Attila
What these tokens would have cost billed per token at each model's published price. Not a bill. It is the one figure that compares across people on different plans. Free profiles show the last 30 days.
KPI figures on Free cover the last 30 days.
Activity
Last 30 days2.98B tokens over 25 active days · last 30 days
Averaging 119.2M on a day worked, with a median of 63.4M. Busiest was Jul 21, 2026 at 380.7M. Hover a day for the models and harnesses behind it.
Last 30 days in full on this profile; older days are muted on Free, last synced Aug 3, 2026. Counters only: no prompts, no project names, no file paths.
Where it goes
$1,739, split four waysLast 30 days
Output is a fraction of the tokens and usually most of the money, because it is priced at several times the rate of anything read back out of cache.
Every model in this archive has a known rate.
Cache
Reuse, not volumeLast 30 days
A raw token total mostly measures how much context got re-read. The ratios underneath it are the part that says anything about how the work was done.
Models
10 models, mostly claude-sonnet-5Last 30 days
Tools
5 harnesses, one archiveLast 30 days
Rhythm
Heaviest on SundayLast 30 days
Tokens by day of the week, in the timezone each machine reported. Last 30 days on Free.
By month
2 months tracked, busiest Jul 26.
Hour of day is not in the sync payload yet, so the rhythm here stops at the day. See docs/tokenstat-sync-cli.md for what the CLI would need to send.
Spread
Across 24 projectsLast 30 days
Project names stay on the machine. What reaches us is an opaque hash we can count but never read, so a profile can show how the work was spread and never which work it was.
Ordered by size, unlabelled on purpose. The top project is 39% of everything.
How the counts were read
Some harnesses report token counts exactly; others have to be derived from what they write to disk. The archive keeps that distinction rather than flattening it, so a number that was inferred never passes itself off as a measurement.
