Real coding sessions · never self-reported
Where the tokens go
50% building · 44.1M classified tokens
7 builds predate classification and aren't included.
Autonomous session on debug_notebook
2.3M tokens
Autonomous 44h 15m agent session using claude-opus-4-8. 108K tokens processed.
Autonomous session on debug_notebook
Welcome back
Autonomous 0s agent session using claude-sonnet-4-6. 2.2M tokens processed.
Agent verified and completed BigFix database migration
Welcome back
Completed BigFix database migration with end-to-end verification, confirming all data differences were benign.
Agent debugged notebook performance bottlenecks
First build tracked for this project
Identified recursive filesystem bottleneck causing delays and refactored processing functions to use API pagination.
Updated release notes with new data quality features
First build tracked for this project
Updated the changelog to document new data quality capabilities.
Agent architected org directory sync and began implementation
First build tracked for this project
Designed org directory sync feature and began store implementation.
Agent synthesized deployment platform evaluation research
7.6M tokens
Produced comprehensive evaluation report with weighted decision matrix and diagrams.
Debugged production
27 prompts
1h 19m session running commands and testing. 728K tokens across 27 prompts using claude-opus-4-8.
One-shot exploration on diff3
First build tracked for this project
3m session reading and analyzing the codebase. 89K tokens across 1 prompts using claude-opus-4-8.
Debugged pipeline
1937 sessions tracked this week across all builders
1h 7m session running commands and testing. 625K tokens across 8 prompts using claude-opus-4-8.
Explored cgc
5m session reading and analyzing the codebase. 200K tokens across 3 prompts using claude-opus-4-8.