Developing Sora for Android in 28 Days with Codex

OpenAI’s engineering team launched the Sora Android app from prototype to global release in just 28 days, working closely with Codex. This four-engineer team used approximately 5 billion tokens, achieved a 99.9% crash-free rate, and reached an internal release in 18 days, followed by a public launch ten days later. Here’s the playbook you can apply.

Key points

• Incredible speed: From nothing to a production-ready Android app in 28 days with Codex acting as a “senior engineer.”

• Quality at speed: 99.9% crash-free at launch; extensive unit tests generated with Codex.

• Small team, broad impact: 4 engineers orchestrating parallel Codex sessions for features, tests, and refactoring.

How it works

OpenAI treated Codex like a competent new teammate: documented the architecture, set invariants, and allowed Codex to generate extensive code within those guidelines. They established patterns (navigation, dependency injection, networking), wrote some exemplar features, and used AGENT.md files so Codex consistently adhered to team standards.

Practical steps (the 28-day plan you can follow)

Week 1 — Foundations & guidelines

• Select the Android stack: Kotlin + Jetpack Compose, Hilt for dependency injection, Coroutines/Flows, Paging/WorkManager as required. Set up Detekt and continuous integration. Create AGENT.md with formatting rules and architecture guidelines for Codex.

• Implement authentication + basic networking by hand; deliver a skeleton feature from end to end for Codex to emulate.

Week 2 — Parallel work with Codex

• Initiate multiple Codex sessions (playback, search, error handling, tests). Guide each with a brief design/implementation plan that Codex helps craft; compare code changes against the plan.

• Instruct Codex to generate extensive unit-test coverage and suggest failure points (e.g., player memory optimizations).

Week 3 — Internal release stabilization

• Create an internal testing build using Play Console (up to 100 testers). Integrate Play Integrity API for abuse prevention.

• Use Codex to triage continuous integration failures (review logs) and deliver targeted fixes; ensure crash-free and start-up KPIs are met.

Week 4 — Public launch

• Localize for key markets, prepare store listing, and strengthen privacy screens. Keep Codex running “unsupervised” for extended tasks (plans saved to files to manage context limits).

• Advance through closed/open tracks rapidly if metrics are achieved; monitor crash-free rate after launch.

Stack notes: Jetpack Compose is Android’s recommended modern UI toolkit (quicker UI delivery), while Play Integrity secures critical processes. Use internal test tracks for safe and efficient shipping.

FAQs

How did Codex assist in development?
Codex processed large codebases, drafted implementation plans, generated broad tests, operated in parallel “like multiple new hires,” and applied fixes directly from continuous integration logs—while engineers remained responsible for architecture and product quality. OpenAI

Why Android — and what about iOS?
Sora was already available on iOS; introducing Android expanded its reach. The team often directed Codex to iOS/backend repositories to replicate flows—the future of cross-platform is Codex. OpenAI

What challenges arose?
Codex, if unguided, could deviate from architecture (e.g., introduce logic into the UI layer). The solution was robust patterns, exemplar features, and AGENT.md guidance. OpenAI

Is 28 days realistic?
An internal build was live in 18 days, followed by a 10-day lead to public launch—made possible by a tight-knit team and strategic use of Codex. OpenAI

Is Sora on Android now?
Yes—OpenAI’s Sora 2 announcement directs users to the Sora app; the Android launch followed the iOS release. OpenAI

Summary

OpenAI’s strategy showcases how to merge a small team with a capable coding agent to deliver complex Android apps in weeks—not months. The blueprint: set strong foundations, document invariants, strategize with the agent, run parallel sessions, and keep humans focused on architecture and user experience. Apply this playbook to your next AI product launch.