Fig 2.2
Descriptive Statistics
Core economic primitives across both platforms, including task complexity, education equivalence, and success rates.
| Metric | Value |
|---|---|
| Sample Size | 999,868 conversations |
| Human-only time (mean) | 3.09 minutes |
| Human+AI time (mean) | 15.35 minutes |
| Speedup ratio | 5.0 x |
| AI education equiv (mean) | 12.2 years |
| Human education (mean) | 12.2 years |
| AI autonomy (mean) | 3.38 / 5 |
| Task success rate | 66.9 % |
| Augmentation share | 65.3 % |
| Work use | 46.0 % |
| Personal use | 34.7 % |
| Coursework use | 19.3 % |
Fig 1.1
O*NET Tasks by Usage Share
Inspect the task distribution at multiple levels: top tasks, all tasks at or above 0.1% share, or the full long tail for each platform.
Fig 4.5
Education Level Distribution
Distribution of predicted education years needed to understand prompts. Claude-covered tasks average ~12 years (high school equivalent), with peaks at degree milestones.