# The Disarmed Countermovement — working notes

New paper, sibling to *Nobody to Blame* (attribution) and *The Third Disembedding*
(the model). This one carries the history, philosophy, and political economy that was
cut from *Nobody to Blame*, re-aimed at the question: **what happens politically if AI
devalues labor at scale — unrest? authoritarian turn? and how does the K/L battle
resolve?**

## 1. Thesis (one sentence)

The countermovement will come — displacement always produces politics — but for the
first time in the industrial era it will arrive **without the structural leverage**
that converted every previous countermovement into reform rather than pacification or
repression; therefore the decisive question is not whether society responds but
whether the response is **bargained** (ownership), **bought** (income without power),
or **broken** (repression) — and the bargaining window is open only while labor is
still needed.

Three legs, each with verified evidence:
1. **Reform was always purchased with leverage.** Franchise ← threat of revolution
   (Acemoglu–Robinson 2000; Earl Grey: "to prevent the necessity of revolution").
   Welfare states and top tax rates ← mass-mobilization war (Scheve–Stasavage:
   WWI raised top rates ~+33pp; ≥2% of population under arms threshold; mass-army era
   1859–1970 per Onorato et al. — exactly the era of democratization). Wagner Act ←
   1934 strike wave. Power resources theory (Korpi); structural vs associational
   power (E.O. Wright, AJS 2000).
2. **AI attacks all three sources of leverage at once**: production (strike collapse:
   470 major US stoppages 1952 → avg 17.8 in 2006–25; union density 20.1%→10.0%;
   robots→union decline causal, IZA DP 17864), soldiering (mass army ended ~1970;
   Ferejohn–Rosenbluth: automation of war → "democracy may atrophy"), and it cheapens
   repression relative to concession (Xu Xu AJPS 2021: surveillance counties CUT
   welfare/co-optation spending by more than they raised security spending — the
   substitution is directly measured; Beraja et al. "AI-tocracy" QJE 2023 loop;
   Kendall-Taylor/Frantz/Wright: digitally repressive autocracies last ~2.5×).
3. **Unrest ≠ leverage.** Weimar: the unemployed went KPD, not NSDAP (Falter); the
   frightened propertied went fascist. Chenoweth: nonviolent-campaign success
   65% (1990s) → <34% (2010s). Caprettini–Voth: machines caused riots (2× machine-
   breaking probability) — and the riots were crushed (Swing: 19 hanged, ~500
   transported; Luddites: 12,000 troops).

K/L endgame (from Third Disembedding §8, reused): if automation completes, labor
share → 0 for any elasticity, but capital is reproducible, so surplus migrates to the
non-reproducible — land, energy, siting, spectrum, (maybe) compute inputs. The battle
is ultimately over RENTS, and ownership is already concentrated: top 1% = 31.9% of US
net worth (Fed DFA Q4-25), top 10% own 87.4% of corporate equities, Mag-7 = 34.3% of
S&P 500 (Dec-25), AI capex = 92% of H1-2025 GDP growth (Furman).

Options judged by ONE criterion derived from the thesis: does the instrument leave
ordinary people holding something the powerful still need? Income (UBI) fails the
criterion alone; ownership-with-governance passes; the window argument (legislate
while labor still matters) supplies the urgency.

## 2. Epistemic honesty constraints (from Aaron's brief)

- "Even AI is not taking jobs like we thought — it is still early." The paper is
  explicitly CONDITIONAL: capabilities are running ahead of displacement (IMO-gold
  2025; Challenger 2025: 54,836 of 1,206,374 announced cuts explicitly AI-attributed
  ≈ 4.5%; Carnegie protest tracker 2025: AI absent from protest drivers). State this
  up front; the paper is about the politics IF displacement arrives, and about why
  the pre-displacement window matters — the conditionality is load-bearing for the
  window argument, not a hedge.

## 3. Attack log (errors caught before drafting)

- **Luddite/Wellington claim is FALSE as usually told** (Hobsbawm). Howes: Wellington
  sailed with 11,000 in 1808 but had ~50,000 by 1812. Use "12,000 troops, comparable
  to the initial expeditionary force" and footnote the correction. DO NOT repeat the
  meme.
- **Speenhamland**: Polanyi's account is historiographically dead (Block & Somers
  2003: allowances localized, intermittent, abandoned before 1834; 1834 Commission
  report written before its own evidence). But the POWER critique of UBI does not
  depend on Speenhamland. Present both; do not launder the myth.
- **Weimar**: do NOT say unemployment → Nazis. Falter: unemployed → KPD; NSDAP was a
  cross-class protest party with middle-class bulge; students 9× overrepresented.
  The elite-overproduction echo (Turchin) is the respectable version.
- **"AI-tocracy" authors**: Beraja, Kao, Yang, Yuchtman (no "Chen").
- **Homestead militia**: ~6,000 verified; the 8,500 figure is not.
- **Swing transportations**: 481 (Hobsbawm–Rudé) vs 505 (some sources) — footnote.
- **OpenResearch employment effect**: 2.0pp participation decline is the verified
  headline (the 4.1 figure floating around is unverified).
- **Lall 2025 preprint** (AI exposure → mainstream right, LESS redistribution
  demand) cuts AGAINST the assumption that AI displacement produces left demand —
  feature it honestly; it supports the "anxiety phase ≠ displacement phase" frame.
- **Scheidel**: leveling came from violence, not policy — this is a base rate AGAINST
  the bargained outcome; the paper must not pretend the odds are good, only that the
  window exists.
- r > g: attribute to Piketty as tendency, not law.

## 4. Key verified numbers bank

(see research dossiers in session; all URLs verified 2026-08-01)
- Franchise: UK electorate 813k/23M (1832); taxes 8.12%→18.8% of national product
  1867→1927; UK Gini 0.627 (1871) → 0.443 (1901).
- War & taxes: top rate UK 8.33→60% (1914→20); US 7→77%; +32.8pp WWI effect.
- Strikes: US peak 470 (1952), record low 5 (2009), 2025: 30; UK days lost 162.2M
  (1926) → 273k (2018). Union density US 20.1% (1983) → 10.0% (2025).
- Digital repression: Stasi 1:66 → AI removes manpower constraint; Golden Shield
  counties +11.6% security spend, −30.75M CNY welfare; autocracy duration ~10yr →
  ~25yr post-2000; Freedom House: 15 consecutive years of decline, US 73/100.
- Robots→politics: Frey et al. 2SLS +1.31pp Republican per unit exposure; MI flips
  at 10% less exposure. Anelli: 75th-vs-25th vulnerability +3.5pp radical right on
  ~5% base. de Bromhead: Germany-sized slump + young democracy + existing fascist
  party → +12.9pp fascist vote; Nazi 2.6% (1928) → 43.2% (Jul 1932).
- Alesina–Perotti: middle-class share is the robust instability predictor (−0.68).
- Turchin: 2010 Nature prediction; PSI = MMP × EMP × SFD; post-2010 unrest ×10.
- Engels' pause: 1780–1840 GDP/cap +46%, real wages +12%; labor share 50%→45%;
  reversal 1840–1900 (+90% output/worker, +123% wages). ~60 years.
- UBI: OpenResearch −2pp participation, −1.3 hr/wk, wellbeing gains fade after yr 1;
  Finland +6 days employment, wellbeing up; Stockton FT employment 28→40%.
- Funds: Norway ~$2.23T (~$400k/capita, my arithmetic); Alaska 2025 dividend $1,000,
  no aggregate employment effect (Jones–Marinescu); ESOPs 15.1M participants, >$2T;
  CT baby bonds live ($3,200/child); DC's repealed — fragility exhibit.
- Elite statements: Amodei 2025: half of entry-level white-collar jobs, 10–20%
  unemployment, ~3% token tax; Altman: American Equity Fund 2.5% market cap + land;
  Hassabis: "radical abundance," new economic theory needed. IMF: 60% of advanced-
  economy jobs exposed.

## 5. Structure

Part I the two engines (Marx, Polanyi — history/philosophy, from old drafts).
Part II how reform was bought, 1832–1975 (leverage history + table).
Part III what AI does to the bargain (three margins + small concession model).
Part IV will unrest spike (base rates, modern signal, answer: unrest yes, leverage no).
Part V the battle between K and L (limit case, rents, options judged by the leverage
criterion, the window argument).
Part VI objections, indicators, conclusion. Appendices in full edition.

## 6. Error log (running)

- (v1.0 build) none yet.

## 6. Error log (running) — v1.0 build
- ℓ / subscript / ∂ glyphs render as boxes in base-14 Times: replaced with
  <i>l</i>, <sub>, and words. Check any future equation for non-Latin glyphs.
- add(figure(...)) crashes ReportLab (figure() returns a list): use ext().
- Fig 1 caption originally said "one in three private workers at mid-century" —
  the standard mid-1950s ~33% figure is ALL wage/salary workers; corrected.
- Luddite/Wellington meme, Speenhamland disaster story, and
  unemployment→Hitler story each explicitly corrected in text, not repeated.

## 7. Draft 1.1 (same day): Part VI added — "Five worlds where the machines work"

Per Aaron: more philosophy + AGI-does-most-work scenarios. New Part VI (old VI → VII,
sections renumbered 18-20 → 21-23):
- §18 The third problem: meaning. Keynes 1930 (the dread, not the 15-hr week);
  Aristotle Politics 1253b (shuttle/plectrum — the automation reductio we now live);
  scholē; Hegel master-slave (the slave develops, the master stagnates — full
  automation makes everyone master); Arendt labor/work/action ("society of laborers
  without labor"); Marienthal 1933 (apathy not revolt — also independent support for
  Part IV's unrest≠leverage); Case-Deaton as modern replication.
- §19 Five worlds on 2 axes (ownership concentrated↔diffused × leverage lost↔
  retained), present in the unstable corner: I Rentier Ancien Régime (drift default),
  II Caretaker State (Tocqueville soft despotism — the pacification attractor),
  III Shareholder Commonwealth (bargained; inherits the meaning problem honestly —
  universal aristocracy unprecedented; Danaher/Bostrom on where meaning migrates),
  IV Perfected Autocracy (the state that needs nothing — severs the social-contract
  assumption from Hobbes forward), V Human Preserve (Baumol as policy; horse/vinyl
  precedents; enclave inside II or III). Fig 3 map with drift paths.
- §20 The last irony: AGI ends labor's commodity fiction by worthlessness, not
  re-embedding; property = the residual claim to mattering; ownership and meaning
  arguments are one argument.
- Fig 3 v1 had incoherent quadrants (caretaker placed under diffused ownership);
  redrawn as positional map with explicit unstable-present region + drift arrows.
- Refs added: Arendt, Aristotle, Bostrom, Case-Deaton, Danaher, Frase, Hegel,
  Jahoda, Keynes, Tocqueville.