Decisions
The reasoning, not just the result
2026
- UX-192decision
Measure the host before building the model
A small model trained to say what a healthcare payload is — which standard and version, which workflow, which direction, which actors, where it is going — failed its own holdout. What should the next build be?
- UX-184postmortem
A check that cannot see the thing it is trusted for
An external FHIR subject-matter expert pressure-tested the public product on 2026-08-27 — one day after an internal truth programme closed GO — and returned "Not Yet": two Critical findings, eight High, five Medium, one Low, with a recommendation not to let engineers rely on the tool for real implementation decisions. Every one of the seventeen was measured against the code rather than accepted, and every one was real at some level: 10 CONFIRMED, 7 PARTIAL, 0 REFUTED. Both Criticals were wrong-workflow defects on the public teaching surface — an inquiry fixture validated as a submission, and an adaptive-form transcript that taught an invented operation contract.
- UX-164decision
Conflict is an annotation on an authoritative edge; 'weak' is not a tier
An owner review of the Payload Workbook asked for five relationship classes — Confirmed pair, Likely related, Weak, Conflict, Unpaired. Three shipped. The two that did not, Weak and Conflict, have no engine source and the correlation design carries explicit refusals against both. Should those refusals be reversed?
- CAN-103postmortem
A draft gate that looked like an environment check and was not one
Draft library articles were invisible on the preview channel — the one deployment where a draft is actually useful — and visible only under `next dev`.
- SEC-113decision
Shipping the method to the browser IS the privacy guarantee
The payload washer runs client-side, so its entire rule set — every pattern, the word lists, the policy matrix — is delivered to every visitor. Can the quality of those rules be treated as proprietary?
- UX-109decision
A true argument on the wrong axis: cost is not readiness
A tool that makes zero network calls costs nothing to serve. Is that sufficient grounds to promote it to the production surface?
- WASH-107bdecision
Provider and facility identity is kept, permanently — and saying so is the control
The payload washer replaces patient identity with synthetic equivalents. Should it do the same to practitioners, organizations and facilities?
- WASH-110decision
Most of a compliance method is the wrong amount to implement
One geography mode keeps the first three digits of a US postal code. The published de-identification method it resembles additionally requires that the small number of low-population three-digit prefixes be zeroed. The list is static, short and cheap to implement. Should it be?