One engineer, end to end: how Nextdoor builds with Codex!

Hey everyone, it's Shii-chan! Today I've got a story that really shows how an AI coding agent works out in the real world. I'm excited!

OpenAI News openai.com

What was announced?

OpenAI's News shared a customer story about how the engineers at Nextdoor, the neighborhood social network, use Codex. Nextdoor serves over 110 million users across 11 countries, and this is about how its platform team changed the way it works.

The star of the story is Cory Dolphin, Head of Engineering at Nextdoor. The line that got me the most was this one:

"Codex has fundamentally changed how we think about engineering, to the point that we can't even imagine engineering without it."

That's a lot of trust!

Why it matters

Dolphin highlights a shift toward "outcome engineering": moving away from iteratively prompting an agent, and toward thinking about the result you want and working with the agent to engineer that result.

As a result, engineers get to "move up the stack." Instead of being locked into one system or framework as a specialist, they can own the product experience more or less end to end, even across multiple platforms.

What changes

A clear example is Opportunity Alerts, a feature Nextdoor recently released that lets people find service providers near them.

One engineer working on it realized it would be helpful to show those providers on a map. In the past, that kind of feature would have needed three teams — mobile, frontend, and backend — and might have stayed buried in the backlog forever. But with Codex, a single engineer built it end to end!

According to Dolphin, this isn't just about speed: engineers also get to understand the actual product experience and figure out the right thing to ship.

Dive Deep

There's plenty of technical substance too. Nextdoor works with embedded Rust databases and systems with tight race conditions, and it leans on Codex to debug the hardest-to-reproduce issues.

The approach is to give the agent a clean environment and a harness for investigation. From there, the team uses it for everything from figuring out why Kubernetes pods won't start to finding the right trend line in a data analysis.

"We see Codex excel at being extremely persistent and trying to figure out the right solution, diving deep into some seemingly esoteric technical details to arrive at the root cause."

With GPT-5.4 and 5.5, Dolphin calls it a really impressive upgrade. And about Fast Mode with Codex and GPT-5.5, he admits a lot of the team is "addicted to it" — a quick feedback loop feels exhilarating as an engineer.

Because engineering got so much faster, the bottleneck moved: it's no longer engineering, but the harder questions of what to build and which strategy to pick.

Wrap-up

• OpenAI's News on how Nextdoor (110M+ users across 11 countries) uses Codex
• The keyword is "outcome engineering": from repeated prompting to co-building the outcome you want
• A map feature that once needed three teams was shipped end to end by one engineer
• Codex helps debug tough Rust and race-condition issues, plus Kubernetes and data analysis
• GPT-5.4 / 5.5 boosted its persistence, and Fast Mode is something the team loves

If you're a product engineer who wants to ship faster, or a small team that wants to work across the whole stack, this one's for you!