
AI DevOps Podcast James World - AI - Episode 394
38 snips
Mar 23, 2026 James World, a Microsoft‑certified Azure architect and seasoned .NET polyglot, talks about practical generative AI adoption and cloud‑native platform work. He covers which AI capabilities are ready today. He explains spec‑driven coding, combining probabilistic models with deterministic tooling, repo and tooling prep, and strategies for testing, context management, and AI‑first project setups.
AI Snips
Chapters
Transcript
Episode notes
Have Agents Explain UI Test Failures
- When UI tests struggle, ask the agent to document how it navigated the UI and suggest frontend changes to make testing reliable.
- James World uses Playwright and then has the agent produce markdown notes and UI improvement tasks.
Supply Fresh Documentation To Agents
- Feed agents up-to-date documentation and links so they don't rely on stale model knowledge.
- James World recommends tools like Context7 or adding documentation URLs into agent memory for live lookups.
Force Agents To Use Official CLIs
- Instruct agents to use the platform CLI (not write raw files) and embed CLI docs so projects use proper commands and compatible versions.
- James World enforces using dotnet new and nuget CLI to avoid mismatched versions.
