Inside OpenAI: How Codex Changed How Everyone Works
How OpenAI Uses Codex to Blur Engineering Roles
This conversation with Tibo (engineering lead) and Ed (design engineer) from OpenAI’s Codex team offers a rare glimpse into how an AI-native organization actually operates. The most striking revelation isn’t about the technology—it’s about how dramatically role boundaries have blurred.
On mandatory AI code review: “Everyone just doesn’t have a choice—the code is reviewed by Codex no matter whether you want it reviewed or not. It’s just been so useful at catching issues.” The key insight: forcing AI review on all PRs eliminated the “should I use this?” decision fatigue. The hit rate was good enough that complaints never materialized.
On non-technical adoption: “I DM someone, I was like, I didn’t know you could code. And he’s like, I couldn’t till a few months ago.” Ed describes designers and go-to-market staff submitting their own PRs—not because they learned to code, but because Codex abstracted away the coding. A copywriter can now change UX strings directly without involving engineering.
On junior engineers thriving: “Ahmed joined as a new grad. Didn’t know Rust, learned Rust super quickly. I’ve never seen someone pick up a new language as fast as that… And then the way he discovers the true potential of agents is faster than most people on the team.” The counterintuitive finding: juniors without ingrained workflows adapt faster than 10-year veterans.
On what adaptability looks like: “Sometimes I’m like, oh I’m just going to go back to Vim… and I’m slowing myself down. And then you look at the way they are using AI today and you get inspired.” Even senior engineers find themselves reverting to old habits—the juniors keep them honest.
On job titles dissolving: “Often it’s like hey Ed, you’re just like an engineer on the team writing PRs and just fixing things. You don’t need to go and talk to anyone. You just do it.” When a designer can ship production code, the distinction between “designer” and “engineer” becomes increasingly artificial.
6 Insights From OpenAI on AI-Native Development Practices
- Mandatory AI review works - OpenAI forces Codex to review every PR, optimizing for signal-to-noise ratio to prevent alert fatigue while catching “four layers deep” issues humans would miss
- Non-technical staff ship code - Go-to-market, design, and other teams submit their own PRs through Codex’s web interface, bypassing traditional engineering handoffs entirely
- Junior > Senior in adaptation - New grads who grew up with AI tools often outperform veterans because they don’t have ingrained workflows to unlearn
- The real bottleneck is ideas - With execution costs plummeting, “good ideas matter more” and “correctly aimed ideas matter more”—the hard part is knowing what to build
- Design-to-production gap closes - Designers now fork repos and build “fully functioning” demos rather than static mockups that engineers productionize later
- Small teams win - The most successful teams at OpenAI are “small, nimble teams that set themselves up to learn and iterate super fast”
What This Means for Engineering Organizations
OpenAI’s internal transformation previews what’s coming for every organization. The key insight isn’t that AI writes code—it’s that AI eliminates the handoffs that used to define organizational structure. When designers ship code, copywriters change strings, and juniors outpace seniors, the question becomes: what’s left for traditional role definitions? The answer, according to this team: curiosity, problem selection, and the willingness to adapt. Everything else is becoming infrastructure.