We publish three kinds of research: annual benchmarks on where the US economy is on its AI adoption curve, data investigations into the labor market as AI takes hold, and visual essays on the ideas that produced the current moment.
Flagship research
Labor market
April 22, 2026
Twelve thousand US mass-layoff filings, normalized and cross-joined against industry AI adoption — where the cuts are concentrated and what the signal predicts.
Headline finding
Unweighted Pearson r = +0.33 between industry AI adoption score and % of workforce in WARN filings. Each +1 point in industry AI adoption associates with +0.11% of that sector entering WARN filings.2026 benchmark
January 15, 2026 · Updated April 10, 2026
Our annual benchmark across 9,691 US companies and 1,016 occupations — adoption score distribution, sector leaders, and the methodology behind every number we publish.
Headline finding
Median US company AI adoption score is 66.7 on a 0–100 scale. Information (NAICS 51) leads at 70.4; Agriculture (NAICS 11) trails at 62.3.Visual essay
April 20, 2026
A scroll-driven visual journey through 68 years of AI — Perceptron to trillion-parameter transformers. Six architectural eras mapped with methodology and citations.
Headline finding
Every era of AI has collapsed or scaled at a specific architectural inflection. We argue an emerging seventh era — agentic — is forming now.Data products
Explorers and rankings built from the same underlying dataset that powers our flagship research. Use them to drill into your industry, state, or occupation.
9,691 US companies scored 0–100, with state + industry leaderboards.
Open →
1,016 US occupations scored for AI replaceability using O*NET and BLS data.
Open →
Category-level analysis of which software categories AI agents are most likely to displace.
Open →
Tile-map + state ranking of average AI adoption across the 50 states.
Open →
Heatmap of 55 industries × 51 states — adoption score per cell.
Open →
Directory of AI automation platforms by capability and use case.
Open →
Every number we publish is reproducible. Source data is cited inline; rollups, scoring formulas, and prediction models are open to inspection. We favor publishing null results when the data supports one.
When we run a prediction, the methodology section explains the assumptions, the confidence bounds, and what would have to be true for the finding to generalize.
Benchmark your sector, audit your workforce AI readiness, or commission a private deep-dive. Typical turnaround: 2-4 weeks.
Start a conversation →