← Evidence index · 中文 · Main post
Each section: Claim · Why · Evidence · Analogue · Would update if · Conf (H/M/L).
Parent: Shared Ci spine · timeline prediction nodes 1 3 expanded
Cross-cut (physical): node physical ai embodied · node physical ai
Cross-cut (politics): crosscut secondary cruxes §7 culture-war coalition
Date: 2026-07-03 (updated 2026-07-04: physical AI second wave, GUARD embodied tail, culture-war actor table)
Settings: Hybrid time (C); modal + tail branches
Purpose: Every probability and timing claim in Node 1, with evidence, analogues, and falsifiers. Feeds (internal note) concentrated-harm / whimper / coordination cruxes.
TL;DR
Node 1 is partially live on the digital layer only (Block −40%, Canaries −16% rel., Challenger AI-cited cuts 101k YTD). Capability runs fast (METR/agents on AI 2027 pace); digital policy salience runs slow (+~30%, peak ~2027 Q2). Physical AI adds a second labor wave 2028–2031 (+12–24 mo vs digital) — see node physical ai embodied. Modal (~60–70%): hearings + SB 53 + labor + culture-war vetting coalition (anti-woke AI + jobs frame), not federal training caps. Tails: Sanders moratorium surprise (3%), GAAIA preemption (12%), agent-disaster vetting law (8% software-only; 12–14% if embodied Trigger E2). Primary p(doom) channel: concentrated harm + whimper, not extinction.
Hybrid timing windows
T = 2026-09 → 2027-03 — Capability window (fast track)
Claim: Agent labor shock becomes economically binding (junior SWE / white-collar automation tasks) in this window, anchored to METR horizon growth and agent deployment, not to DC calendar.
Why: Hybrid rule (C) keeps capability on AI 2027 / METR pace. Tracker shows METR doubling Ahead (89-day doubling since 2024); agent autonomy Confirmed (6/6 predictions). AI 2027 C4 = “Agent-1-mini; junior dev labor shock” — capability spine, not yet full macro salience.
Evidence:
- (internal note) — speed ratio 0.70× overall but METR time horizon Ahead; Economic Impact mixed (stock backlash Ahead, some job targets Behind)
- METR Time Horizon 1.1 — post-2024 P50 doubling 88.6 days; frontier models at 6–12 h 50%-horizon (Opus 4.6 ~12h, GPT-5.2 ~6.5h, Feb–Mar 2026)
- (internal note) §Hybrid time rule — fast track examples: agents, METR, coding automation
- Anthropic RSI essay (2026-06-04): >80% production code AI-written — internal productivity signal at frontier lab ((internal note) §2026.06.04)
Analogue: Y2K remediation — capability/deployment deadlines hit before political institutions feel urgency.
Would update if: METR 50%-horizon stalls <4h through 2027-06 despite new frontier releases; or junior employment recovers while horizon >24h (automation → augmentation pivot).
Conf: H (capability); M (exact calendar endpoints)
T = 2027-02 → 2027-09 (peak ~2027 Q2) — Policy salience window (slow +30%)
Claim: Mass political salience (HELP hearings, state labor bills, sustained media cycle) peaks ~2027 Q2, ~30% later than AI 2027’s late-2026 job-shock / DC-protest anchor.
Why: Tracker lag on politics/economy (0.70× on governance items). Macro null (Yale) decouples from micro canaries — policymakers react to visible layoff waves + union pressure, not ADP regressions. COVID analogue: 2–3 month institutional delay even on obvious tails.
Evidence:
- (internal note) — salience window explicit
- (internal note) — “Large-scale anti-AI protest (10,000+)” = Emerging, not Confirmed; Economic Impact job displacement Ahead on market narrative but Behind on some quant targets
- (internal note) §8 — Yale occupational dissimilarity still flat; Canaries lead macro by ~1–2 years (speculative lag)
- CA Newsom EO N-6-26 (2026-05-21): first governor EO on AI workforce displacement — study/recommendations, not binding rules (gov.ca.gov PDF)
Analogue: COVID Jan–Mar 2020 Western policy lag ((internal note) §Method evidence library).
Would update if: Second Block-class F500 cut + Challenger AI-cited cuts >30% YoY in one quarter → pull salience forward 1–2 quarters (Trigger E1). Or zero hearings through 2027-12 after Block-scale events → shock normalized/absorbed.
Conf: M
T = Now (2026-07) — Node partially live
Claim: Node 1 triggers are already firing on capability/micro-employment layers; not yet on mass protest or federal binding labor law.
Why: Block, Challenger, Canaries, METR thresholds met; 10k DC protest and F500 eng-freeze wave not met.
Evidence:
- Block: 4,000+ cuts (~40% HC), Feb 27 2026, CEO letter ties to “intelligence-native” model (Fortune, VentureBeat)
- Challenger H1 2026: 101,743 AI-cited cuts (~23% of 443,604 total); AI #1 reason 4 consecutive months (Challenger Jun 2026, PDF)
- Stanford DEL Canaries: 22–25 in top AI-exposure quintile −16% relative employment vs least-exposed, firm FE (paper; (internal note) §3)
- Cloudflare ~20% cut, May 2026, explicit agentic-AI reorg (blog)
Analogue: Canaries in coal mine — micro signal before macro collapse (Brynjolfsson framing).
Would update if: Yale SDID turns significantly positive on AI exposure × unemployment; or Canaries cohort recovers.
Conf: H (partial live status)
Physical AI second wave (cross-cut X2-A)
Full node: node physical ai embodied · Evidence: node physical ai
Definition (for Node 1 scope): Physical AI = frontier models closed-loop with actuators (humanoid/logistics robots, warehouse AMRs, field drones). Same Ci capability spine as digital agents; different deployment surface (factory integration, liability, embodiment data) and policy coalition (OSHA/UAW vs HELP/white-collar).
Rule: Anchor embodied capability to Ci (VLA policies C4–C5); anchor embodied macro labor salience to Node 1 digital peak +12–24 mo unless T-PA1 or vivid embodied failure.
T = 2028–2031 — Physical labor salience window (slow +12–24 mo vs digital)
Claim: Manufacturing/warehouse political salience (OSHA incidents, UAW hearings, factory-floor displacement headlines) peaks 2028–2031, 12–24 months after digital Node 1 peak (~2027 Q2), unless T-PA1 or embodied Trigger E2 pulls forward.
Why: Factory integration timelines 12–24 mo per site; union bargaining cycles slower than tech layoff waves; verified humanoid base still ~100–500 shift-work units globally vs millions of white-collar tasks already automatable at METR 6–12h.
Evidence:
- node physical ai embodied §Hybrid timing — policy/labor salience row
- (internal note) — Agility 100k+ totes, Figure 1,250 hr @ BMW; single-digit robots per line
- Robonaissance deployment gap — ~13k shipped, ~100 sustained productive globally
- (internal note) — digital partial-live triggers already met; physical KPIs Behind AI 2027 factory-automation beats ((internal note))
Analogue: ATM adoption — bank-teller political salience lagged technology by decade; factory robots may lag digital agents by 2–4 yr not decade because visibility is higher.
Would update if: ≥3 OEMs each >500 shift-work units with public KPIs by 2028-H1 (→ T-PA1; pull salience to 2027–28). Or zero new industrial humanoid KPI disclosures 2027–2028 (→ push physical wave to 2030+).
Conf: M
P = 0.65 — Digital Node 1 peak precedes physical labor coalition by ≥12 mo
Claim: Node 1 digital policy peak (~2027 Q2) precedes manufacturing/UAW mass mobilization by ≥12 months in modal path.
Why: White-collar cuts (Block, Canaries, Challenger) already live; factory humanoids still pilot; AFL-CIO Workers First frames general AI displacement before embodied-specific campaigns.
Evidence:
- node physical ai §P=0.65
- node physical ai embodied modal table — P(digital peak before physical) 0.65
- Yale macro null on national unemployment — physical displacement not yet aggregate-binding
Analogue: Offshoring 2000s — white-collar service shock headlines preceded manufacturing automation political cycles.
Would update if: UAW or AFL-CIO embodied-AI campaign launches 2026-H2 with major strike threat tied to humanoid deployment.
Conf: M
P = 0.75 — Humanoids remain pilot-scale through 2029
Claim: <5,000 humanoids worldwide doing paid multi-shift industrial work by 2029-12 — physical AI real in pilots, not macro economically binding.
Why: Only Agility + Figure have verified shift KPIs; integration bottleneck; Unitree volume = research/demos not shift-work KPIs; Tesla Optimus still data-collection-heavy (Musk Jan 2026: zero useful factory work).
Evidence:
- node physical ai embodied §Verified deployment baseline
- Figure F02: 1 robot, 1,250 hrs over 11 mo @ BMW — not fleet scale
- Agility Digit: 100k+ totes @ GXO but ~40–150 units global installed base (est.)
Analogue: Autonomous trucking — years between “remove safety driver in pilot” and “macro freight displacement.”
Would update if: Two additional OEMs publish >200 shift-work units each with KPIs by 2028-H1.
Conf: M–H
P = 0.12 — T-PA1 embodied OEM scale surprise (merges digital + blue-collar waves)
Claim: ~12% tail: ≥3 OEMs each >500 paid shift-work humanoids by 2028 + replicated throughput KPIs (tote/parts/hr) → pulls physical labor salience forward to 2027–2028; AFL-CIO + UAW frame merges with digital shock.
Why: BotQ 12k/yr nameplate + Unitree 10–20k 2026 target + Hyundai 25k/yr roadmap — capacity exists; tail requires simultaneous customer pull + integration success across OEMs.
Evidence:
- node physical ai embodied §T-PA1
- Figure BotQ ramp — 12,000/yr line capacity
- Unitree IPO narrative: 10k–20k 2026 shipment target; Hyundai Atlas 25k/yr roadmap (JPM May 2026)
Analogue: iPhone ramp — supply chain exceeds expectations once product-market fit hits; humanoid PMF unproven but capital committed.
Would update if: 2027 shows zero fleet expansion beyond GXO/Figure/BMW — collapse tail to <5%.
Conf: L–M
P = 0.10 — T-PA2 physical catastrophe → embodied Trigger E2 (feeds GUARD tail)
Claim: ~10% tail: humanoid/AMR death or >$100M logistics loss with clear AI causation chain → satisfies embodied Trigger E2 → raises Node 1 T3 vetting from 8% → 12–14% (increment +4–6pp vs software-only).
Why: Injury/death more legible than API agent errors; Boeing 737 MAX / Uber AV 2018 precedents; Hawley GUARD + Blackburn RESPONSIBLE frame pre-deployment vetting, not training pause; warehouse humanoid > chat agent for media cycle duration.
Evidence:
- node physical ai embodied §T-PA2 — Node 1 T3 8% → ~14% conditional
- node physical ai §T-PA2 — +4–6pp increment
- Uber AV pedestrian death (2018) — immediate regulatory response vs Knight Capital 2012 software-only
- (internal note) §2.3 — GUARD Act Senate Judiciary watch list
Analogue: 737 MAX — single vivid kinetic failure → binding certification changes; software agent losses rarely achieve same.
Would update if: Major software-only agent loss >$100M passes vetting bill without physical component (→ embodied increment overstated).
Conf: L–M
P = 0.40 — UAW / AFL-CIO embodied-AI campaign by 2028-H2
Claim: ~40% major US manufacturing union launches public embodied-AI campaign (humanoid deployment, line surveillance, injury liability) by 2028 H2 — distinct from digital Workers First frame.
Why: UAW has factory jurisdiction; AFL-CIO Tech Institute brief already covers surveillance/deskilling; physical robots visible on factory floor → stronger populist regulation appetite than API automation (node physical ai embodied §p(doom)).
Evidence:
- AFL-CIO Workers First principles — guardrails, collective bargaining on deployment (report)
- node physical ai embodied — physical AI visible → populist regulation
- No embodied-specific UAW strike threat yet (2026-07) — lag supports +12–24 mo rule
Analogue: Auto industry robotics fights 1980s–90s — slow union cycle, high salience when visible.
Would update if: UAW 2027 contract negotiations explicitly exclude embodied AI (→ downgrade to <20%).
Conf: L–M
P = 0.80 — No federal embodied-AI licensing separate from software agents (modal)
Claim: ~80% no separate federal embodied-AI training/licensing regime through Node 1 + physical window — policy stays OSHA/state/factory lanes.
Why: Natsec humanoid race (Figure $39B, Tesla) + anti-pause default; no embodied-AI moratorium in playbook Tier 1; battlefield autonomy separate frame.
Evidence:
- node physical ai embodied modal — P(no federal embodied licensing) 0.80
- Playbook §2.3 Tier 1 — GUARD/RESPONSIBLE are deployment vetting, not embodied training caps
Analogue: Industrial robot safety (ISO 13849) — voluntary/consortia lag, not federal AI licensing.
Would update if: Federal embodied-AI licensing passes without major incident (policy faster than model — falsifier).
Conf: M
Node 1 × Physical AI interaction table
| Effect on Node 1 | Direction | Magnitude | Source |
|---|---|---|---|
| Peak salience timing | Split or extend | Digital unchanged (~2027 Q2); physical +12–24 mo unless T-PA1 | node physical ai embodied |
| Federal training cap | ↓ | Physical “robots take jobs” → economic menu; cap stays <5% | Same |
| GUARD / vetting (T3) | ↑ | Embodied failure → +4–6pp on T3 (8% → 12–14%) | §T-PA2 above |
| Concentrated harm | ↑↑ | Injury liability, factory deskilling, wearables surveillance | §p(doom) below |
| Whimper | ↑ modest | Physical + digital marginalization compound C8–C9 | Same |
| Extinction | ~0 direct | Unless battlefield LAWS spillover (Node 3) | Same |
Key crux: Physical AI widens concentrated harm without pulling forward DC digital peak — unless vivid embodied failure satisfies Trigger E2 (>$100M / enforceable judgment / fatality with clear AI chain).
Culture-war coalition — vetting not pause (cross-cut §7)
Full crux: crosscut secondary cruxes §7
Playbook: (internal note) §1.4, §3.1
Claim: US AI politics fuses populist-right anti-”woke AI” (xAI/Musk, Cruz/Banks lane) with labor/creative anti-displacement coalitions. Equilibrium policy = vetting, licensing, transparency, child safety, deepfake bans — explicitly not training pause or compute caps. Constrains Node 4 tail-gov and FLI-class advocacy; conditions Node 1 policy menu.
P = 0.62 — Vetting coalition modal wins (2026–2028)
Claim: ~62% modal: BMIA + SB 53 + NO FAKES + GUARD-class vetting pass or advance; no federal training pause or compute cap.
Why: Cruz moratorium stripped 99–1; Encode SB 53 path; Hawley GUARD unanimous Judiciary Apr 2026; 55% oppose regulation moratorium (2025); federal mandatory pause dead (playbook §1.2).
Evidence:
- crosscut secondary cruxes §7 — P(vetting modal) 0.62
- Blackburn 99–1 killed Cruz preemption — culture-war carve-outs (ELVIS, KOSA, NO FAKES) preserved (
entity_master_list.mdBlackburn entry) - (internal note) §5 Tier 1 — “labor + deepfake + natsec on same legislative package”
Analogue: Post-Cambridge Analytica — hearings + sector rules, not industry shutdown.
Would update if: Binding federal training moratorium signed (falsifies coalition claim).
Conf: M
P = 0.08 — Populist moratorium floor vote (Cruz-class)
Claim: ~8% tail: populist-right moratorium gets floor vote or passes one chamber — overlaps Node 1 T1 Sanders tail but different coalition (R-led preemption fight, not HELP progressive).
Why: Cruz BBB moratorium existed; stripped; Altman endorsed uniformity at Cruz hearing; still in Overton window as adversarial text.
Evidence:
- Node 1 §T1 P=0.03 Sanders — lower than Cruz-class because R Senate blocks Sanders; Cruz-class higher as negotiation lever
- crosscut secondary cruxes §7 — P(moratorium floor) 0.08; Node 4 tail-overreaction 0.06
Analogue: Repeal ACA floor votes — symbolic force, rarely law.
Conf: L
P = 0.20 — Anti-regulation fusion blocks even vetting (GAAIA O1 tail)
Claim: ~20% culture-war + industry fusion passes broad preemption without ELVIS/KOSA/NO FAKES carve-outs → blocks state vetting path (Encode/Wiener defensive war lost).
Why: Trump EO Jun 2026 voluntary acceleration; GAAIA 3-year preemption; $8.5M Q1 2026 preemption lobbying; Blackburn TRUMP AMERICA AI Act is selective preemption — blanket Cruz-class revival is the tail.
Evidence:
- crosscut secondary cruxes §7 — P(anti-regulation tail) 0.20
- Node 1 §T2 GAAIA preemption P=0.12 — subset of this tail
- Playbook §2.1 — preemption = highest negative risk
Analogue: FCC net neutrality preemption — federal uniformity kills state labs.
Would update if: TRUMP AMERICA AI Act passes with child-safety/deepfake carve-outs intact (→ collapse to <10%).
Conf: M
P = 0.10 — Cross-ideological pause coalition tail
Claim: ~10% C10 capability + live incident (Node 4 Trigger E3) sustains cross-ideological halt >60 days — falsifies “vetting not pause” equilibrium.
Why: Requires alignment scare + labor + natsec rare alignment; historical: board crisis, Leike, Saunders → no halt.
Evidence:
- crosscut secondary cruxes §7 — P(pause coalition) 0.10
- Node 4 P(halt | whistleblower modal) ≈ 0.08
Conf: L
P = 0.30 — Labor + populist-right coalition on vetting (explicit bridge)
Claim: ~30% sustained legislative package bridging AFL-CIO/labor + Hawley/Blackburn vetting (not pause) — e.g. displacement notice + child-safety/deployment vetting in same bill.
Why: Playbook “weird but workable coalition”; Cruz 99–1 showed populist-right + state-rights + labor-adjacent opposition to blanket preemption; GUARD passed Judiciary unanimously.
Evidence:
- crosscut secondary cruxes §7 — generalizes Node 4 actor row P(coalition w/ labor/populist right on vetting)=0.30
- Playbook §3.1 — Bannon/Hawley killed Cruz moratorium; GUARD Act
- Playbook §1.4 — winnable coalition → pre-deployment vetting, not moratorium
Analogue: Strange-bedfellows copyright fights (musicians + tech critics) — ELVIS Act precedent.
Would update if: Hawley/Encode co-sponsor joint workforce + vetting bill with committee hearing.
Conf: M
Culture-war coalition — expanded actor table
| Actor | Node 1 role | Key P(modal) | Conf | Notes |
|---|---|---|---|---|
| Musk / xAI | Anti-”woke AI” accelerant | P(jobs-frame dominates x-risk) 0.85 | M | ”Truth-seeking” vs Anthropic safeguards; Pentagon Grok Feb 2026; Optimus PR — (internal note) |
| Sen. Josh Hawley (R-MO) | GUARD / child-safety vetting | P(GUARD advances) 0.70 | M | Bipartisan w/ Blumenthal; Judiciary unanimous Apr 2026; opposes Trump preemption EO |
| Sen. Marsha Blackburn (R-TN) | Selective preemption gatekeeper | P(carve-outs hold) 0.65 | M | Killed Cruz 99–1; TRUMP AMERICA AI Act w/ ELVIS/KOSA/NO FAKES; not frontier catastrophic-risk lane |
| Sen. Ted Cruz (R-TX) | Anti-state-reg uniformity | P(broad preemption retry) 0.25 | M | Commerce chair; industry-friendly; not pause advocate |
| Sen. Jim Banks (R-IN) | Natsec visibility, not pause | P(RSI eval mandate) 0.55 | M | Jun 2026 letter — CAISI visibility, not veto; correlates anti-pause cluster |
| Encode / Wiener | State transparency path | P(SB 53 stands) 0.85 | H | SB 53 signed; opposes GAAIA preemption; vetting not halt |
| AFL-CIO Workers First | Labor shock partner | P(sustains agenda) 0.75 | M | Oct 2025 launch; Apr 2026 Sanders meeting; jobs not x-risk frame |
| Sen. Bernie Sanders (I-VT) | Progressive labor + infra | P(S.4214 floor fight) 0.15 | L | Datacenter moratorium bundles jobs + safety; GovTrack ~2% enactment |
| Trump admin / EO 2026 | Acceleration default | P(voluntary EO wins over GAAIA) 0.25 | M | Jun 2026 EO — voluntary frontier review, not mandatory licensing |
| Big Tech / a16z | Block pause | P(federal cap) <0.05 | H | Preemption lobbying $8.5M Q1 2026 |
| Natsec hawks | Strong anti-pause | P(training halt) <0.03 | H | Beat-China frame; DoD AI spend Confirmed scaling |
| AI Now / labor-adjacent | Near-term harm bridge | P(coalition w/ Encode) 0.40 | M | Surveillance, deskilling — overlaps Workers First |
p(doom) direction (culture-war overlay): Vetting modal → ↓ bio (BMIA ↑); ↑ race continues (no pause) — net ±1–3pp on ~17% doom region (emergent sim) (crosscut secondary cruxes §7). Anti-regulation tail → +2–3pp via P(no pause).
Observable triggers (capability & salience thresholds)
P(trigger met) ≈ 1.0 — METR 50%-horizon ≥ 8 h sustained
Claim: Frontier agent task horizon ≥8 hours (50% success) is already met across multiple 2026 frontier releases — sufficient for meaningful junior-dev / ops task classes.
Why: METR TH1.1 puts Opus 4.6 at ~12h, GPT-5.2 at ~6.5h, Gemini 3.1 Pro ~5.8h (50%-horizon). Threshold used as Node 1 “agents economically relevant” gate.
Evidence:
- METR Time Horizon 1.1 — model table Feb–Mar 2026
- (internal note) — “METR time horizon doubles every ~4 months” = Ahead
- Wikipedia METR summary with Apr 2026 Mythos preview ~16h+ (link)
Analogue: Moore’s-law style capability curves — policy always lags measured doubling.
Would update if: METR retracts TH1.1 estimates downward >35% on recent models (methodology revision).
Conf: H
P(met by 2027-Q1) ≈ 0.65 — METR 50%-horizon ≥ 24 h
Claim: ~3× current horizon (~24h tasks at 50% reliability) likely by 2027 Q1 on fast track — junior SWE week-fraction automation class.
Why: Extrapolate 89-day doubling from ~12h baseline → ~24h in one doubling (~3 months) to ~48h in two. AI 2027 junior-dev shock assumes multi-hour → day-scale tasks.
Evidence:
- METR TH1.1: 88.6-day doubling since 2024
- (internal note) — falsifier: “junior employment recovers while METR >24h” would break automation link
- AI 2027 C4 anchor: late 2026 junior labor shock (capability, not policy)
Analogue: Autonomous vehicle miles — capability milestones precede regulatory response by years.
Would update if: Doubling time reverts to 7-month TH1.0 rate; or agent evals show plateau on software tasks.
Conf: M (extrapolation)
P(met) ≈ 0.95 — Canaries junior cohort −10% relative (threshold −16% observed)
Claim: 22–25 year-olds in highest AI-exposure occupations show ≥10% relative employment decline vs least-exposed quintile (firm fixed effects) — already met at ~−16%.
Why: ADP microdata through 2025-09; automation-heavy tasks, not augmentation-heavy; holds excluding tech firms.
Evidence:
- Brynjolfsson, Chandar, Chen (2025/2026): Canaries in the Coal Mine — Fact 4: ~−16% relative
- (internal note) §3 — six facts table; Yale macro null coexists (§2)
- Yale Budget Lab — no national occ-level AI unemployment signal (33 mo post-ChatGPT)
Analogue: Entry-level offshoring 2000s — narrow cohort hurt before aggregate stats move.
Would update if: ADP update through 2026 Q2 shows recovery to −5% relative or less.
Conf: H
P(met) ≈ 0.90 — Challenger: AI = #1 layoff reason ≥2 consecutive quarters
Claim: AI cited as leading stated reason for US job cuts for ≥2 consecutive quarters — met (4 months through Jun 2026).
Why: May 2026: 38,579 AI-cited (40% of monthly cuts, record); Jun 2026: 14,029 (31%). YTD 101,743 vs 54,836 all of 2025.
Evidence:
- Challenger May 2026 PDF — AI 40% of May cuts
- Challenger Jun 2026 — 4th consecutive month AI #1
- CNBC Jun 2026 — Glassdoor economist caution: stated reason ≠ true cause
Analogue: “Restructuring” label in 2008 — disclosure category lags reality but shapes narrative.
Would update if: AI-cited share <10% for 2 consecutive quarters while agents scale (AI-washing thesis).
Conf: M–H (measurement is disclosed reason, not causal proof)
P(partial) ≈ 0.55 — Fortune 500 ≥15% eng/tech cut explicitly AI-driven
Claim: At least one Fortune 500 engineering/tech headcount reduction ≥15% with explicit AI attribution — partially met (Block ~40% but fintech not classic F500 eng; Cloudflare ~20%).
Why: Block is S&P 500 (XYZ) but narrative is “intelligence-native” fintech, not FAANG eng freeze. No wave of MSFT/GOOG/META eng cuts at 15%+ explicitly AI-labeled yet (Meta 2026 cuts cite AI investment + overhire).
Evidence:
- Block 40% (Fortune); NBER/Oxford cited in Forbes: many CEO AI layoffs = pandemic overhire correction (Forbes Apr 2026)
- Cloudflare >1,100 (~20%): blog
- (internal note) §4 — Meta ~8k cuts, AI + overhire mixed narrative
Analogue: IBM 1990s workforce restructurings — headline % cuts vary in true AI/automation content.
Would update if: ≥3 F500 tech firms announce ≥15% eng cuts with AI as primary stated driver in same quarter.
Conf: M
P(not yet) ≈ 0.88 — Public protest ≥10,000 (anti-AI / x-risk)
Claim: No 10,000+ DC anti-AI march yet; slow-track places this 2027 H1–H2 if at all.
Why: Tracker governance: large protest = Emerging. Third Act / data-center actions smaller. x-risk frame lacks labor-scale organizing.
Evidence:
- (internal note) — “Large-scale anti-AI protest (10,000+)” Late 2026 = Emerging
- FLI pause letter Mar 2023: massive media, zero mass street movement ((internal note) §Executive summary)
- Seismic UK 2025 — 40% want stop development; not organized US veto power
Analogue: Occupy Wall Street — economic anger ≠ sustained 10k+ single-issue march without union/NGO infrastructure.
Would update if: ≥10k DC march before 2027-06 with >2 mo media cycle (falsifier for slow-track timing).
Conf: L–M
Modal branch
P = 0.60–0.70 — Modal path branch mass
Claim: ~60–70% of Node 1 worlds follow modal path: transparency + labor hearings + state disclosure bills; no federal training cap; economic coalition beats x-risk coalition.
Why: SB 1047 → SB 53 lesson; Cruz 99–1; Trump EO 2026 voluntary not mandatory; labor + Encode + populist right align on vetting/preemption fights, not stop training (Stix & Maas bridge partial success).
Evidence:
- (internal note) — modal definition
- (internal note) §1.4 — winnable coalition = state transparency/licensing, not moratorium
- (internal note) §2 — SB 53 signed; SB 1047 vetoed
- Saunders 2024 testimony → hearings, no pause ((internal note) Node 4 analogues)
Analogue: Post-2008 financial reform — hearings + disclosure + sector stress; no structural break of core industry.
Would update if: Federal training moratorium signed; or zero hearings through 2027-12 post-Block.
Conf: M
Modal actor table
P = 0.75 — US Congress AI & jobs hearings (by Node 1 peak)
Claim: Senate HELP or relevant committees hold AI + workforce oversight hearings during Node 1 salience window (~2027 Q2 peak).
Why: Sanders chairs HELP; AFL-CIO Apr 2026 meeting; 2023–25 established AI hearing pattern; Saunders 2024 precedent (transparency not pause).
Evidence:
- (internal note) §1.4 — labor + populist right coalition table
- AFL-CIO Workers First launch Oct 2025
- S.4214 rollout Mar 2026 — Sanders forcing AI-labor-data-center nexus (Roll Call)
- (internal note) Node 4 — P(hearings within 90d | whistleblower) = 0.75 same reference class
Analogue: Facebook/Cambridge Analytica hearings 2018 — scandal → hearings, limited binding law.
Would update if: Zero HELP/jurisdiction AI-labor hearings through 2027-12 after Block-scale events.
Conf: M
P < 0.05 — Federal frontier training cap / licensing
Claim: <5% chance of binding federal training FLOP cap or licensing motivated primarily by Node 1 labor shock (distinct from Node 4 alignment scare).
Why: Natsec hawks + Big Tech block; Cruz moratorium stripped 99–1; Trump EO Jun 2026 explicitly not mandatory licensing; Anthropic conditional pause requires verification that doesn’t exist.
Evidence:
- (internal note) §1.2 — Cruz BBB stripped; Trump EO voluntary 30-day review
- (internal note) §3.4 — GAAIA has audits/reporting, not training cap
- GovTrack S.4214: ~2% enactment probability (GovTrack)
Analogue: Post-Sandy Hook gun legislation — salient harm, structurally blocked federal cap.
Would update if: Binding federal training cap passes and is signed (falsifier for entire Node 1 modal).
Conf: H
P = 0.40 — US executive “AI workforce transition” EO (voluntary, pro-growth)
Claim: ~40% Trump-admin (or successor) issues workforce-transition EO/OSTP initiative — reskilling rhetoric, voluntary employer guidance — without training halt.
Why: Deregulation default but political need to “do something” on jobs; CA Newsom N-6-26 creates template; could be mirrored or preempted federally with lighter touch.
Evidence:
- CA EO N-6-26 (2026-05-21): WARN review, severance study, AI dashboard — no private employer mandates (Fisher Phillips, signed PDF)
- Pending CA SB 951 — AI-specific WARN-style notice (Kaufman Dolowich)
- Trump EO 2026-06-02: voluntary frontier review, acceleration frame ((internal note) §3.4)
Analogue: Obama “Skillful” / Trump Pledge to America’s Workers — executive workforce branding without structural labor market intervention.
Would update if: Federal mandatory AI-displacement WARN passes; or explicit admin statement ruling out any workforce EO.
Conf: M
P = 0.30 — Bipartisan workforce-transition / reskilling bill (no compute cap)
Claim: ~30% Congress passes narrow workforce bill (WIOA expansion, displacement notice, tax credits) — no compute cap attached.
Why: Stix & Maas short-harm bridge is politically tractable; Sanders S.4214 bundles worker protection with datacenter moratorium (harder pass); historical precedent for retraining bills without industry caps.
Evidence:
- (internal note) §1.4 — bridging short ↔ long harm = most tractable political problem
- S.4214 worker-protection preconditions for lifting datacenter moratorium (Congress.gov PDF)
- Pew: 64% public expects fewer jobs over 20 years — demand-side pressure (Pew Apr 2025)
Analogue: Trade Adjustment Assistance — compensation politics without stopping trade/tech.
Would update if: 119th Congress adjourns with zero workforce AI bills out of committee despite sustained Challenger AI-cited cuts.
Conf: L–M
P = 0.20 — GAAIA enacted (frontier reporting + IVO audits + preemption)
Claim: ~20% Great American AI Act (Obernolte–Trahan discussion draft) becomes law with frontier audits, incident reporting, CAISI codification — includes 3-year state development preemption fight.
Why: Strongest federal safety text yet (Jun 2026) but not introduced; preemption repeats Cruz dynamic; OpenAI aligned with lighter Trump EO.
Evidence:
- Trahan press release Jun 4 2026 — discussion draft, feedback to GAAIA@mail.house.gov
- (internal note) §3.4 — GAAIA vs Trump EO table; preemption vs SB 53
- ArentFox Schiff analysis — 269 pages; workforce + frontier oversight combined
- House Democratic Commission statement: draft “does not meet the enormity of the moment” (Subject to Inquiry)
Analogue: GDPR passage — years of draft → compromise; US federal AI may follow but 2027 uncertain.
Would update if: GAAIA formally introduced with bipartisan committee chairs + White House sign-on; or preemption stripped early (→ separate P).
Conf: M
P = 0.85 — CA SB 53 compliance path proceeds
Claim: ~85% frontier labs operating in CA meet SB 53 transparency / catastrophic-risk reporting obligations (2026 effective) — law stands and is enforced absent successful federal preemption.
Why: Signed Sep 2025; Anthropic endorsed; Encode path; Newsom pro-innovation but signed; industry prefers transparency over SB 1047 caps.
Evidence:
- (internal note) §2 — SB 53 signed 2025-09-29; §22757.x B&P Code structure
- Brookings SB 53 explainer
- CA Gov signing release
- Playbook: SB 53 = Encode-backed, Anthropic endorsed
Analogue: CA privacy (CCPA) — state lab for rules that industry learns to comply with.
Would update if: GAAIA preemption passes and SB 53 dev-stage duties enjoined; or major lab publicly refuses compliance without penalty.
Conf: H
P = 0.50 — CA new labor-facing AI bill (displacement disclosure / hiring algo)
Claim: ~50% CA passes additional labor-oriented AI legislation (displacement disclosure, hiring algorithm rules, whistleblower expansion) in Node 1 window.
Why: Wiener/Encode lineage; Newsom N-6-26 WARN review due Nov 2026; SB 951 pending; NY RAISE weakened but labor plank survives in discourse.
Evidence:
- Newsom EO: 180-day WARN Act revision recommendations (gov PDF)
- (internal note) §3.5 — Wiener vs GAAIA preemption fight
- AFL-CIO principles: collective bargaining on AI deployment (workers-first-ai report)
Analogue: CA AB 5 gig-worker classification fights — state labor experimentation.
Would update if: Newsom signals opposition to any new AI labor bills beyond EO study; or SB 951 dies in committee without substitute.
Conf: M
P = 0.08 — CA compute cap / training licensing
Claim: ~8% California enacts compute cap or training licensing in Node 1 window.
Why: SB 1047 veto lesson; Newsom caution; industry + a16z opposition; SB 53 chose transparency lane deliberately.
Evidence:
- (internal note) §1.2 — SB 1047 vetoed; SB 53 transparency survives
- (internal note) Node 4 — P(CA cap passes) = 0.08 same reference class
- Primer §2.4 — SB 1047 vs SB 53 comparison table
Analogue: CA nuclear moratorium — single-state hard stop rare when industry HQ’d in state.
Would update if: Wiener or successor introduces cap bill with Newsom non-opposition signal.
Conf: M
P = 0.85 — Frontier labs: restate safety commitments, no unilateral slowdown
Claim: ~85% frontier labs respond to labor shock with Seoul-style voluntary safety restatements + continue training — no revenue-peak voluntary halt.
Why: Anthropic RSI: pause only with multilateral verification; no precedent for voluntary halt at revenue peak; labor shock increases deploy pressure.
Evidence:
- Anthropic When AI Builds Itself (2026-06-04): conditional multilateral pause — (internal note)
- Seoul 2024 Frontier AI Commitments — no enforcement ((internal note) §1.3)
- OpenAI board crisis 2023: internal governance reverted commercial default ((internal note) Node 4 analogues)
Analogue: Oil majors’ climate pledges during record profits — rhetorical safety, operational growth.
Would update if: ≥2 frontier labs public halt >30 days explicitly citing labor/external pressure (not internal incident).
Conf: H
P = 0.85 — Labs “reskilling partnership” PR campaign
Claim: ~85% at least one frontier lab launches high-visibility reskilling / community college partnership PR within Node 1 salience window.
Why: Standard crisis PR; deflects regulation; matches Anthropic/OpenAI existing edu partnerships pattern.
Evidence:
- Playbook §1.3 — indirect pressure includes workforce rhetoric
- AFL-CIO demands worker voice in deployment — labs counter with partnership optics (AFL-CIO AI page)
- Block/Dorsey narrative: “efficiency” not “abandon workers” framing in shareholder letter (VentureBeat)
Analogue: Tech company diversity pledges 2020 — visible programs, limited structural change.
Would update if: Major lab explicitly refuses any workforce program while under hearing pressure (unlikely).
Conf: M
P = 0.55 — Frontier labs material public junior-eng hiring freeze
Claim: ~55% at least one frontier lab announces material junior SWE hiring freeze or “AI-native small team” headcount policy publicly.
Why: Canaries + Block/Cloudflare; counter: OAI/Anthropic still hire juniors for alignment/safety; mixed signals.
Evidence:
- (internal note) §5 — posting ↑ but junior employment ↓ in ADP
- Anthropic RSI: >80% prod code AI-written — implied reduced junior need internally
- Meta 2026: cuts + AI rehire simultaneously ((internal note) §4.3)
Analogue: Finance 1990s — associate class sizes shrink while industry revenue grows.
Would update if: All major labs publish 2027 campus hiring numbers at or above 2024 baselines.
Conf: M
P = 0.85 — Public salience: economic anxiety ≫ x-risk
Claim: ~85% mainstream media / polling salience during Node 1 frames AI as jobs/economy not extinction — conditioning policy menu.
Why: Pew gap enormous: 56% public extremely/very concerned about job loss vs 25% AI experts; 64% expect fewer jobs vs 39% experts.
Evidence:
- Pew Apr 2025 report — job loss concern 56% vs 25%
- Pew workers survey Feb 2025 — 52% worried, 32% fewer long-run opportunities
- Playbook §2.2 Tier 4 — Pew, Seismic as opinion anchors
- Sanders S.4214 frames AI via jobs/data centers/environment — not x-risk (Roll Call)
Analogue: Climate vs jobs framing in coal country — economic frame wins legislative attention.
Would update if: Post-Node-4 whistleblower, x-risk share of AI media mentions exceeds jobs for >8 consecutive weeks.
Conf: H
P = 0.12 — Mass x-risk protest (≥10,000, DC)
Claim: ~12% organized x-risk / pause march ≥10k in DC during Node 1 window.
Why: FLI letter → zero street movement; PauseAI niche; Sharma “exit not organize” signal; labor frame dominates.
Evidence:
- (internal note) — 10k protest Emerging on slow track
- Playbook §Executive summary — FLI 1000+ signers, zero policy
- (internal note) Node 4 — P(organized pause >10k activists) = 0.15 same order of magnitude
Analogue: Anti-GMO marches — expert concern without mass US movement.
Would update if: Encode/FLI/PauseAI coalition achieves sustained 10k+ turnout (would upgrade to M conf and revise timing).
Conf: L–M
P = 0.30 — Economic / anti-DC protest (jobs, rates, data centers) by 2027 Q3
Claim: ~30% protest ≥10k focused on jobs, utility rates, or data-center buildout (not primarily x-risk) by 2027 Q3.
Why: Third Act data-center actions; AFL-CIO 2026 convention AI plank; Sanders/AOC coalition potential; higher than x-risk march but still sub-majority.
Evidence:
- S.4214 datacenter moratorium coalition language (Congress.gov)
- AFL-CIO Oct 2025 Workers First — nationwide mobilization program (press release)
- Playbook §1.4 — UK Seismic 40% want stop; US lacks equivalent organized street power
Analogue: Dakota Access / local infrastructure NIMBY — economic/environmental coalition, not existential.
Would update if: Regional utility-rate spikes + midterm economics create visible march before 2027 Q2.
Conf: L–M
P = 0.65 — Markets reward AI-attributed layoffs (margin narrative)
Claim: ~65% equity markets positive or neutral on AI-attributed large layoffs when revenue strong — reinforcing CEO incentive to cut + cite AI.
Why: Block +22% AH on announcement; political economy pushes acceleration.
Evidence:
- Fortune Feb 2026 — stock surge on cut news
- Forbes Apr 2026 — investor reward loop pattern
- (internal note) — “Stock market impact/backlash from AI job displacement” = Ahead (narrative) but “$3T valuation” Behind
Analogue: 2010s stock buyback era — shareholders reward headcount reduction.
Would update if: Block-style cuts repeatedly punished by >10% drawdowns across multiple firms.
Conf: M
P = 0.75 — AFL-CIO Workers First sustains cross-sector agenda
Claim: ~75% AFL-CIO Workers First Initiative on AI remains active organizing frame through Node 1 peak — meetings, state/local pushes, convention planks.
Why: Oct 2025 launch with 63 unions / ~15M workers; integrated with state/local AI task force; Apr 2026 Sanders meeting; eight principles document.
Evidence:
- AFL-CIO Workers First launch — Oct 15 2025
- Principles report — collective bargaining, guardrails, IP, R&D worker voice
- Tech Institute issue brief — surveillance, deskilling, job elimination harms
- Playbook §1.4 — unions + populist right blocked Cruz preemption Jun 2025
Analogue: AFL-CIO NAFTA fight 1990s — sustained agenda, mixed legislative wins.
Would update if: Shuler departs + successor deprioritizes AI; or federation fractures on AI stance across major affiliates.
Conf: M
P = 0.12 — Federal binding AI workplace standards
Claim: ~12% OSHA/NLRB/DOL issues binding federal AI workplace standards (not sector CBAs) in Node 1 window.
Why: No tech union density like WGA; federal gridlock; CA EO is study-only.
Evidence:
- Playbook §1.4 — sector deals vs federal standards
- WGA 2023 CBA AI limits — sector-specific ((internal note) §2023)
- Newsom EO: recommendations only, 180-day WARN review (Fisher Phillips)
Analogue: OSHA ergonomics rule fight 2000 — federal workplace tech rules hard.
Would update if: DOL publishes NPRM on AI workplace monitoring with enforceable penalties before 2027.
Conf: L
P = 0.55 — Unions + populist right block federal preemption of state AI laws
Claim: ~55% coalition of unions + populist Republicans prevents 10-year state AI regulation ban (Cruz-style) or weakens GAAIA preemption — preserving SB 53-class state law.
Why: Cruz moratorium stripped 99–1; ~$8.5M Q1 2026 preemption lobbying (Playbook); AFL-CIO explicit on preemption fight.
Evidence:
- (internal note) §1.2 — Cruz stripped 99–1 Jul 2025; §2.1 highest negative risk = federal preemption
- AFL-CIO: “Remove Ban on State-Level AI Accountability Laws from Budget Reconciliation” (aflcio.org/issues/future-work/ai)
- GAAIA 3-year preemption — active fight ((internal note) §3.4)
Analogue: Cannabis federalism — state labs survive despite industry wanting uniform federal light touch.
Would update if: Preemption passes in reconciliation with <10 dissenting Senate votes.
Conf: M
P = 0.70 — DoD/IC AI contracting expands despite labor anxiety
Claim: ~70% US executive branch continues scaling DoD/IC AI lab contracts through Node 1 — labor shock does not reduce natsec AI spend.
Why: Tracker DoD scaling Confirmed; natsec hawks strong anti-pause; jobs frame doesn’t cut defense AI budget.
Evidence:
- (internal note) — “DoD scales up AI lab contracting” Late 2026 = Confirmed
- Playbook §1.4 — natsec hawks vs full stop
- (internal note) — 2026 policy week: EO + GAAIA + Banks letter all pro-innovation/security frame
Analogue: Post-9/11 defense tech spending during civilian job shocks.
Would update if: Congress cuts AI line items in NDAA explicitly citing labor displacement (unlikely).
Conf: M
Tail branches
P = 0.03 — T1: Sanders moratorium surprise (S.4214 or successor)
Claim: ~3% Sanders S.4214 (AI datacenter moratorium until worker-protection + safety legislation) or successor gets floor vote / passes one chamber — surprise tail, not modal.
Why: GovTrack ~2% enactment; R Senate + industry opposition; but bundles jobs + environment + x-risk rhetoric; Cruz 99–1 shows extreme text can force negotiation.
Evidence:
- S.4214 introduced Mar 25 2026 — moratorium until worker displacement policies + safety review + datacenter conditions
- Roll Call Mar 2026 — three legislative categories to lift ban
- Playbook §1.2 — hard moratorium versions exist in Overton window but don’t pass
Analogue: Green New Deal floor votes — symbolic/adversarial passage possible, law unlikely.
Would update if: Moratorium passes Senate committee with R crossover; or attached to must-pass bill.
Conf: L (tail); M (order-of-magnitude)
P = 0.12 — T2: GAAIA preemption wins (SB 53 dev duties weakened)
Claim: ~12% GAAIA passes with 3-year preemption of state “model development” rules → SB 53 development-stage obligations weakened; Encode/Wiener defensive war lost.
Why: Strongest draft yet but preemption fight repeats Cruz dynamic; OpenAI prefers lighter EO; industry $8.5M lobbying.
Evidence:
- (internal note) §3.4 — GAAIA preemption vs SB 53 table
- Trahan/Obernolte draft Jun 2026
- Playbook §2.1 — preemption = highest negative risk for state wins
Analogue: FCC net neutrality preemption fights — federal uniformity vs state labs.
Would update if: Preemption stripped in markup (→ collapse tail to ~3%); or House floor vote scheduled with White House support (→ raise to ~25%).
Conf: M
P = 0.08 — T3: Agent catastrophe → federal vetting law (GUARD / RESPONSIBLE) — software baseline
Claim: ~8% publicized software/digital agent failure (Trigger E2: >$100M loss or enforceable judgment, no embodied/kinetic component) → Hawley GUARD / Blackburn RESPONSIBLE-class pre-deployment vetting passes — first binding federal deployment gate, not training cap.
Why: Child safety / financial harm frame bridges populist right + labor; precedent for Node 4; still hard in R Congress without vivid incident; software-only errors less legible than physical (Knight Capital vs Uber AV).
Evidence:
- Playbook §2.3 Tier 1 — Hawley GUARD Act, Blackburn RESPONSIBLE Act watch list
- (internal note) — T3 definition
- Sydney 2023 / agent error media cycles ~2 weeks ((internal note) Node 4 media analogue)
Analogue: Knight Capital 2012 — automated loss, regulatory attention but no binding pre-deployment vetting law.
Would update if: Major agent-caused financial loss with clear liability chain in H1 2027; or vetting bill passes committee on bipartisan vote without incident (exogenous).
Conf: L–M
P = 0.12–0.14 — T3 conditional on embodied Trigger E2 (GUARD tail reweight)
Claim: If embodied Trigger E2 fires (humanoid/AMR fatality or >$100M loss with clear AI chain), P(T3 vetting law) rises to ~12–14% — +4–6pp increment over software-only 8% baseline. Combined P(any Trigger E2 × T3 pass) ≈ P(E2_embodied ~0.10) × P(T3|E2_emb ~0.35–0.45) + software path.
Why: T-PA2 mechanism; injury/death satisfies Kingdon problem stream faster than API errors; GUARD frames deployment vetting; culture-war coalition (Hawley + labor) already aligned on vetting not pause.
Evidence:
- node physical ai embodied §T-PA2 — 8% → ~14% conditional
- node physical ai §T-PA2 — +4–6pp increment
- Uber AV 2018 — immediate regulatory response; Boeing 737 MAX — certification overhaul
- crosscut secondary cruxes §7 — vetting coalition modal 0.62 provides legislative substrate
Analogue: 737 MAX — single vivid failure → binding certification; software agent losses rarely achieve same without physical harm.
Would update if: Embodied incident occurs but vetting bill still dies in committee (→ increment <3pp); or software-only >$100M loss passes vetting first (→ embodied premium overstated).
Conf: L–M
Trigger E accelerators (optional)
P(E1 fires by 2027-H1) ≈ 0.25 — Second Block-class F500 cut + Challenger AI cuts >30% YoY quarter
Claim: ~25% second household-name ≥15% AI-attributed cut plus Challenger quarter with AI-cited cuts >30% YoY — pulls policy salience forward 1–2 quarters.
Evidence: Block precedent; May 2026 AI = 40% of monthly cuts; Dorsey predicts peer cuts (Fortune)
Analogue: Lehman after Bear — cascade salience.
Would update if: Q3 2026 Challenger shows AI-cited deceleration.
Conf: L–M
P(E2 fires by 2027) ≈ 0.10 — Agent error >$100M loss / enforceable judgment (software-only)
Claim: ~10% high-profile software/digital agent failure triggers liability/vetting frame (feeds T3 8% baseline).
Evidence: Playbook populist-right vetting bills; Node 2 CBRN uses similar Trigger E structure.
Analogue: Knight Capital 2012 — automated system, immediate regulatory attention.
Conf: L
P(E2_embodied fires by 2028) ≈ 0.10–0.15 — Physical agent failure (warehouse/factory/AV-class)
Claim: ~10–15% embodied failure (fatality, >$100M logistics loss, enforceable judgment with clear AI actuator chain) by 2028 — separate from software E2; feeds T3 12–14% conditional.
Why: Low base rate but fat-tail salience; pilot-scale humanoids + AMRs expanding 2026–2028; OSHA incident reporting; GUARD/RESPONSIBLE already in Judiciary pipeline.
Evidence:
- node physical ai embodied §Key crux — Trigger E2 definition (>$100M / judgment / fatality)
- §T-PA2 above; Uber AV 2018 analogue
- (internal note) — eval / GUARD-adjacent frames
Analogue: Uber AV pedestrian death 2018 — kinetic failure punctuates policy window.
Would update if: Zero OSHA-reportable humanoid incidents through 2028 despite >1k shift-work units (→ downgrade to <5%).
Conf: L
p(doom) bucket links
P(Δ extinction | Node 1 modal) ≈ 0 — Direct extinction channel
Claim: Node 1 modal resolution ~does not move P(extinction by 2050) directly — no misalignment or CBRN Tier 3 path opened by labor shock alone.
Why: Mechanism is economic/political, not capability escape or misuse enablement.
Evidence: (internal note) — four buckets; Node 1 scoped in (internal note) cross-node synthesis
Analogue: 2008 financial crisis — severe harm, not extinction.
Conf: H
P(Δ whimper | Node 1 modal) ≈ +small — Permanent agency loss
Claim: Modal path slightly increases P(permanent agency loss / whimper) — economic coalition beats x-risk coalition → ↓P(effective pause later); acceleration frame normalizes.
Why: Stix & Maas bridge partial success only; Node 4 modal still “continue training.”
Evidence:
- (internal note) — Node 1 ”↓ pause later”; cross-node synthesis
- (internal note) — crux “No effective pause/shutdown”; hypothesis whimper > extinction for this author profile
Analogue: Social media business model lock-in — gradual agency erosion, not bang.
Conf: M
P(Δ severe recoverable | Node 1 modal) ≈ +moderate
Claim: Modal path increases P(severe recoverable catastrophe): mass displacement, political instability, mis-governance under shock — not extinction.
Evidence: Canaries −16%; Challenger 101k AI-cited; Pew 56% job-loss concern; no federal transition law modal
Conf: M
P(Δ concentrated harm | Node 1 modal) ≈ +large — Primary channel
Claim: Node 1 modal is primary driver of concentrated harm bucket: junior pipeline collapse, workplace surveillance AI, hiring algo bias, inequality.
Evidence:
- AFL-CIO Tech Institute brief — surveillance, deskilling (link)
- Canaries cohort; Cloudflare “measurer” roles ((internal note) §4.2)
- (internal note) — concentrated harm ongoing bucket
Conf: H
P(Δ concentrated harm | physical AI modal) ≈ +large incremental — Blue-collar channel
Claim: Physical AI modal adds incremental concentrated harm: factory-floor deskilling, injury liability asymmetry, algorithmic line supervision, surveillance wearables — primary embodied channel per node physical ai embodied.
Why: Digital Node 1 already loads white-collar harm; physical extends timeline 2028–2032 without requiring macro unemployment spike.
Evidence:
- node physical ai embodied §p(doom) — concentrated harm primary channel for physical
- Node 1 × Physical AI interaction table (§Physical AI second wave)
Conf: M–H
P(Δ whimper | physical + digital compound) ≈ +2–4pp by C8–C9
Claim: If physical + digital marginalization compound by C8–C9, P(whimper / permanent agency loss) +2–4pp vs digital-only Node 1 modal.
Why: Bimodal labor market — millions of digital remote workers/agents + thousands of humanoids on shift (node physical ai embodied §Non-doom society preview).
Evidence:
- node physical ai embodied §p(doom) whimper row
- crosscut secondary cruxes §6 RSI locus — embodied favors whimper before extinction
Conf: M (speculative timing)
P(no effective pause | Node 1 modal) ↑ ~2–4pp — Coordination crux update
Claim: If Node 1 resolves modal, update P(no effective pause/shutdown) upward ~2–4 percentage points — labor wins transparency, not brakes; labs treat shock as deploy-faster mandate.
Why: Composite from actor table: federal cap <5%, labs no slowdown 85%, hearings without caps 75%.
Evidence: Playbook §1.5 — verification window closing; conditional pause needs treaty-grade metering
Analogue: Climate agreements — pledges without enforcement.
Conf: M
Confidence summary (node-level)
| Claim | Conf |
|---|---|
| METR/agents on fast hybrid track | H |
| Mass F500 eng shock + 10k protest on slow track (~2027 Q2 peak) | M |
| Digital-only partial live; physical second wave +12–24 mo | M |
| Modal = transparency/labor/vetting coalition, not federal cap | H |
| Junior cohort hurt already measurable | H |
| T3 8% software; 12–14% conditional on embodied E2 | L–M |
| Macro US unemployment spike from AI by 2027 | L |
| Tail T1 moratorium passes | L |
| Tail T2 preemption passes | M |
| Culture-war vetting modal 0.62 | M |
Falsifiers (master list)
| Observation | Implication |
|---|---|
| Federal frontier training moratorium or cap signed | Modal path wrong; revisit all pause priors |
| Zero HELP AI-labor hearings through 2027-12 post-Block | Salience model wrong |
| 22–25 SWE employment recovers to parity while METR >24h | Capability–employment link broken |
| ≥10k DC x-risk march before 2027-06, >2 mo media | Slow-track protest timing wrong |
| GAAIA preemption passes and SB 53 enforcement continues unscathed | Legal model wrong |
| Challenger AI-cited cuts <10% of total for 2 consecutive quarters while agents scale | AI-washing; downgrade Node 1 triggers |
| ≥10k humanoids paid shift-work w/ KPIs by 2028 | T-PA1 fires; physical salience pulls to 2027 |
| Embodied fatality + GUARD passes within 12 mo | T3 conditional validated; upgrade embodied E2 |
| Embodied fatality + no vetting bill in 18 mo | Embodied E2→T3 increment <3pp |
| TRUMP AMERICA AI Act passes without ELVIS/KOSA carve-outs | Culture-war anti-reg tail validated |
| Cross-ideological halt >60 days post-C10 incident | Vetting-not-pause equilibrium falsified |
Source index
Phase 2b — physical AI, GUARD, T-PA1, culture-war synthesis
Cross-refs: §Physical AI second wave (above); §Culture-war coalition (above); node physical ai; crosscut secondary cruxes §7 → Node 1 ext
Phase 2b-1. Two-wave labor model — digital then physical
Claim: Node 1 resolves as two sequential labor shocks: digital/agent peak ~2027 Q2, physical/embodied peak 2028–2031 (+12–24 mo), unless T-PA1 or T-PA2 pulls physical forward.
Why: Same Ci spine; different deployment surfaces (API vs factory integration). Policy coalitions differ: HELP/white-collar vs OSHA/UAW.
Evidence: node physical ai embodied §Hybrid timing; §T=2028–2031 above; Canaries already live on digital layer only.
Analogue: ATM decade-lag on teller politics — here 2–4 yr because factory visibility higher.
Would update if: UAW embodied-AI campaign 2026-H2 with strike threat tied to humanoids.
Conf: M
Phase 2b-2. P = 0.65 — digital peak ≥12 mo before physical coalition
Claim: Manufacturing/UAW mass mobilization lags digital Node 1 peak by ≥12 months in modal path (point 0.65).
Why: White-collar cuts live (Block, Canaries); humanoids <5,000 shift-work units by 2029; AFL-CIO Workers First frames general AI before embodied-specific.
Evidence: node physical ai §P=0.65; Yale macro null — physical not aggregate-binding.
Analogue: Offshoring 2000s — service headlines before factory automation politics.
Conf: M
Phase 2b-3. P = 0.75 — humanoids pilot-scale through 2029
Claim: <5,000 humanoids worldwide in paid multi-shift industrial work by 2029-12 — physical AI real in pilots, not macro-binding.
Why: Agility + Figure only verified shift KPIs; Tesla Optimus still data-collection-heavy (Musk Jan 2026: zero useful factory work).
Evidence: Figure F02: 1 robot, 1,250 hrs @ BMW; Agility ~40–150 units global est.; Robonaissance ~100 sustained productive globally.
Analogue: Autonomous trucking — years between pilot and macro displacement.
Conf: M–H
Phase 2b-4. T-PA1 — P = 0.12 OEM scale surprise
Claim: ~12% tail: ≥3 OEMs each >500 paid shift-work humanoids by 2028 + replicated throughput KPIs → pulls physical salience to 2027–2028; merges AFL-CIO digital + UAW blue-collar waves.
Why: Would collapse digital–physical lag; Figure/Agility pipelines could scale if integration bottleneck breaks.
Evidence: node physical ai embodied §T-PA1; (internal note).
Analogue: iPhone scale-up — single OEM hit changes entire supply-chain politics timeline.
Would update if: Two additional OEMs publish >200 shift-work units each with KPIs by 2028-H1 → revise T-PA1 to 0.25+.
Conf: L–M
Phase 2b-5. T-PA2 — P = 0.10 embodied catastrophe Trigger E2
Claim: ~10% vivid embodied failure (injury, fatality, >$100M factory loss) before 2028 → Trigger E2 class; pulls GUARD/RESPONSIBLE vetting salience forward 12–18 mo.
Why: Physical harm more legible than API errors; Boeing 737 MAX / Uber AV 2018 precedents; OSHA reporting mandatory.
Evidence: §T-PA2 above; (internal note); Node 12 §11 embodied before C10.
Analogue: Uber AV pedestrian death — immediate regulatory cycle.
Conf: L–M
Phase 2b-6. GUARD baseline — P = 0.08 software-only T3
Claim: ~8% software/digital agent catastrophe (Trigger E2: >$100M loss or enforceable judgment, no kinetic component) → Hawley GUARD / Blackburn RESPONSIBLE pre-deployment vetting passes federally.
Why: Kingdon problem stream weaker for API errors; Knight Capital precedent shows possible but rare federal response; Judiciary pipeline already moving on child-safety frame.
Evidence: Playbook §2.3 Tier 1; Hawley GUARD Judiciary unanimous Apr 2026; §P=0.08 T3 above.
Analogue: Knight Capital 2012 — automated harm → reg attention, not industry halt.
Conf: L
Phase 2b-7. GUARD embodied conditional — P = 0.12–0.14
Claim: P(federal deployment vetting law | embodied Trigger E2) = 0.12–0.14 — +4–6pp vs software-only 0.08 baseline.
Why: T-PA2 mechanism; injury/death satisfies problem stream faster; Hawley + labor coalition already aligned on vetting-not-pause; GUARD frames deployment gate.
Evidence: §P=0.12–0.14 T3 conditional above; Node 12 L3 embodied first-tail; crosscut §7 vetting modal 0.62.
Analogue: Medical device MAUDE reports — physical harm accelerates FDA pathway vs software SaMD alone.
Would update if: Software-only >$500M agent loss with sustained media ≥ embodied salience → collapse embodied premium.
Conf: L–M
Phase 2b-8. GUARD vs RESPONSIBLE — parallel Judiciary tracks
Claim: P(at least one of GUARD (Hawley–Blumenthal) or RESPONSIBLE (Blackburn) advances to floor by 2028) = 0.55; P(both in package deal) = 0.30.
Why: Unanimous Judiciary on GUARD Apr 2026; Blackburn selective preemption preserves child-safety/deepfake carve-outs; not frontier catastrophic-risk lane but sets deployment-vetting precedent.
Evidence: Playbook §2.3; entity_master_list Blackburn/Hawley entries; §Culture-war actor table above.
Analogue: KOSA/ELVIS parallel tracks — multiple vetting bills, one may pass.
Conf: M
Phase 2b-9. Culture-war vetting modal — P = 0.62
Claim: Equilibrium US AI policy 2026–28 = BMIA + SB 53 + NO FAKES + GUARD-class vetting; explicitly not training pause or compute caps — 62% modal.
Why: Cruz moratorium stripped 99–1; federal mandatory pause dead; Encode SB 53 path live; 55% oppose regulation moratorium (2025).
Evidence: crosscut secondary cruxes §7 → Node 1 ext; §P=0.62 above; (internal note) §5.
Analogue: Post-Cambridge Analytica — sector rules, not shutdown.
Conf: M
Phase 2b-10. Anti-regulation fusion tail — P = 0.20
Claim: ~20% GAAIA-class broad preemption without carve-outs → blocks even state vetting (Encode defensive war lost).
Why: Trump EO Jun 2026; GAAIA 3-year preemption; $8.5M Q1 2026 lobbying; subset of Node 1 §T2 P=0.12.
Evidence: Crosscut §7; Node 6 GAAIA branch; playbook §2.1 preemption highest negative risk.
Analogue: FCC net neutrality preemption killing state labs.
Would update if: TRUMP AMERICA AI Act passes with ELVIS/KOSA carve-outs → collapse to <10%.
Conf: M
Phase 2b-11. Culture-war actor table — load-bearing rows (synthesis)
Claim: Twelve-actor table (Musk/xAI, Hawley, Blackburn, Cruz, Banks, Encode, AFL-CIO, Sanders, Trump EO, Big Tech, natsec hawks, AI Now) sets P(federal training cap) <5% and P(vetting advances) >50% — Node 1 political menu bounded.
Why: Generalizes Node 4 actor row P(coalition w/ labor/populist-right on vetting)=0.30; Musk jobs frame dominates x-risk P≈0.85.
Evidence: Full table §Culture-war coalition — expanded actor table above; (internal note); playbook §3.1 Bannon/Hawley killed Cruz moratorium.
Analogue: Climate coalition fragmentation — many actors, no pause bloc majority.
Would update if: Hawley + Encode co-sponsor joint workforce + vetting bill with committee hearing.
Conf: M
Phase 2b-12. Node 1 × Node 10 — hidden near-miss dulls labor punctuation
Claim: If Node 10 hidden:public ratio >3:1 (modal), agent/finance near-misses fail to trigger Trigger E2 → GUARD embodied tail stays 0.12–0.14 not 0.18+ even with physical incidents contained internally.
Why: Lobstar Wilde public but not mapped to Trigger E; embodied factory incidents may stay OSHA confidential until vivid failure.
Evidence: node10 §10, §18; Node 10 P(agent finance hidden)=0.60.
Analogue: Workplace injury underreporting — true rate >> headline rate.
Conf: L–M
Phase 2b-13. Node 1 × Node 12 — L3 embodied first lowers extinction, raises whimper
Claim: If Node 12 resolves L3 embodied first (P=0.12), Node 1 physical wave pulls forward + ↑ GUARD tail — but ↓ P(extinction) 2–3pp (whimper/severe reweight, Hanson mix ↑).
Why: Branch selector interaction; physical harm legible, not paperclip; factory vetting ≠ training halt.
Evidence: node12 §21; Node 12 cross-node matrix N1 row.
Analogue: Industrial revolution — concentrated harm ↑, extinction risk not the frame.
Conf: M
Phase 2b-14. p(doom) Phase 2b net on Node 1 channels
Claim: Phase 2b additions do not move P(extinction) directly; ↑ concentrated harm + whimper; culture-war vetting modal ±1–3pp net on ~17% doom region (emergent sim); GUARD embodied tail ↓ bio friction indirectly (deployment gate) while ↑ race (no pause).
Why: Node 1 scoped to economic/political in my pdoom; crosscut §7 mixed direction.
Evidence: §p(doom) bucket links above; crosscut §7 p(doom) direction table; Node 10/12 interaction §Phase 2b-12/13.
Analogue: Phase 4a stitch — labor node orthogonal to extinction except via coordination crux.
Conf: M
Update log
| Date | Change |
|---|---|
| 2026-07-03 | Initial evidence rationale from Node 1 subagent pass + web verification |
| 2026-07-04 | Phase 2b — 14 subsections: physical AI second wave, GUARD 12–14%, T-PA1/PA2, culture-war synthesis |
| 2026-07-04 | Physical AI second wave (+12–24 mo); T-PA1/T-PA2; GUARD T3 8%→12–14% embodied; culture-war coalition actor table; cross-links X2-A |