Nobody at OpenAI knows what a “designer” is anymore. Or an “engineer.”
And that’s not as scary as it sounds!
I spent an hour with Ed and Tibo from the Codex team—Ed designs the product, Tibo leads engineering—and what struck me wasn’t the growth numbers or the technical architecture. It was watching Ed struggle to answer a simple question about his job title.
He commits code every day now. Submits pull requests. Fixes bugs without asking permission from anyone. He’s been a designer for years, and when I asked what that means now, he paused and said, genuinely: “There’s an identity crisis here. It’s like—what am I?”
That moment stuck with me because I think it’s where a lot of us are headed.
The junior engineers at OpenAI are scaling and accelerating faster than in the past—not because they’re smarter, but because they have less muscle memory to override—and a tool to help them do it. A new grad who’d never touched Rust picked it up in weeks via Codex and now levels up to engineers with a decade of experience elsewhere. It’s the closest thing to Neo downloading kung fu in the Matrix we’ve actually seen.
The model taught itself to spawn copies of itself to solve harder problems, which nobody asked it to do. Non-technical staff are shipping working products at hackathons, not mockups or prototypes but actual functioning things.
The boundaries that used to define who builds what are dissolving faster than anyone expected. Including by the people building the tools that are dissolving them.
This isn’t a productivity story. It’s a story about what work means now.
Here’s what’s inside:
What’s actually happening — the three tiers of usage emerging inside OpenAI, why mandatory AI review became one of their most loved features, and how they meet people where they work instead of forcing everyone into terminals
The emergent behaviors — a model that bootstrapped its own multi-agent system, memory that’s embarrassingly simple, and why minimal tools beat elaborate frameworks
Where the bottlenecks are moving — code generation is mostly solved, so now everything downstream is overwhelmed
The human side — why traditional interviews don’t test what matters anymore, and the $20 equalizer changing who gets to build
Two paradigms crystallizing — tight cockpit loop versus parallel coworker, and why it matters which one you’re building toward
The Codex Playbook Prompts — six prompts that turn these patterns into something you can actually use:
The Signal System — Make AI feedback automatic without creating noise everyone ignores
The Three Doors — Roll out AI tools so everyone from interns to executives actually adopts them
The Copywriter Pattern — Ship a real change to your product without writing code or asking engineering
The Parallel Coworker — Hand off a task, go do something else, come back to finished work
The Memory Kit — Keep a multi-hour AI task from losing the plot halfway through
The Bottleneck Map — Figure out what breaks next when you speed up one part of the workflow
The gap between “I have an idea” and “I built the thing” is collapsing. Not just for engineers. For anyone willing to work differently.
Let’s get into it.
Listen to this episode with a 7-day free trial
Subscribe to Nate’s Substack to listen to this post and get 7 days of free access to the full post archives.
