We brought together Steve Yegge and Nathan Sobo to explore the future of the IDE. Steve, a veteran engineer from Amazon, Google, and Sourcegraph, has become convinced that traditional IDEs are obsolete, arguing that developers still staring at source code are "doing it wrong" in the AI era. Nathan, CEO and co-founder of Zed and co-creator of Atom, has spent over a decade building code editors.
Their opposing perspectives reveal surprising common ground. Both agree that current IDE architectures are fundamentally broken for the AI era, that collaboration must be central to future development tools, and that we're still in the early stages of understanding effective human-AI workflows.
You can watch the lively debate on YouTube or read below for some selected quotes.
The Death and Rebirth of IDEs
"I think that the IDE monolith architecture is kind of flawed... your AI needs to be able to do everything for you. And that means it can't sit around waiting for somebody to build a plugin for it." — Steve Yegi
"The IDE is dead. Long live the IDE. Or maybe the IDE needs to be reinvented in some very fundamental ways in order to be useful." — Nathan Sobo
The Role of the Engineer is Changing
"My role has changed as an individual contributor. I'm no longer building cars, I'm now running a factory full of workers building cars." — Steve Yegi
"I envisioned myself like coming in in the morning with a cup of coffee and there's a bunch of like supervisors there all with their own teams. And they all brief me on where things are at." — Steve Yegi
Collaboration as the Future
"We really view this as a collaborative environment with multiple humans and multiple agents interacting in this very continuous way." — Nathan Sobo
"The ideal is putting AI in Zed as every bit of a collaborator as another human. Ideally that would be the goal." — Nathan Sobo
The Reality of Agentic Engineering
"I review a lot of the code that I write, but I will say that I don't necessarily understand it, definitely don't understand it as well as if I'd written it myself." — Nathan Sobo
"They're toddlers on ice skates with chainsaws. Right. It's like getting this word across has been so hard." — Steve Yegi
The IDE as a Platform Opportunity
"One of the great things about IDEs... they're designed to wrap things. That's kind of what they do. They wrap other tools... There's a really thin, thin layer over a really rich platform that can be repurposed and extended in a lot of different directions." — Steve Yegi
"The quality of a platform needs to be very high, which takes time. Both of us are building platforms and those things are effectively a lifetime of work. Because a platform, it's almost like a living thing, right? It can grow to do anything you want." — Nathan Sobo
What IDEs Need to Become
"I can tell you a laundry list of things that I do every single day in my outer loop or in my inner loop with the agents that I wish an IDE were helping me with. Right. But VS Code is not helping me, Cursor's not helping me." — Steve Yegi
"Once you start working with these autonomous coding agents and you're sitting at dinner and you're like, man, that one's not working right now, but it could be. And you could be checking in with it on your phone." — Steve Yegi
"If you can get intelligent summaries of the diffs... Our diff technology is decent, but it wasn't really designed for the scale that we're seeing. Right. And so if we can think about scaling up the consumption of diffs, probably by introducing an AI layer in there somewhere." — Steve Yegi
"I wanna come in with my cup of coffee and I wanna have my supervisors tell me how my projects are going, man. I mean, like, fundamentally, I think in terms of multiple projects now." — Steve Yegi