Work
The work, with the mechanism shown.
Case studies: what the work was, how it actually worked, and what I would do differently.
· Case study
A green run is not proof: building an agent platform around silent failure
A production coaching agent for a sim racing team, built solo on Cloudflare and Claude. Every control in it exists because something failed without saying so.
Read the case study
$2.06
Cost per debrief run
639
Sessions discarded
8
Incidents written up
2 of 10
Replay eval bundles
· Case study
Replacing guessed delivery dates with calibrated forecasts
An agent-native forecasting tool for engineering managers that grew from a set of Claude Skills into an internal MCP server, and ended up inside the org's roadmapping process.
Read the case study
80%+
Manager adoption
75%+
Inside the p85 target
1 in 10
Delivered ahead of forecast
3
Targets, not one date