Cable commissioning · browser-local field record · 2026 · Solo — product framing, test-first parsing logic, interaction design, accessibility, responsive verification, production deployment
Patchproof
An industrial continuity bench that turns a loose endpoint list into an explicit, portable cable-walk record.
The problem
Small AV and network crews commonly begin a commissioning walk with an endpoint list, mark physical test results somewhere else, and then reconstruct a fault handoff from those separate scraps. Patchproof makes that bounded field loop causal: paste up to 24 FROM > TO pairs, reject malformed or reused endpoints before assigning stable cable IDs, activate each full-width run to cycle it through untested, pass, and fault, then copy the same visible state as a plain-text commissioning record. It records operator observations rather than measuring electrical continuity or certifying installed cabling. Inputs and observations stay in the tab; there is no upload, account, persistence, paid API, tracking SDK, email, payment, or destructive operation. Production deployment dpl_DaNTbgmdgTjTi9E9yvJrR3f7N18R reached READY before independent Foundry harness checks proved HTTPS markers and the seven-step commissioning interaction at the stable alias.
Architecture
Key decisions
Make the physical run the control
Each endpoint pair is rendered as one cable-like circuit, and the entire row is its state button. The wire, endpoints, status, completion percentage, and record update from one operator action.
Fail the manifest closed
Malformed lines, blank endpoints, case-insensitive endpoint reuse, and manifests beyond 24 runs produce line-addressed errors without partially replacing the current board.
Keep the measurement claim narrow
Patchproof records the technician's stated observation. It does not connect to test hardware, measure continuity, certify cable category, preserve history, or coordinate multiple technicians.
Use an industrial Operate bench
Warm ruled stock, a near-black handoff panel, safety-yellow controls, continuity green, fault red, Georgia display copy, and monospaced field evidence create direct manifest-to-run-to-record causality without feature tiles, a generic center stack, blur, or dashboard data slop.
Repair the race exposed by the harness
The first timed local walk found old run controls remained clickable during a 260-millisecond rebuild. Clearing the board immediately on submit made the loading boundary explicit; the final local and production seven-step paths passed.