May 2026
May: AI on both sides of the diff, browser tests as state guardrails, MCP as production integrations
Three threads from the month. The honest tension of using one AI vendor for both codegen and review, browser-test-as-code (Playwright at PR time plus Checkly and Datadog as synthetic monitoring) becoming the practical guardrail against AI-introduced regressions in stateful systems, and MCP servers maturing into the agent runtime's SDK layer without the operational discipline that usually carries.