OpenAI Slashes Sora Android App Dev to Just 28 Days

"We leaned on Codex to do a huge amount of heavy lifting inside well-understood patterns and well-bounded scopes, while our team focused on architecture, user experience, systemic changes, and final quality," the authors wrote. OpenAI said Codex excelled at reading large codebases, translating logic across platforms, and generating broad test coverage. "Codex is (uniquely) enthusiastic about writing unit tests," the blog noted, adding that engineers frequently pasted CI logs into prompts to diagnose failures.

Codex "isn't yet great at inferring what it hasn't been told," and struggled with "deep architectural judgment" when left unguided. To address this, the team invested heavily in documentation, such as AGENTS.md, to enforce patterns, coding standards, and tooling requirements. One notable technique was to use Codex as a cross-platform translation layer rather than as a shared framework.

"Forget React Native or Flutter; the future of cross-platform is just Codex," the engineers wrote, explaining that Codex translated Swift logic from the iOS app into Kotlin while preserving behaviour. As development accelerated, the bottleneck shifted from writing code to reviewing and coordinating parallel Codex sessions. "Our bottleneck in development shifted from writing code to making decisions, giving feedback, and integrating changes," OpenAI said.

In the company's State of Enterprise AI 2025 report, released a few days ago, the company stated that over the last six weeks, there was a 2x increase in weekly active Codex users. Further, the company observed a ~50% increase in Codex messages over the same period. In October, Sam Altman, the CEO of the company, revealed that "Almost all new code written at OpenAI today is from Codex users." He added that engineers in OpenAI complete 70% more pull requests (PRs) each week using Codex.