>Building AI Governance & Games
Real coding sessions · never self-reported
Where the tokens go
40% building · 2816.0M classified tokens
2248 builds predate classification and aren't included.
Debugged SDLC loop termination, identified environment blocking verify
8.8M tokens vs typical 3.7K
Diagnosed why automation exits before verification; environment issues prevent test execution.
One-shot exploration on dirigible2D
721.5K tokens vs typical 3.7K
50s session reading and analyzing the codebase. 230K tokens across 1 prompts using claude-sonnet-5.
One-shot exploration on dirigible2D
230.0K tokens vs typical 3.7K
5m session reading and analyzing the codebase. 721K tokens across 1 prompts using claude-sonnet-5.
Agent validated VG infrastructure implementation plan and held at gate
3.4M tokens vs typical 3.7K
Prepared and validated Visual Graph infrastructure plan; correctly held pending approval decisions.
Finalized M06 Combat roadmap with 18 implementation tickets
15.5M tokens vs typical 3.7K
Formalized M06 Combat roadmap into 18 implementation tickets and validated M06 P0 Combat Core milestone structure.
Formalized Combat Core and organized MVP into three execution pushes
2.7M tokens vs typical 3.7K
Organized MVP roadmap into three sequential pushes with execution-ready M06 detail and target dates set.
Shipped M04 with complete API documentation coverage
3.2M tokens vs typical 3.7K
Closed M04 milestone with comprehensive public API documentation and zero Doxygen warnings.
Marathon debugging on dirigible2D
Previous best: 14.6M · +59%
9h 9m session running commands and testing. 23.2M tokens across 65 prompts using claude-sonnet-5.
Restructured M5.1 Infrastructure roadmap with sequential phases
911.0K tokens vs typical 3.7K
Reorganized M5.1 infrastructure roadmap phases and created initial ClickUp task mapping.
Shipped sub-milestone support to roadmap planning tool
2.1M tokens vs typical 3.7K
Fixed and deployed the roadmap tool to handle phase-based milestones; verified against live roadmap.
Shipped room classification system with event-triggered moodlets
10.7M tokens vs typical 3.7K
Delivered complete room classification system; rooms now identify by contents and trigger contextual moodlets.
Room type text pipeline and colonist moodlet integration
82.4K tokens vs typical 3.7K
Completed a focused local build session that added room type import/export coverage, expanded editor tests, and wired colonist need and task logic to room-event moodlets.