Guides for engineering leaders

Your team has AI coding tools. The open question is whether anyone is getting good with them — and your current dashboards can't answer it. These guides are the playbooks we use for measuring and improving AI-assisted engineering: what to measure, what the standard frameworks miss, and how to coach the gap closed.

A note on where this comes from. We build software that parses real Claude Code sessions and classifies the practices inside them — planning before coding, delegating to subagents, managing context, committing work. The guides are written from what session data makes observable, not from vendor surveys. The caveats are included, because the honest answer to several questions here is "it depends" or "we don't have public numbers yet." We're pre-launch; treat these as working playbooks from a practitioner, not settled science.

Start with the measurement guides if you own the budget question. Start with the coaching guides if you own the outcome.

Measuring

How to Measure AI Coding ROI (Beyond the Accept Rate) — For the leader who has to justify the spend. A layered measurement model, and why accept rates and seat counts mislead: they measure usage, not value. You'll leave with a defensible way to answer "is it working?"

Adoption Is Not Proficiency — The thesis essay. Everyone has a license; that tells you nothing about who's good with it. Explains the 4x proficiency gap between your strongest and weakest AI-tool users, and why no dashboard you have today can see it.

Developer Productivity Metrics in the AI Era — For teams already running DORA, SPACE, or DX Core 4. Where those frameworks stop, why AI-coding skill is invisible to them, and what a session-practice layer adds underneath.

Improving

Rolling Out AI Coding Tools: The Phase Everyone Skips — Most rollouts end at deployment. This guide covers the proficiency-development phase that should come after — the part that determines whether the tools were worth buying.

Coaching Engineers on AI Tools — The practice curriculum and how to run the coaching loop: observe what your strongest people do differently, spread it, re-measure. For EMs and platform leads doing the actual work.

AI Coding Best Practices That Show Up in Session Data — Eight practices that are observable in real sessions, with the mechanism behind each and the failure modes when they're skipped. Useful for individual engineers too.

Beyond the playbooks

Guides tell you how. Research is where we publish original findings from session data — what teams actually do, not what they should.

If you want the measurement itself, join the early access list. If you want to shape what we measure, work with the founder directly as a design partner.