According to OpenAI, workplace coordination platform Asana utilized OpenAI Codex to complete a complex code migration in two weeks that was previously projected to take five more years. Facing an outdated and unmaintained testing system called Enzyme that blocked frontend stack modernization, Asana deployed Codex-powered coding agents to tackle the codebase changes. Operating from a five-sentence prompt, up to four coding agents worked in parallel in separate copies of the codebase while engineers reviewed progress twice daily and approved every proposed change. The effort resulted in the full removal of Enzyme within 1.5 weeks of engineering effort spread across two calendar weeks. Beyond framework migration, the project also improved test coverage, fixed poor-quality tests, and cleaned up legacy testing infrastructure. OpenAI reports that model and infrastructure costs totaled about $12,000, while Asana estimates the same scope of manual engineering work would have represented roughly $6 million in fully loaded engineering effort. The successful completion of the project has shifted Asana's perspective on which long-running software initiatives are practical, opening the door to testing agents on other migrations, rewrites, and performance problems that were once assumed to require years of manual work.
Asana completed a years-long code migration in 2 weeks with Codex
Asana used OpenAI Codex to complete a years-long code migration in just two weeks. The tool helped engineers remove the outdated testing system Enzyme, which had blocked modernization of the company's frontend stack. Model and infrastructure costs for the project totaled approximately $12,000.
Why it matters
This demonstrates how frontier AI models can radically compress software engineering timelines and lower costs for large-scale enterprise codebase migrations. It changes which long-running technical projects companies consider practical to tackle.
Topic dossiers OpenAI