Real coding sessions · never self-reported
Where the tokens go
36% building · 3528.0M classified tokens
72 builds predate classification and aren't included.
One-shot debugging on jeremy
13.8M tokens vs typical 1.7M
4m session running commands and testing. 516K tokens across 1 prompts using MiniMax-M3.
Debugged startaitools
326 sessions tracked this week across all builders
28m session running commands and testing. 3.3M tokens across 2 prompts using claude-opus-5.
Delegated three review agents and staged the validated blog post
277 sessions tracked this week across all builders
Orchestrated three review agents to validate today's post and stage it for publishing.
Shipped a daily engineering journal entry through full verification
271 sessions tracked this week across all builders
Compiled a write-up on the auth rollup fix and cleared every verification gate before publish.
Stood up a blog-backfill run with a delegated classifier
2641 sessions tracked this month across all builders
Inspected the workspace, kicked off a classifier subagent, and began the next pipeline stage after it completed.
Debugged source
First build tracked for this project
25m session running commands and testing. 5.8M tokens across 4 prompts using MiniMax-M3.
Organized service registration intake and filing process
256 sessions tracked this week across all builders
Created structured intake checklist and coordinated first service registrations with Anthropic.
Shipped Longbox customer case study and partner data
70 prompts
Published Longbox customer story with testimonial and verified partner certification status.
One-shot debugging on jeremy
Previous best: 4,000 lines
5m session running commands and testing. 495K tokens across 1 prompts using MiniMax-M3.
One-shot session on comehomealabama
Previous best: 57.3K · +12%
2m coding session using gpt-5.6-sol. 64K tokens across 1 prompts.
One-shot session on comehomealabama
Previous best: 57.2K · +0%
2m coding session using gpt-5.6-sol. 57K tokens across 1 prompts.
One-shot session on comehomealabama
First build tracked for this project
3m coding session using gpt-5.6-terra. 57K tokens across 1 prompts.