Asana Uses Codex to Finish Five-Year Migration in Two Weeks
Asana used OpenAI Codex to remove Enzyme, an outdated testing system that was blocking upgrades to its frontend technology. The migration was completed with 1.5 weeks of engineering effort spread across two calendar weeks, despite an earlier plan estimating at least five years of work.
Up to four coding agents worked simultaneously in separate copies of the codebase after receiving a five-sentence prompt. An engineer checked their progress twice daily and reviewed every proposed change, with the team finding that simple instructions produced better results than a more elaborate setup.
Model and infrastructure costs totaled about $12,000, compared with roughly $6 million under the previous staffing estimate. The result has encouraged Asana to reconsider other migrations, rewrites, and performance projects that previously seemed too expensive or time-consuming.
Why it matters
- —The project demonstrates how supervised coding agents can accelerate large, repetitive changes across a complex codebase.
- —The reported $12,000 cost was dramatically lower than Asana's roughly $6 million estimate for its earlier staffing plan.
- —Successful automation could make long-delayed migrations, rewrites, and performance improvements practical for engineering teams.
Key facts
- Asana fully removed the outdated Enzyme testing system in two calendar weeks.
- The work required about 1.5 weeks of engineering effort and approximately $12,000 in model and infrastructure costs.
- The previous staffing plan was expected to take at least five years and cost roughly $6 million.
- Up to four Codex agents worked in parallel, while an engineer monitored progress and reviewed every proposed change.
- Asana found that a simple five-sentence prompt worked better than a more elaborate instruction system.
The full text is in the original source. Here we provide a brief summary and key facts.
