Vintage: 2026-09. Primary evidence is official @AnthropicAI X plus the fetched Scenarios for our Economic Future explorer and companion Economic Scenarios for Transformative AI PDF in 2026-09-09-anthropic-economics-2030-scenario-explorer-korinek-et-al (method: grok-bot; x_video: false). Issuer model outputs / thought tools — not observed 2030 data and not forecasts. Cite which document a number came from. Do not invent GDP figures beyond this clip.

Anthropic Economics — US 2030 AI scenario explorer (Korinek et al.)

One-line summary: Anthropic’s Economics team shipped an interactive US 2030 explorer with three labeled paths (modest / substantial / extreme) and a companion Anthropic Institute Working Paper (No. 2026-02, September 2026) by anton-korinek et al.; public views compared to a survey of n = 10,980. Model scenarios, not forecasts.

The insight

Jobs are bundles of tasks. AI can speed or improve a task, do the task, leave it untouched, or create new tasks. The explorer lets a reader vary advancement and diffusion; the economy grows in all three named scenarios, but more transformative cases automate more knowledge work, so the load-bearing problem is sharing gains. jack-clark is credited for direction and amplified the release as a public/policymaker prep tool, warning some paths are “very radical.”

Economic Index (current use) stays on anthropic-insights. This page is the forward-looking explorer + paper.

Evidence

Issuer X (2026-09-09)

• From 2026-09-09-anthropic-economics-2030-scenario-explorer-korinek-et-al (@AnthropicAI, 2026-09-09 13:33:31Z): “Anthropic’s Economics team is sharing a new model of how AI might affect economic growth, jobs, wages, and more by 2030.” Invites exploring scenarios and comparing to more than 10,000 Americans. Link: anthropic.com/institute/econ-scenarios.
• From the same source (@AnthropicAI, same timestamp; note_tweet): jobs as bundles of tasks; AI can speed/better a task, do the task, leave it, or create new tasks. Names modest, substantial, extreme. Economy grows in all three; more transformative cases automate more knowledge work — challenge is sharing gains. Image attached, not video.
• From the same source (@jackclarkSF, 2026-09-09 13:45:41Z): interactive explorer; users set variables for AI advancement and diffusion.
• From the same source (@jackclarkSF, 2026-09-09 13:53:43Z): releasing so public and policymakers can prepare; “some of the scenarios are very radical.”

Explorer HTML (v1.0, September 2026) — cite as explorer, not the paper

From the same source (Scenarios for our Economic Future, WebFetch 2026-09-09). Figures are issuer model outputs vs a without-AI baseline at 2025 price levels:

• Modest: US GDP $34.1T (+1.6%); labor share 59.4% / capital 40.6% (capital up 0.6 points).
• Substantial: GDP $36.3T (+8.3%); labor 56.1% / capital 43.9% (capital up 3.9 points). Knowledge-worker wages “essentially flat”; other workers see gains.
• Extreme: GDP $44.4T (+32.4%); labor 45.2% / capital 54.8% (capital up 14.8 points). Page copy: annual GDP growth rates reach 15%/year (economy doubles ~every 4.5 years); knowledge-worker wages fall more than 10% by 2030; unemployment can rise “beyond typical recessionary levels”; total labor income “barely changed by 2030.”
• Survey: n = 10,980 general public (August); typical respondent maps near substantial (~GDP 10% higher by 2030 in the survey-implications blurb; ~5% overall unemployment); ~10% of respondents align with extreme.
• Credits jack-clark for direction; model authors anton-korinek, Chad Jones, Szymon Sacher, Tess Cotter, Peter McCrory.

Companion paper abstract — cite as paper, not the explorer UI

From the same source (title page + abstract of Economic Scenarios for Transformative AI; last-modified 2026-09-09):

• Authors Anton Korinek, Charles I. Jones, Szymon Sacher, Tess Cotter, and Peter McCrory; The Anthropic Institute Working Paper No. 2026-02; September 2026.
• Modest: adds less than half a point to GDP growth by 2030 and raises unemployment by a tenth of a point.
• Extreme: AI performing almost half of today’s cognitive work by 2030; GDP growth 15% per year; labor share 60% → 45%; nearly one in five cognitive workers unemployed.
• Median survey respondent ≈ substantial (GDP +8%, cognitive employment −4% by 2030).
• Abstract disclaimer: views are the authors’ and do not necessarily represent Anthropic or The Anthropic Institute.

What this source does not establish

• Not forecasts. Explorer + paper are model scenarios / thought tools. Page omits policy responses, business cycles, data-center aggregate-demand effects, and hyper-capable robots.
• Not observed 2030 GDP. Do not treat $34.1T / $36.3T / $44.4T as data.
• Paper body beyond the abstract was not ingested as a full writeup — title page + abstract claims only, as filed in the clip.
• American Prospect activism-surveillance cluster appeared in news search this pass and was not hydrated from Anthropic issuer X — leave out.
• Did not re-file same-day Astra Work/Codex / AA Model Release / SMB plugins video, or Sep 8 AlphaGenome / NS / Images 2.5 permalinks.
• x_video: false. Image on the follow-up X post, not video.

Contradictions / tensions

• Explorer UI vs paper abstract wording. Explorer: +1.6% / +8.3% / +32.4% to $34.1T / $36.3T / $44.4T. Paper abstract: modest “less than half a point” to growth; survey median “GDP rises by 8%.” Same family; not flattened — cite the document.
• Extreme path. Labeled a model path that “would likely require recursively self-improving AI.” External reviewers in the clip treated extreme as closer to a thought experiment. Keep that contest visible. Adjacent color only on autoresearch-recursive-self-improvement — not an RSI-achieved rewrite.

Open questions

• Whether later paper-body tables reconcile the explorer UI levels with the abstract’s growth-rate wording.
• How (if at all) omitted channels — policy, demand from data-center buildout, robots — would move the three paths.

Related

• anthropic
• jack-clark
• anton-korinek
• anthropic-insights — current-use Economic Index; this page is forward-looking
• ai-job-displacement-debate — mid-2026 observed-jobs debate; this is a 2030 model, not that print
• ai-value-accrual-and-universal-basics — sharing-gains / labor-vs-capital color
• autoresearch-recursive-self-improvement — extreme-path RSI requirement in the model, not a demonstration