Cost
What it costs to produce this site.
Every piece and page was written and checked with AI models. This panel shows the real cost of each one, read from the session logs, with no estimates.
What is measured
- 01Input and output tokens
- Summed across every session that touched the page, including cache reads.
- 02Cost in US$
- Tokens × each model’s public price on the session date.
- 03Split by model
- How much of the cost was Opus, Sonnet and Haiku. Shows whether the work went to the right model.
- 04Cache hit rate
- Share of input tokens read from cache. Higher is cheaper.
- 05Session time
- Summed session minutes, first turn to last.
- 06Rounds to approval
- How many times the piece came back from checking before it was published.
By page
Nothing here. Yet.
The numbers go in once this production’s logs are processed. No value is estimated or rounded to look better.
How to generate
python3 tools/cost-audit.py --window 90 --verbose