← Selected work Developer Delivery Instrumentation
Pathmeter Built a local-first CLI that measures idea → scaffolded → first-run → validated → shipped cycles, runs real quality gates, and identifies the recurring delivery phase that is actually slow.
Public technical artifact Problem
What this was built against. Developers often optimize scaffolding and generation even when the expensive delay is later in first-run, validation or shipping.
System
How the work closes the loop. Feature commitment → five milestone events → configured validation commands → ship refusal on failed gates → repeated-cycle bottleneck report.
Evidence now
What an employer can safely inspect. Public repository documents a five-stage golden path, JSONL event log, local dashboard and validation gate recording. The design explicitly refuses to call a failed validation state shipped. The roadmap requires measurement evidence before adding more automation. Claim boundary
What this does not prove. No external developer cohort or before/after delivery-time result is claimed. The strongest next proof is repeated real feature cycles, not more product surface. Next evidence event
What would upgrade the proof. Instrument 7–10 real feature cycles, identify the recurring bottleneck, automate exactly one repeated gap, then compare before/after cycle time.
P2 — inspectable CLI / measurement system · Canonical case-study page v1