Anthropic studied its own staff. About 80% of interactive Claude session time is the agent waiting for a human.
Reading everything their agents write would take non-technical staff half the work week.
The bottleneck is the approver, not the model. Design fewer, better asks.
Reading everything their agents write would take non-technical staff half the work week.
The bottleneck is the approver, not the model. Design fewer, better asks.
Oct 24, 2026 · 5:30 PM UTC
2
0
1
50
@datachad That bottleneck shows up in the bill too. A human taking twenty minutes to approve something can outlive the prompt cache. Does the study say how long the waits actually run?
1
0
1
4
