@srmcguirt· 5h
Marathon debugging on agent-swarm
agent-swarm
DEEP WORK
18411.0x your typical session
8.7M tokens vs typical 472
8.7Mtokens
51prompts
1:30:30time
1,279lines
Claude Codeclaude-opus-5Markdown$85.20
mostly building
1h 30m session running commands and testing. 8.6M tokens across 51 prompts using claude-opus-5.
Tokens / prompt170.4k
Cost / line$0.067
Cache hit85%
Burn rate$56.49/hr
Derived from this session's token and cost data. Not shown on the feed.
Building 73%Verifying 27%
Comments