Anthropic says experienced Claude users tackle higher-value work and succeed more often

On March 24, 2026, Anthropic’s X account highlighted a new Economic Index report examining how Claude usage changes with experience. The post distilled a core finding from the report: longer-term users are less likely to hand Claude full autonomy, more likely to iterate carefully, and more likely to use the system on higher-value tasks.

What the report found

Anthropic’s official Learning curves report studies Claude usage in February 2026 and frames the question as one of learning-by-doing. The company says Claude.ai use became less concentrated, with the top 10 tasks falling from 24% of traffic in November 2025 to 19% in February 2026, while personal queries became a larger share of usage. At the same time, Anthropic reports that higher-tenure users are more likely to choose the more capable Opus class for higher-wage tasks.

The headline claim in the report is that users with six months or more of tenure show a different operating pattern: 10% fewer personal conversations, a 6% higher education level reflected in their inputs, and a 10% higher success rate in conversations. Anthropic says this relationship is not explained away by task selection, country, or other simple confounders, although the company also notes possible cohort and survivorship effects.

Why it matters

This is one of the clearer public attempts to measure how AI skill compounds over time. If the pattern holds, the upside from AI may accrue faster to people and organizations that learn how to scope tasks, iterate with the model, and match model class to problem difficulty. Anthropic also raises the opposite possibility: if better outcomes increasingly cluster around early and higher-skill adopters, AI could reinforce existing gaps in productivity and labor-market outcomes.

Source materials include Anthropic’s X post and the March 24, 2026 Economic Index report.