How engineers at Nextdoor use Codex to build without limits

Nextdoor, a platform serving over 110 million users across 11 countries, relies on Codex with GPT-5.5 to transform its software engineering processes. According to Cory Dolphin, Head of Engineering at Nextdoor, Codex represents a strategic shift from iteratively prompting an agent to outcome engineering, where developers focus on desired results rather than specific frameworks or platforms. This evolution allows individual engineers to move up the stack and own product experiences end-to-end, dismantling traditional silos between mobile, frontend, and backend teams. For example, Nextdoor recently used Codex to build Opportunity Alerts, enabling a single engineer to add a map view feature that historically would have required collaboration across three separate teams and likely stalled in the backlog. Beyond feature development, the engineering team utilizes Codex to investigate hard-to-reproduce issues involving embedded Rust databases, tight race conditions, Kubernetes pod startup failures, and data analysis trend lines. Dolphin notes that newer models like GPT-5.4 and 5.5 excel at persistence and diving into esoteric technical details to locate root causes. Features like Fast Mode have created rapid feedback loops that the team finds exhilarating. Ultimately, the dramatic increase in engineering velocity has shifted pressures within Nextdoor, meaning the primary bottleneck is no longer how to build software, but rather identifying the right strategy and features to build next.