OpenAI cofounder Greg Brockman is sketching out a future in which software largely disappears from view — and, along with it, the need to learn how any given app works.
According to The Decoder, Brockman envisions an "almost no interface" world where "nobody learns software anymore." Rather than people clicking through menus and buttons, the idea is that an invisible, context-aware agent would understand what you want and act on it directly.
That vision comes paired with an unusually frank admission. The Decoder reports that Brockman acknowledged ChatGPT's plugins — the app extensions OpenAI marketed heavily in 2023 — failed "because the models weren't ready." In other words, the technology at the time couldn't reliably deliver on the promise.
Brockman's framing suggests OpenAI now sees the path forward not in bolting features onto existing apps, but in a single agent that quietly handles tasks in the background. Instead of learning a tool, you would simply state a goal.
The Decoder tempers the pitch with a reality check: OpenAI's own Codex, its coding-focused system, remains "light-years" from that seamless, interface-free ideal. The gap between the ambition and what today's products can actually do is still wide.
Why it matters: if agents really do replace the screens and menus we've spent decades learning, it would upend how ordinary people interact with computers — but Brockman's own account of the plugins stumble, and the distance Codex still has to travel, are a reminder that the vision is a long way from arriving.