED (Layer B) | Amodei–Altman–Musk (+Hassabis) public pace = coordinated restraint among majors; xAI can price it financially without loving Coxon |
| Musk/xAI ≠ EA cult by that agreement alone | CONFIRMED separation | Policy rhyme ≠ Singham/SFF/Tallinn membership |
| Anthropic (and OpenAI-if-same-stack) disproportionately benefit | CONFIRMED structure (Sep 12 METR named + unilateral permanent embeds); equity-causal open | Amodei essay names METR; Anthropic unilaterally commits now to permanent employee-level access; METR-gate setup → lab install |
Frontier cartel BLUF: Shape of Musk’s agreement → financial/cartel strategy among frontier labs that especially helps Anthropic — not proof Elon is an EA cultist. .
Scoreboard — exact dollars
| Item | Verdict | Figure |
|---|---|---|
| GV → Coxon scholarship | HOLD | $20,159 (EIN 46-1008520); ≠ LTFF label |
| SFF → Encode AI Corp | HOLD | $516,000 (SFF-2025); Encode Justice $50k separate |
| SFF → AI Futures | HOLD | $1,535k + $500k‡ (2025); $505k (2024) |
| SFF → AIPI (Hack) | HOLD | $1,635k (2025) + $142k (2024) + IC $300k — not AIPN payee |
| OP/GV → ARC → METR | HOLD path | ARC $1,515,000; spin-off $4,477,169; direct OP→METR DEBUNK |
| Frame stipend | HOLD | $10,000 / 10 weeks |
Full table: Thesis audit.
Report polish (open)
Geo chart reproduction: Geo repro method
Gaps still open (honest)
- Org-account first-wave Coxon QTs (AIPN / AIPI / Encode /
@AI_Futures_orgs) — personal Wildeford CONFIRMED; orgs still thin - Creator X archive Sep 7–14 (Aella/Grimes/PDKU/Frame) — needed if we want to upgrade the creator-aisle row
- Nadeau “14th QT” ordinal — timing CONFIRMED; ordinal still alleged
- DigitalBorders raw sample for geo repro — method in Geo repro method; charts already CONFIRMED as published
Closed since first draft of this scoreboard: Wildeford personal QT · WSJ publish Δ · Marks status ID · GV 990-PF PDF · Lightcone ‡ face — see receipt crumbs on the research map.
Epistemic bar — smoke / fire vs statute / invoice
Locked 2026-09-14 (Mark): Do not erase a confirmed setup with pedantic “not law yet / no invoice yet.” Editorial rule from the DSC editorial: where there’s smoke, there’s fire — score the setup, not only the final stamped statute.
Two layers (always split; never collapse)
| Layer | What it is | How to score | Example |
|---|---|---|---|
| A — Setup / trajectory (the fire) | Institutions, playbooks, embeds, revolving doors, donor/advocacy adjacency being built so one meter becomes the gate | CONFIRMED when primary shows the apparatus is real | METR writes RSP templates labs adopt; NIST/UK/EU chairs; free tokens + access; EA philanthropy/personnel circuit |
| B — Final formality (the stamp) | Already written into statute as licensing choke; paid psyop invoice; foreign-% export on file | Score separately: NOT YET / UNSUPPORTED / OPS_GAP — without downgrading Layer A | METR not yet a statutory license gate; no DigitalBorders Coxon CSV; no boost invoice |
Wrong Otto move: Label R-D7 “UNSUPPORTED as statute” and let that read as “Roemmele is wrong about the choke point.”
Right Otto move: CONFIRMED — institutional choke-point setup / trajectory; NOT YET — statute.
Same split for Thayer: CONFIRMED — SFF-funded advocacy first wave + Tallinn-adjacent capital + WSJ prep (smoke→fire). NOT ESTABLISHED — covert psyop / paid boost invoice.
Ask: “Is the activation/setup confirmed?” first. Ask “Is the invoice/statute confirmed?” second. Never answer the first with the second.