Our own AI bill, published
$412.88 to build the Analytics brick, every run itemized, since we ask you to work this way.
We ask you to run inference on your own key, in the open, and to watch the meter as it climbs. It would be strange to hide ours. So here it is.
Building and verifying the Analytics brick cost $412.88 across 61 runs. Not a rounded figure, not a range, the actual number off the provider invoice.
Where it went
- Audit passes were the biggest line. Reading a repo well is expensive, and it’s where the quality of everything downstream is decided.
- Plan drafting was cheaper than expected once we stopped re-reading context the audit had already produced.
- Verification probes were a rounding error per run, and the best money we spent.
The 30% we ground out
About a third of early spend was waste: re-crawling files we’d already parsed, re-running an agent because a later step lost its context, retries that a better checkpoint would have avoided. Every one of those is a thing you should never have to pay for, so we hunt them down and cut them, and the build log is where we show our work.
No markup, no credits, no bundle. When the meter runs on your key, we have the same incentive you do: make each run count.