Skip to content
Slow Variables

Ask the data

Answers come from the same store as the site; every number is checked against the record it cites.

Early adoption · Adopters

Usage as a share of theoretical exposure

consistent with normal60/95 confidence, mixed or hard to operationalisegrade Bcoincident

1What this measures

Weekly work use of generative AI (Bick-Blandin-Deming) divided by the share of the US workforce with at least a tenth of tasks exposed to LLMs (Eloundou and co-authors' "GPTs are GPTs"), derived metric exposure_vs_usage_gap.

Why it matters. The paper's central gap - what the technology could touch versus what people actually use it for weekly. Exposure is a ceiling; usage is the diffusion curve under it.

Proxy types
behaviour
Unit
ratio
Cadence
quarterly
Valve
product to adoption

2How we track this

  • series fred.us_workers.work_use_weekly_share.q
  • series arxiv.us_workforce.llm_exposed_10pct_tasks_share.pt
  • derived metric exposure_vs_usage_gap (formula in the semantic layer)
  • source arXiv abstract pages · default tier 6 · arXiv; abstract quotation
  • source FRED (Federal Reserve Economic Data) · default tier 6 · FRED terms of use; RPS series by Bick, Blandin & Deming
Normal band
≤ 0.60×
Fast band
≥ 0.80×
Falsifying
—

Normal = under 60% of exposed workers use the tools weekly three years in; fast = 80% or more, the gap closed. Between is `emerging`.

Applied to metric:exposure_vs_usage_gap.

3Tracker interpretation

Roughly half of the exposed workforce uses generative AI weekly; the ratio has climbed steadily since 2024.

4Evidence

Latest point
0.49×as of 2026-06-30(2 obs)
Value the bands apply to
0.49×as of 2026-06-30(2 obs)

9 observations. Hollow points are disputed (see counterevidence). Every point links to its observation.

Derived rows (8)
as ofdimsvalueinputs
2026-06-300.49×obs:115415a8obs:72b0b6f7
2026-03-310.47×obs:115415a8obs:8d4f8bd1
2025-12-310.44×obs:115415a8obs:56e28ca7
2025-09-300.40×obs:115415a8obs:58f79eef
2025-06-300.38×obs:115415a8obs:8aa0a471
2025-03-310.37×obs:115415a8obs:2c173e2d
2024-12-310.33×obs:115415a8obs:972d9250
2024-09-300.35×obs:115415a8obs:3fa468ca

5Status and reasoning

consistent with normalsince 2026-09-10 · evaluate

Evaluator: 0.4904 (arXiv abstract pages, 2026-06-30) is inside the consistent band (hi=0.6). Auto-reason; band rationale: Normal = under 60% of exposed workers use the tools weekly three years in; fast = 80% or more, the gap closed. Between is `emerging`.

6Timeline notes

  • 2026-06-30 · 0.49×as of 2026-06-30(2 obs)
  • 2026-03-31 · 0.47×as of 2026-03-31(2 obs)
  • 2025-12-31 · 0.44×as of 2025-12-31(2 obs)
  • 2025-09-30 · 0.40×as of 2025-09-30(2 obs)
  • 2025-06-30 · 0.38×as of 2025-06-30(2 obs)
  • 2025-03-31 · 0.37×as of 2025-03-31(2 obs)

7Counterevidence

What cuts against this reading

The denominator is a 2023 model-based exposure estimate; the numerator counts any weekly use, however shallow; exposure and use are measured on different populations.

8Update history

  1. 2026-09-10unmeasured to consistent with normalconf — → 60 · evaluate

    Evaluator: 0.4904 (arXiv abstract pages, 2026-06-30) is inside the consistent band (hi=0.6). Auto-reason; band rationale: Normal = under 60% of exposed workers use the tools weekly three years in; fast = 80% or more, the gap closed. Between is `emerging`.

    obs:115415a8obs:72b0b6f7

9Confidence

60 / 95 — mixed or hard to operationalise

Confidence is independent of status: 90–95 multiple strong independent sources; 70–89 good evidence, some ambiguity; 50–69 mixed or hard to operationalise; below 50 limited or vague.

10Related