For years, a lot of thought and investment went into Integrated Development Environments (IDEs). Agentic development seemed to be their triumph, with tools like VSCode + Copilot, Windsurf, and Cursor bringing development experiences to the next level. But the same agentic development led to the decline of IDEs. The issue is that agents, and not human developers, are writing code. And agents don’t need IDEs to write code.
The role of the IDE is now that of a glorified code reader. It makes it somewhat easier for a human to read and understand what an agent did. With IDEs’ recent security shortcomings (e.g., here), I’m not sure that’s worth it.
Personally, I believe that at this point, investments in IDEs are largely a waste. The same goes for attempts to improve the security posture of agentic IDEs. As mentioned above, the solution here is to stop using IDEs for coding tasks – agents don’t need them, after all. What’s left when you keep one running is not a development environment, but just a large attack surface.