Claude’s new usage logs now read like an early sensor for how AI is entering work.
Rohan Paul Twitter · Rohan Paul (@rohanpaul_ai) · 2026-06-26
Anthropic's 'Cadences' report, based on nearly 10,000 anonymized Claude conversations, reveals temporal and occupational usage patterns showing Claude is more heavily used for personal tasks on weekends, higher-wage work burns more compute, and clear output is produced in 93% of sessions.
Extraction
Topics: ai-usage-analyticsanthropichuman-ai-interactionai-economy
Claims
- Personal prompts to Claude rise from 35% on weekdays to nearly 50% on weekends, with weekend Claude Code work shifting toward AI agent design and quant trading rather than backend architecture.
- Claude produces a clear output in 93% of conversations, with explanations (17%), documents/reports (15%), and guidance (11%) being the most common output types.
- Higher-wage occupations use approximately 2.07x as many tokens per conversation as lower-wage occupations, and app-building conversations use more than 3x the median tokens.
- Real-world temporal events are mirrored in Claude usage: US tax requests spiked 8x before the filing deadline then collapsed immediately after.
- Work done through Claude at nights and weekends skews toward higher-wage occupations, not lower-wage clerical tasks.
Key quotes
Claude's new usage logs now read like an early sensor for how AI is entering work.
Work done through Claude at nights and weekends skews toward higher-wage occupations, not lower-wage clerical tasks.
App-building conversations use more than 3x the median tokens, while basic explanations use about 1/5 of the median.