What breaks, what it costs, and how to run it in production. Written by the engineers doing the work.
Production, security, AI costs, and agents. What we learn taking AI-built software to production.
AI-generated code is now a due-diligence line item. What reviewers look for, what it does to your valuation, and how to be ready before the data room opens.
Audits find over half of public MCP servers are dead and fewer than one in five are production-ready. The failures repeat: auth, tool design, and a moving spec.
Missing access rules, leaked keys, deleted databases. The documented failures in AI-built software, and the five fixes that prevent most of them.
In June 2026 AI coding tools switched from flat rates to meters and budgets broke. What happened, why it was inevitable, and how to get costs predictable.
Monthly digest of decision frameworks, architecture templates, and upcoming events.