@premsai-pendela· 3d
Marathon debugging on Agent
Agent
MARATHON
16.5x your typical session
1.2M tokens vs typical 71.7K
1.2Mtokens
78prompts
8:29:20time
4,444lines
Claude Codeclaude-sonnet-4-6Python$19.57
mostly building
8h 29m session running commands and testing. 1.1M tokens across 78 prompts using claude-sonnet-4-6.
Reconstructed from session history
Tokens / prompt15.1k
Cost / line$0.004
Cache hit98%
Burn rate$2.31/hr
Derived from this session's token and cost data. Not shown on the feed.
Building 100%