How VS Code Went from Monthly to Weekly Releases with AI — Harald Kirschner

After 10 years of monthly releases, VS Code now ships weekly. It wasn't more AI that made it work. It was rebuilding the whole system around it. Harald Kirschner, principal product manager on VS Code at Microsoft, explains how a small team shipping to more than 50 million users moved from monthly to weekly releases with AI agents, and what broke along the way. He covers three stages: shipping faster (agent-ready codebases, skills, TypeScript Go's 10x faster builds, letting agents drive the app), holding quality (mandatory AI code review, AI issue triage, error stacks that turn into auto-fix PRs, staged rollouts) and learning faster (VSC-Bench evals, daily prototyping and smaller squads). In this talk: • Code survival: how the share of agent-written code that gets committed rose from 55% to 86% • Why slow CI compounds when 20 agents hit it at once • Giving agents a self-correcting loop by letting them use your app • Turning 51 billion daily telemetry events into auto-filed issues and fix PRs • Why a 5-character 'hello world' eval cost one model 70x more tokens SPEAKER Harald Kirschner, Principal Product Manager, Microsoft (VS Code) LinkedIn: https://www.linkedin.com/in/digitarald X: https://x.com/digitarald GitHub: https://github.com/digitarald LINKS What 50,000 Runs of a 5-Line Eval Taught Us: https://code.visualstudio.com/blogs/2026/06/19/what-50000-runs-taught-us Improving token efficiency in GitHub Copilot: https://code.visualstudio.com/blogs/2026/06/17/improving-token-efficiency-in-github-copilot CHAPTERS 0:00 Intro 0:12 From monthly to weekly releases 1:12 Code survival: 55% → 86% 1:57 When success creates new problems 2:42 Ten years of monthly releases 3:07 Evolve the system, not just the AI usage 3:42 Ship faster, ship quality, learn faster 4:52 Make your codebase agent-ready 5:46 Encode expertise as skills 6:26 The test: can a PM vibe code here? 7:01 TypeScript Go: 10x faster builds 7:56 From scrappy PR to polished feature 8:31 Let agents use your app 9:06 Automated component screenshots 10:06 A self-correcting loop with Playwright 11:20 Mandatory AI code review 12:05 Holding quality: AI issue triage 13:25 From error stacks to auto-fix PRs 15:24 Staged rollouts 15:59 Learning faster with VSC-Bench 17:04 70x the tokens for a 5-character file 17:39 Prototyping in daily sprints 18:19 Smaller squads, durable ownership 18:39 Find your next bottleneck Recorded at the AI Engineer World's Fair 2026 in San Francisco. Subscribe for more talks from the engineers building with AI. AI Engineer: https://ai.engineer YouTube: https://www.youtube.com/@aiDotEngineer X: https://x.com/aiDotEngineer LinkedIn: https://www.linkedin.com/company/aidotengineer/ #VSCode #AIAgents #AIEngineer