# Source note — Anthropic Economic Index report: Cadences

- **Managing Expectations slug:** `anthropic-economic-index-cadences`
- **Date prepared:** 2026-06-29
- **Evidence label:** company economic research report / AI labor-use telemetry
- **Primary source:** Anthropic Research — “Anthropic Economic Index report: Cadences,” published Jun. 26, 2026
- **Primary URL:** https://www.anthropic.com/research/economic-index-june-2026-report
- **PDF URL surfaced by page:** https://www-cdn.anthropic.com/7b76335c444876a93fa22a63aabb4aeb820aff25.pdf
- **Library placement:** AI Papers Library / Anthropic lane / Societal and economic impacts

## Why this item was selected

This is a meaningful update for the AI Papers Library because it is not a social-media forecast or valuation headline. It is a primary-source Anthropic report describing how Claude usage is changing as Claude Code and Cowork shift some sessions from simple chat into longer-running agentic work. It also reports initial results from the Anthropic Economic Index Survey launched in April 2026.

## Key source-visible points

From Anthropic’s report page:

- Anthropic says it changed the Economic Index data pipeline by sampling at a higher rate, introducing a new classifier for the output of each conversation, and sharing more granular monthly data split across chat/Cowork conversations and first-party API usage.
- The report says Claude usage mirrors the workweek: personal-use conversations rise from about **35% on weekdays** to **just under 50% on weekends** during the sample period.
- The report says hourly usage reflects daily routines: news requests cluster around **7 a.m. local time**, recipe requests are **2.3× more frequent at 6 p.m.** than average, and sleep-advice requests peak in the hours before dawn.
- Anthropic says Claude Code weekend clusters shift away from backend architecture, API debugging and data storage, and toward AI agent design, quant trading and gaming.
- The report says more compute is associated with more valuable artifacts: token consumption rises with Anthropic’s estimated value of the work output.
- The survey preview says users who use Claude in the most automated way expect AI to take on more of their tasks in the next year, while also reporting more optimistic expectations for pay, job security and work meaning.

## What the source does not prove

- It does not measure the whole economy; it measures Anthropic/Claude usage patterns and survey responses.
- It does not prove that AI automation is net-good or net-bad for workers.
- It does not prove that all Claude users are representative of all workers, companies, countries, or occupations.
- It does not replace independent labor-market data, payroll data, productivity studies, or cross-platform comparisons.

## Managing Expectations framing

Treat this as a useful early instrument panel, not as a verdict. The report shows that AI use already has a cadence — weekdays, weekends, morning news, evening recipes, tax deadlines, coding sprints, higher-wage after-hours work — and that more agentic usage changes what needs to be measured. The practical takeaway is not “AI will replace everyone” or “AI makes everyone happier.” The better takeaway is: frontier-AI deployment is now measurable enough that leaders should track usage, outputs, automation level, worker experience and distributional effects separately.

## Sources checked during the 2026-06-29 cron run

- Anthropic Research feed: https://www.anthropic.com/research — surfaced this Jun. 26, 2026 Economic Index report as the newest meaningful Anthropic research item.
- Anthropic report page: https://www.anthropic.com/research/economic-index-june-2026-report — used as primary source for article facts.
- OpenAI Research index: https://openai.com/research/ — checked; no newer comparable Jun. 2026 technical report or system card surfaced in the fetched page excerpt.
- Google DeepMind News: https://deepmind.google/discover/blog/ — checked; several June 2026 items surfaced, but none was selected over the Anthropic report for this run.
- LawZero: https://lawzero.org/en — checked; latest visible news included ICML 2026/recruiting and governance/team updates, not a new technical paper selected for this run.
- Transformer Circuits: https://transformer-circuits.pub/ — checked; latest visible May 2026 interpretability updates were already within the AI-library watch lane.

## Editorial caution

This source is company-authored and based on Anthropic’s own product telemetry and survey pipeline. Use the label `company economic research report / AI labor-use telemetry` and preserve the caveat that it is a partial view of AI use, not a full labor-market finding.