writing

What we measured.

Notes from building this: what the agent costs per request, what it refuses to collect and why, and what the numbers looked like when a deploy went wrong. Every post has something in it we had to run something to find out.