What Anthropic Found When Its Own Staff Started Running Claude Agents
Anthropic measured what happened inside its own company once Claude agents did real work. The numbers, why most of your Claude time is spent waiting, three habits that follow from it, and the catches in the data
On October 16, Anthropic's Economics team published a study of its own company. It tracked how work changed once Claude agents started doing real tasks next to staff.
Here are the numbers, what they say about your own Claude time, three habits that follow, and the catches.
The numbers
| What Anthropic measured | Result |
|---|---|
| Hours worked by agents vs. human staff | Agents passed humans in March 2026; by August, 6x as many hours |
| Code commits per engineer | About 5x their level six months earlier |
| Time in a Claude session spent idle, waiting for a human | About 80% |
| Sessions spent building or watching agent tooling | 10% up to 22% |
| Junior share of technical staff | 21% down to 14% since early 2025 |
| Junior share of technical hires | 27% down to 9% |
Engineers who adopted coding agents added more code, with no measured difference in code removed or reverted.
What it means for you
The biggest number is the idle one. If about 80% of a session is Claude waiting for you, the limit isn't how fast Claude works. It's how much work you have queued up for it.
Free, for your email
Get the full setup steps + the copy-paste setup prompt
The rest of this guide (3 more sections, 1 copy-paste prompt) opens right here. The setup prompt, one paste into Claude that sets it all up, comes by email only. Use your real address.
50,000+ follow where these guides come from. No spam, unsubscribe in one click.