"Across eight months of data (V1-V3), computer and mathematical tasks still dominate Claude.ai at 36%, but education (from ~9% to ~12%) and science (from ~6% to ~7%) grew. New features (web search and Research mode) nudged usage toward information retrieval and research. Coding composition changed: program creation rose while debugging fell, suggesting users accomplish more in a single pass. Collaboration patterns also shifted: directive "automation" interactions increased from 27% to 39%, overtaking augmentation for the first time."
"Anthropic's AI Usage Index (AUI) - usage relative to working-age population - shows a strong correlation with income. Singapore (≈4.6×) and Canada (≈2.9×) over-index; India (≈0.27×) and Nigeria (≈0.2×) under-index. In the U.S., DC (3.82×) leads per capita, with Utah (3.78×) close behind. As adoption deepens, usage diversifies beyond coding into education, science, and business tasks. Notably, high-AUI countries favor augmentation, while low-AUI regions lean toward automation, even after controlling for task mix - hinting at cultural and structural differences in workflows."
AI diffusion is accelerating, reaching 40% of U.S. employees using AI at work, up from 20% in 2023. Claude.ai usage shifted toward creation and delegation, with computer and mathematical tasks still at 36% while education rose from about 9% to 12% and science from about 6% to 7%. New features like web search and Research mode increased information retrieval and research usage. Coding changed as program creation rose and debugging fell, and automation interactions grew from 27% to 39%, overtaking augmentation. Higher-income regions show higher per-capita usage; enterprise API usage is automation-heavy at 77%.
Read at Medium
Unable to calculate read time
Collection
[
|
...
]