Blog
Field notes on enterprise delivery, reliable systems, local AI, and what it takes to make software useful in production.
Explore by topic
7 articlesAug 01, 2026Graceful Degradation Is Good Product Design: Building Software That Still Helps When Something FailsWhat HomeCost Canada and VachaVox reveal about fallbacks that preserve user value without hiding failure.Jul 18, 2026RAG Is Not a Support Strategy: Designing AI That Knows When to Cite, Act, and Escalate.How support AI can validate citation provenance, restrict tool authority, and turn weak evidence into a useful human handoff.Jul 01, 2026Requirements Are Not Test Cases: Building Traceability Across 300+ Requirements and 2,000+ Tests.An anonymized field guide to turning hundreds of enterprise requirements into risk-based tests, traceable evidence, and defensible release decisions.Jun 15, 2026Reliability Is a Product Feature: Lessons From Supporting Always-On Retail SystemsWhy green dashboards are not enough—and how user-centred observability, safe change, rehearsals, and runbooks protect retail operations.Jun 01, 2026The Demo Is the Easy Part. Enterprise Software Wins or Loses in the Last Mile.Why workflow discovery, migration rehearsal, adoption, and ownership transfer—not the demo—decide whether enterprise software creates value.May 15, 2026Open Models Now Match 90% of Frontier Quality at up to 84% Lower Cost. So Why Are You Still Just Renting?A practical framework for deciding when to rent frontier AI APIs, use cloud-hosted open models, or own inference through self-hosting.May 01, 2026Vendor Lock-In Was Yesterday's Problem. Operational Lock-In Is What's Killing AI Adoption.Why retail AI adoption fails when legacy operations, fragmented data, batch systems, and institutional habits become harder to move than vendors.