Observed arrival · 2026-09-24
Can Agents Work? maps benchmarks onto 997 occupations
A source-linked atlas maps AI agent benchmark results across US occupations and job families.
○Open
⊠Login
$Paid
†Ads
✦Pretty
●Pro
◎Niche
◉Human
⚑Risk
ƒJS
- For
- Readers tracking AI’s impact on work
- Worth noticing
- The atlas flags 13 benchmarks as saturated, solved, or retired, acknowledging that its picture of the frontier has a moving edge.
Field notes
The atlas groups US occupations into job families and distinguishes the largest unit of work completed on direct benchmarks from the share of work those benchmarks actually measure. Its displayed totals include 70 benchmarks from 62 institutions and 252 source-linked results; it also marks 13 benchmarks as saturated, solved, or retired. These are benchmark-coverage indicators, not direct proof that an entire occupation can or cannot be automated.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
○OpenPublic substance visible
✦PrettyNotable craft visible
●ProPolished or operationally mature
◎NicheUnusually specific use
One card from the complete issue