Descriptions:
Marlene Mhangami, senior developer advocate at Microsoft and GitHub, opens with striking platform-level data: GitHub is seeing approximately 275 million commits per week in 2026, a rate that would produce 14 billion commits by year-end — 14 times the 1 billion recorded across all of 2025, which was itself GitHub’s biggest year ever. A growing share of those commits are co-authored by AI agents, with Claude, Copilot, and others now leaving traceable signatures.
The central argument, backed by a Stanford University study of 120,000 developers, is that AI amplifies productivity in clean codebases and amplifies entropy in messy ones — meaning AI-assisted development makes disciplined testing practices more important, not less. Mhangami advocates for test-driven development adapted for the agentic era: agents write failing behavioral tests first (the red phase), then rapidly generate code to pass them (green), while developers focus their effort on the refactor phase to ensure quality.
The practical focus is Playwright, Microsoft’s open-source end-to-end browser testing framework, and how it integrates with coding agents. Mhangami covers three integration paths — the Playwright MCP server, the CLI tool, and Playwright Agents — with the latter installing three agent instruction files (planner, generator, and healer) that coordinate to design, write, and automatically repair test suites. The session is directly applicable to teams running Claude Code, Cursor, or GitHub Copilot in automated coding workflows and looking to maintain code quality at agent-generated scale.
📺 Source: AI Engineer · Published May 16, 2026
🏷️ Format: Tutorial Demo
