AI DevOps Podcast Pierce Boggan: AI Workflows - Episode 398
19 snips
Apr 20, 2026 
Pierce Boggan, Product leader for Visual Studio Code and GitHub Copilot at Microsoft, steers VS Code toward an AI-native future and created the Primer tool to prep repos. He talks about recent Copilot and VS Code innovations, how local and cloud agents reshape workflows, Primer/AgentRC for AI readiness, faster release cadences, and the evolving IDE with sandboxing, verification, and multi-agent UX.
AI Snips
Chapters
Transcript
Episode notes
Agents Must Verify Real Results
- Agents are moving from line-completion helpers to trusted teammates that execute multi-step tasks autonomously.
- VS Code added a sandbox and embedded browser so agents can run, boot apps, drive navigation, take screenshots, and verify the real result not just compiled code.
Use Adversarial Reviews To Boost Quality
- Use multiple models and adversarial reviews to improve agent output quality by having models grade each other's work.
- GitHub Copilot features like Rubber Duck and Slash Review spin up different models to generate and critique results, improving resolution rates about 5% in hard cases.
Execution Cheapens Decision Making Bottlenecks
- Faster code generation shifts the bottleneck from writing code to product decisions and PR reviews about whether ideas belong in the codebase.
- VS Code team now spends more time in targeted meetings deciding priorities and endorsing PRs because execution is cheap but long-term maintenance cost matters.