← Research themes · Last updated 24 Oct 2026

A gleaming plateau of towers separated by a deep chasm from a plain of modest houses, joined by a thin bridge

AI and the great divergence

Government debt, jobs, civil unrest and the concentration of wealth as AI scales.

Working thesisAI will widen the gap between owners of capital and everyone else: the ultra-rich get richer, labour markets come under pressure, already-high government debt gets harder to carry, and civil unrest rises.

This page tracks four linked claims. Every chart is drawn from public data fetched by scripts/update_data.py; anything that could not be fetched is shown as data pending, not estimated.

Government debt

The question: does AI raise growth and tax receipts fast enough to stabilise debt, or do lower labour income, bigger transfers and higher interest costs make it worse?

General government gross debt, % of GDP

% of GDP

IMF World Economic Outlook via DataMapper API. Values from 2026 onward are IMF projections (dashed).
Source: IMF World Economic Outlook, DataMapper (GGXWDG_NGDP) · Data as of 2025 · fetched 24 Oct 2026 · dashed = projection from 2026
View the data / CSV

US federal debt as % of GDP

% of GDP

Source: FRED, Federal Reserve Bank of St. Louis (GFDEGDQ188S) · Data as of 1 Jan 2026 · fetched 24 Oct 2026

US federal interest payments (annualised)

US$ billions, SAAR

Source: FRED, Federal Reserve Bank of St. Louis (A091RC1Q027SBEA) · Data as of 1 Apr 2026 · fetched 24 Oct 2026

US total public debt

US$ trillions

Source: FRED, Federal Reserve Bank of St. Louis (GFDEBTN) · Data as of 1 Apr 2026 · fetched 24 Oct 2026
View the data / CSV

What to watch

  • IMF Fiscal Monitor (April and October): global and US debt paths and the interest-burden projections.
  • US federal interest outlays as a share of receipts, and the 10-year Treasury yield at auctions.
  • Productivity-growth estimates that include generative AI (Penn Wharton, Goldman, CBO), compared with what is assumed in budget baselines.
  • Any proposal to tax AI rents, compute or automation to fund transfers.

Reading

Unemployment and labour

Exposure is not the same as displacement. Watch the headline rate, but also the groups most exposed early: young graduates and entry-level white-collar workers.

US unemployment rates

%

Source: FRED, Federal Reserve Bank of St. Louis (UNRATE, U6RATE) [2] · Data as of 1 Sep 2026 · fetched 24 Oct 2026

US unemployment: young workers vs. graduates

%

Entry-level and white-collar workers are where early AI labour effects would be expected to show.
Source: FRED, Federal Reserve Bank of St. Louis (LNS14024887, LNS14027662) [2] · Data as of 1 Sep 2026 · fetched 24 Oct 2026

World unemployment rate (ILO modelled)

% of labour force

Source: ILO modelled estimates via World Bank, processed by Our World in Data · Data as of 2025 · fetched 24 Oct 2026
View the data / CSV

What to watch

  • Unemployment among 20–24 and 25–34 year-olds and recent graduates compared with the overall rate.
  • Employment in AI-exposed occupations (software, customer service) versus less-exposed ones, from studies using payroll data.
  • Usage data from the Anthropic Economic Index showing automation versus augmentation.
  • The labour share of income and the gap between productivity and pay.

Reading

Civil unrest

Protest counts measure mobilisation, not grievance. Rising numbers can mean a healthy democracy as much as instability, so arrests and property damage are tracked separately.

US protest events per month (Crowd Counting Consortium)

events / month

Counts computed by this site from CCC event records (online-only events excluded). Phase 2 (2021–24) and Phase 3 (2025–) use slightly different coding; treat the Jan 2025 join as a series break. Coverage depends on media reporting.
Source: Crowd Counting Consortium (Harvard Ash Center / UConn), Harvard Dataverse [2] · Data as of May 2026 · fetched 24 Oct 2026
View the data / CSV
Data pending
Global demonstrations & riots (ACLED)
ACLED requires a free account + API key. Register, then add ACLED_KEY/ACLED_EMAIL support to this script.
Intended source: ACLED

What to watch

  • Monthly US protest events from the Crowd Counting Consortium, and the share involving arrests or damage.
  • Global demonstration and riot counts from ACLED (they need a free API key; see the data status page).
  • Protests that explicitly target AI, automation or data centres.
  • Carnegie Global Protest Tracker entries driven by the cost of living.

Reading

The ultra-rich getting richer

Two lenses: billionaire wealth (Forbes) and the distribution of US household wealth (Federal Reserve Distributional Financial Accounts).

Forbes World's Billionaires: combined net worth and headcount

US$ trillions / people

Forbes annual list (snapshot each March/April), transcribed from the Forbes-cited table on Wikipedia (revision 1381326652).
Source: Forbes World's Billionaires list, via Wikipedia table [2] · Data as of 2026 · fetched 24 Oct 2026

Share of US household net worth: top vs. bottom

% of total net worth

Federal Reserve Distributional Financial Accounts (DFA), quarterly since 1989.
Source: FRED, Federal Reserve Bank of St. Louis (WFRBST01134, WFRBSTP1300, WFRBSB50215) [2] [3] · Data as of 1 Apr 2026 · fetched 24 Oct 2026

What to watch

  • The top 1% and top 0.1% shares in the Fed DFA each quarter.
  • Forbes and Bloomberg rankings: how much of the gain comes from AI-linked equity (Nvidia, hyperscalers, frontier labs).
  • Private valuations of AI labs, which are not captured in public-market wealth.
  • Proposals for wealth or windfall taxes on AI gains.

Reading

Prediction markets

Live prices from public prediction markets, snapshotted by scripts/update_markets.py. A price of 30% means traders pay about 30¢ for a contract that pays $1 if the event happens; treat it as a crowd estimate, not a fact. Polymarket and Kalshi are real-money markets (Kalshi is CFTC-regulated); Manifold uses play money and is often thin. Long-dated and low-volume markets are poorly calibrated. Each market's exact resolution rules on the platform decide what counts, so read them before relying on a headline question.

Recession and jobs

7%
$2.4M vol · $38k liquidity · closes 2027-01-31 · as of 2026-10-30 17:06 UTC
32%
$18k vol · $19k liquidity · closes 2027-12-31 · as of 2026-10-30 17:06 UTC
5%
$169k vol · $2k liquidity · closes 2027-02-01 · as of 2026-10-30 17:06 UTC
2%
$218k vol · $14k liquidity · closes 2027-02-01 · as of 2026-10-30 17:06 UTC
1%
$109k vol · $5k liquidity · closes 2027-02-01 · as of 2026-10-30 17:06 UTC
2%
$38k vol · $1k liquidity · closes 2027-02-01 · as of 2026-10-30 17:06 UTC
2%
$54k vol · $1k liquidity · closes 2027-02-01 · as of 2026-10-30 17:06 UTC
92% bid 19¢ / ask 93¢
2k contracts · 187 open interest · closes 2030-01-04 · as of 2026-10-30 17:06 UTC
87% bid 81¢ / ask 94¢
758 contracts · 248 open interest · closes 2030-01-04 · as of 2026-10-30 17:06 UTC
90% bid 65¢ / ask 100¢
957 contracts · 154 open interest · closes 2030-01-04 · as of 2026-10-30 17:06 UTC
56% bid 17¢ / ask 58¢
2k contracts · 1k open interest · closes 2030-01-04 · as of 2026-10-30 17:06 UTC
49% bid 35¢ / ask 49¢
3k contracts · 834 open interest · closes 2030-01-04 · as of 2026-10-30 17:06 UTC
23%
Ṁ34k play-money · 175 traders · closes 2029-12-31 · as of 2026-10-30 17:06 UTC

Government debt

>99% bid 100¢ / ask 100¢
63k contracts · 18k open interest · closes 2029-03-31 · as of 2026-10-30 17:06 UTC
93% bid 93¢ / ask 100¢
2k contracts · 127 open interest · closes 2029-03-31 · as of 2026-10-30 17:06 UTC
86% bid 85¢ / ask 86¢
19k contracts · 6k open interest · closes 2029-03-31 · as of 2026-10-30 17:06 UTC
64% bid 61¢ / ask 64¢
3k contracts · 2k open interest · closes 2029-03-31 · as of 2026-10-30 17:06 UTC
24% bid 23¢ / ask 26¢
31k contracts · 9k open interest · closes 2029-03-31 · as of 2026-10-30 17:06 UTC
3%
$11k vol · $4k liquidity · closes 2027-01-01 · as of 2026-10-30 17:06 UTC

Unrest and emergency powers

20%
$179k vol · $17k liquidity · closes 2027-01-01 · as of 2026-10-30 17:06 UTC
0%
$911 vol · closes 2026-10-01 · as of 2026-10-30 17:06 UTC
20% bid 19¢ / ask 20¢
389k contracts · 144k open interest · closes 2027-01-01 · as of 2026-10-30 17:06 UTC
53% bid 33¢ / ask 54¢
308 contracts · 226 open interest · closes 2028-01-01 · as of 2026-10-30 17:06 UTC
48% bid 47¢ / ask 48¢
228k contracts · 48k open interest · closes 2029-01-20 · as of 2026-10-30 17:06 UTC
1%
$551k vol · $40k liquidity · closes 2026-11-01 · as of 2026-10-30 17:06 UTC
8%
$2.1M vol · $136k liquidity · closes 2027-01-01 · as of 2026-10-30 17:06 UTC
0%
$45k vol · closes 2026-10-22 · as of 2026-10-30 17:06 UTC

Wealth concentration

23% bid 20¢ / ask 23¢
7k contracts · 2k open interest · closes 2033-01-01 · as of 2026-10-30 17:06 UTC
18% bid 17¢ / ask 19¢
10k contracts · 1k open interest · closes 2033-01-01 · as of 2026-10-30 17:06 UTC
10% bid 2¢ / ask 9¢
8k contracts · 2k open interest · closes 2033-01-01 · as of 2026-10-30 17:06 UTC
4% bid 5¢ / ask 11¢
4k contracts · 1k open interest · closes 2033-01-01 · as of 2026-10-30 17:06 UTC
24%
$4.5M vol · $93k liquidity · closes 2026-11-04 · as of 2026-10-30 17:06 UTC
15%
Ṁ3k play-money · 33 traders · closes 2030-12-31 · as of 2026-10-30 17:06 UTC

Snapshot fetched 2026-10-30 17:06 UTC from public APIs (Polymarket Gamma/CLOB, Kalshi trade API, Manifold v0). Prices are what traders are paying, not my forecasts and not advice; thin or long-dated markets can be far off. Sparklines show daily prices over up to 120 days. Metaculus: API now requires an authenticated account token (HTTP 403 without one) and the website is behind a bot challenge, so Metaculus forecasts are not included. Add a ref type with a personal token later if wanted.

How to add a market

Add a ref to "markets" → "groups" → "refs" in src/themes/<slug>.json: {"platform":"polymarket","event":"<slug after /event/ in the URL>"}, {"platform":"kalshi","event":"<EVENT_TICKER>"} or {"platform":"manifold","id":"<market id or slug>"}. Optional: "markets" (list of specific sub-market slugs or tickers), "max", "sort" (date, prob or api). Then run python3 scripts/update_markets.py <slug> and rebuild.

Notes & journal

Dated observations from src/notes/ai-great-divergence.md. Add a new entry with a ## YYYY-MM-DD Title heading.

24 Oct 2026

Tracker started

Baseline as of today: US unemployment, Fed DFA wealth shares and IMF debt are loaded (see the charts for their as-of dates). ACLED global protest data is pending until an API key is added.

Question to revisit each quarter: is the youth/graduate unemployment gap widening faster than the headline rate?

Data behind this figure