The BaseResearchAnthropic
Anthropic Economic Index report: Learning curves
Anthropic's Economic Index report, based on 1 million Claude.ai and 1 million first-party API conversations sampled on 5-12 February 2026, finds coding still the largest use of Claude, with personal use rising. It reports that longer-tenured users succeed more often and cautions that self-selection or survivorship may partly explain this.
About Anthropic or ClaudeClaude, made by Anthropic, wrote this entry. It follows the same rules as every other entry, whoever the entry is about.
Why it matters
The same thing, for different readers.
What it means for you
Anthropic's own counts show coding is the largest use of Claude and personal use is growing; longer-term users succeed more often, which could favor early adopters.
For lawmakers
One AI company's usage data, with a warning that unequal skill and access could widen inequality; it measures task use, not jobs lost or pay changes.
For researchers
Comparisons within the same task find longer-tenure users succeed more, but success is Claude's own rating and the authors cannot separate learning from self-selection or survivorship.
The question it raises
The authors suggest benefits of early adoption may be self-reinforcing, raising a fairness question about who gets the chance to learn to use AI well.
In their own words
Every quotation was found, word for word, in the text we saved from the source. That shows the words are there; it does not make them true, or prove the speaker meant them as quoted.
From the source’s own text.
“Coding remains the most common use on our platforms, with tasks associated with Computer and Mathematical occupations accounting for 35% of conversations on Claude.ai”
Found word for word in the source
“About 49% of jobs have seen at least a quarter of their tasks performed using Claude.”
Found word for word in the source
“Usage remains heavily concentrated: the top 20 countries account for 48% of all per-capita usage, up from 45%, underscoring a persistent gap in global adoption.”
Found word for word in the source
“These results suggest that high-tenure users have more success in their Claude conversations, and that this is not due to simple factors like language or the task being performed.”
Found word for word in the source
“Further, there’s an inherent survivorship bias: people who signed up a year before our data pull may be seeing positive results from their usage. We do not observe people who signed up a year ago but are no longer using Claude.”
Found word for word in the source
How the finding was reached
In our words, from the source. Weigh the finding by how it was made.
- What was done
- Company report from Anthropic's Economic Index. It classifies a sample of 1 million conversations from Claude.ai and 1 million from Anthropic's first-party API, collected on 5-12 February 2026, by the occupational tasks (O*NET) they involve, using a privacy-preserving analysis system, and compares them with earlier samples (November 2025 and August 2025). It also compares users who signed up at least six months earlier with newer users, with regressions that control for task, model, use case and country.
- What it found
- Tasks tied to computer and mathematical occupations were 35% of Claude.ai conversations, and about 49% of jobs have seen at least a quarter of their tasks performed using Claude, barely changed since November 2025. The ten most common tasks fell from 24% to 19% of Claude.ai conversations; personal use rose from 35% to 42% and coursework fell from 19% to 12%. The average hourly wage of US workers who do the tasks seen on Claude.ai fell from $49.3 to $47.9. The top 20 countries account for 48% of per-capita usage, up from 45%. Users who signed up at least six months earlier had a success rate about 5 percentage points higher in a simple comparison, about 3 within the same task and about 4 with full controls (the report's summary calls this a 10% higher success rate). The authors say such differences in success could deepen inequalities in the labor market.
- What it does not show
- Describes one company's users over one week in February 2026, which overlapped with a Super Bowl advertising push that brought many first-time users and, in some countries, a school winter break (which the authors say explains part of the fall in coursework). It counts which tasks Claude is asked to do and values them with US wage data; it does not observe employment or pay outcomes. Success is Claude's own assessment of whether a conversation succeeded. The authors say the experience-and-success pattern may reflect self-selection (early adopters may be more technical) or survivorship bias, since users who left are not observed, and that they cannot yet separate these from learning by doing.
- Status
- The company’s own report about its own work
The record
- Questions
- What happens to work, income and the economy?
- People
- Maxim Massenkoff (Wrote it, Lead author, Anthropic Economic Index report) · Eva Lyubich (Wrote it, Lead author, Anthropic Economic Index report) · Peter McCrory (Wrote it, Lead author, Anthropic Economic Index report) · Ruth Appel (Wrote it, Author, Anthropic Economic Index report) · Ryan Heller (Wrote it, Author, Anthropic Economic Index report)
- What we read
- We read the whole source.
- Found by
- A research pass: work, income and the economy, gathered October 13, 2026, via
research-work(tier 1: the source’s own channel or an official record) - Cite this
- Anthropic, “Anthropic Economic Index report: Learning curves”, March 24, 2026. Entry in the Ctrl AI Base, https://ctrlai.com/base/anthropic-2026-03-24-economic-index-learning-curves
Our notesAnthropic's own report about its own product, so the entry is marked as about Anthropic and written by Claude. The summary section says high-tenure users have a 10% higher success rate, while the regression section reports 3 to 5 percentage points; the text read does not reconcile the two (the appendix is a separate page and was not read), so check whether 10% is a relative figure. The page credits authors without individual titles and marks the first block as lead authors. The report refers to a separate labor-market-impacts report and to a previous Economic Index report from January 2026; neither was read here.