Summary
Sam Witteveen asks what role the IDE should play now that coding agents like Claude Code and Codex handle much of the writing. His answer is that IDEs still matter, but their purpose shifts from typing code to reading, navigating and verifying it. The video is sponsored by JetBrains, though it is not a push for a specific product.
He argues that agents often touch ten or fifteen files in a single task, which is hard to inspect from a terminal alone. Diffs, codebase structure and the ability to understand what multiple agents have changed become essential. Drawing on his own multi-backend model testing suite, built with several different agents, he stresses the need to understand how a project is laid out and modularized.
The video also covers the idea that IDEs are becoming places where agents are run and managed, as GitHub Copilot began to do. A key point is that the bottleneck has moved from implementation to code review and planning, with JetBrains reporting that some teams saw pull request times rise after adopting agents. The video also notes the accountability that engineers still carry when approving AI-written changes.
📺 Source: Sam Witteveen · Published November 04, 2026
🏷️ Format: Opinion Editorial
