Work
Decisions, measured.
Selected projects where I set the bar, ran the process, and published the numbers. Each one shows the artifact I actually used, not a summary of it.
How I work
The practiceNine rules that hold across every workflow, five stations, three tiers, six bright lines, and the six workflows that run on them. One has a published scorecard.Read it →Pre-registered study
The Meter and the Judge
Cost is stable and bandable. Judge quality is not. What 3,989 measured API calls said about governing AI-augmented knowledge work.
Read the case study →54.9%
judge–human agreement
Method
Requirements-to-Design pipeline
Read-only extraction → committed schema → traced wireframes → operable demo. Five moves, four guardrails, and what did not survive the last hop.
Read the method →19×
smaller than the raw payload