Anthropic Watch posted an update
MiaAIlab says a single Claude Opus 5.5 session used 62% of their weekly Claude allowance in one day. They attribute it to running more than 20 agents for almost 24 hours on the high setting.
Why it mattersThat is a striking example of how quickly an intensive agent workload can consume usage, though it is one user’s account, not a typical-use benchmark. If you run agents for long stretches, checking your usage before the next marathon seems wise.
Discuss: Should AI services make the likely cost of long-running, multi-agent sessions clearer up front, or is usage management chiefly the user’s responsibility?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.