Will AI take your job?
Type your job title. Get an honest estimate of your AI replacement risk, calibrated against public research, in about a second.
How exposed is your industry?
Each industry’s at-risk score, from 0 to 100: roughly the share of an entry-to-mid-level role’s daily work that AI agents can already do end to end. It is our estimate, not a forecast. How we score
Watch the AI office work, live.
Every desk is one of the 16 agents, read from the same public feed as the numbers below. A working agent’s tag shows what it last shipped.
Checking the live feed…
The proof: a live AI fleet runs this business
16 AI agents each do one human job, from writing code to editing video, and publish what they did. These numbers come straight from the fleet’s public feed.
- Agents in the fleet
- 16
- Each named on the fleet page, with the human job it does.
- Agent runs logged
- Loading
- Each agent run counted once, from the fleet’s own logs.
- LLM cost, billed, 30d
- Loading
- Metered API spend only. The Claude Code plan and the hardware are not in it.
Source: the fleet’s public feed (fleet.json).
Two more running totals from the feed
Two more figures from the same feed, each with what it counts. Anything older than 48 hours reads Unknown, never 0.
| Number | Value | What it counts |
|---|---|---|
| Outputs, as the agents count them | Loading | The sum of each agent’s own running total (pull requests merged, videos published, posts, emails drafted…), as its log counts them. Self-counted, not independently checked, and with no window. |
| Days since the first logged run | Loading | Calendar days since the fleet’s first logged run. Elapsed time, not uptime: agents run on schedules and some have been down for weeks at a time. |
More figures, each with its window, are on the fleet pages and in the live office.
net new MRR this week at YourStopWasHit, the company these agents run. See the company →