AI Agents, Developer Experience & GitHub Copilot — Cassidy Williams

Cassidy Williams joins Ido Evergreen to discuss GitHub Copilot, trust in AI agents, and why engineering judgment and product direction still matter. We explore how developer experience is changing, how Cassidy decides what to delegate, and why trying to keep up with every new AI tool can work against you. Cassidy also shares her perspective on using multiple models, MCP and skills, WebMCP, and the pressure AI-generated pull requests put on open-source maintainers. In this episode: • Choosing when to write code yourself and when to use agents • Staying relevant without chasing every new tool • GitHub Copilot’s evolution beyond autocomplete • Building trust in agents through small, practical experiments • Why faster implementation still needs product judgment • Accessibility and the future of an agent-driven web • Contributing responsibly to open source • Cassidy’s predictions for the next five years of software development What do you delegate to AI—and what do you still prefer to do yourself? Share your approach in the comments. Subscribe for more conversations about building software, and share this episode with a developer friend. 00:00 intro 02:37 Introduction to AI and Developer Experience 04:44 The Evolution of Developer Tools 07:23 The Role of AI in Software Development 10:04 Navigating the Future of Development 12:52 GitHub Copilot: Beyond Autocompletion 15:27 The Agentic Framework of GitHub Copilot 17:24 Trusting AI in Development 20:18 The Changing Landscape of Developer Roles 21:46 Open Source in the Age of AI 24:06 Predictions for the Future of Development #SoftwareDevelopment #GitHubCopilot #aiagents #github