Journal

Notes on where the AI budget goes.

1. No. 05How-toClaude Code OpenTelemetry setup: a step-by-step guideTurn on Claude Code's OpenTelemetry metrics the safe way: metrics only, logs and traces off, repository attribution, managed settings for the whole team, and a local collector to verify it works.04 Nov 2026
   9 min read
2. No. 04MetricsAI coding cost per pull request: how to measure it without fooling yourselfCost per pull request is the most intuitive AI spend metric — and the easiest to get wrong. How to compute it from OpenTelemetry cost and pull request counters, and the traps to avoid.03 Nov 2026
   7 min read
3. No. 03For agenciesHow agencies can pass AI coding costs through to clientsA practical playbook for agencies and software houses: measure AI coding spend per client, choose a billing model, word the contract clause, and send a report clients actually accept.30 Oct 2026
   8 min read
4. No. 02How-toClaude Code cost per project: a practical guide to attributionHow to find out what Claude Code costs per project, per branch and per pull request using its built-in OpenTelemetry metrics, repository attributes and a small session hook.22 Oct 2026
   8 min read
5. No. 01Engineering managementAI coding spend by team: how to see who's using what, and why it mattersYour AI coding bill is one number. Here's how to break it down by team, project and developer from OpenTelemetry metrics — and what to do with the numbers once you have them.14 Oct 2026
   7 min read