Asana Removes Enzyme in 2 Weeks With Codex, Cutting a $6 Million Plan to $12,000

Q: What hidden engineering harness allowed Asana to compress a five-year modernization project into just two weeks?

The hidden accelerator was not just Codex itself, but what Asana called a strong engineering “harness” around it: a codebase with clear React Testing Library examples, established testing conventions, fast validation loops, and reliable test infrastructure. That harness gave the AI agents a controlled environment to work in. Asana’s engineers used only a five-sentence prompt, then let up to four Codex agents run in parallel on separate copies of the repository. The agents could infer the right migration pattern from existing examples already embedded in the codebase, while type-checking, linting, tests, and CI quickly exposed mistakes. Engineers then reviewed every proposed change before approving it. Asana’s own write-up stressed that the result depended less on elaborate prompting than on the surrounding guardrails. Good examples, accurate conventions, and fast feedback made the Enzyme-to-React Testing Library migration easy for the model to verify and repeat at scale. By contrast, stale internal documentation and flaky tooling were among the biggest sources of friction, because they misdirected the agents or slowed correction loops. In practice, the “hidden harness” was the combination of clean target patterns, dependable automated checks, parallel workstreams, and human review gates. That is what allowed Asana to compress what it had estimated as a five-year, roughly $6 million modernization project into about two weeks at a cost of roughly $12,000.

2 articles · Updated · OpenAI · Aug 18

Two weeks of work let Asana fully remove Enzyme, a legacy testing system the company had expected would take at least five years to unwind.
OpenAI Codex drove the migration from a five-sentence prompt, with up to four coding agents working in parallel while engineers checked progress twice daily and reviewed every proposed change.
About $12,000 in model and infrastructure spending replaced a previous staffing plan Asana had estimated at roughly $6 million.
Enzyme had fallen out of active maintenance and was blocking upgrades to Asana’s frontend stack; with the migration done, the company says it can now pursue other rewrites and performance projects once deemed impractical.

Sources

OpenAI2d ago

Asana Removes Enzyme Testing System in Two Weeks Using OpenAI Codex

OpenAI2d ago

Asana cleared 5 years of engineering work in 2 weeks with Codex | OpenAI

What hidden engineering harness allowed Asana to compress a five-year modernization project into just two weeks?

Could a rapid AI-driven code migration hide subtle bugs that eventually cost more than the initial savings?

Will shifting from writing code to merely reviewing AI outputs fundamentally redefine the software engineering profession?

Sources

2 total

OpenAI2d ago

Asana Removes Enzyme Testing System in Two Weeks Using OpenAI Codex

OpenAI2d ago

Asana cleared 5 years of engineering work in 2 weeks with Codex | OpenAI

Asana Removes Enzyme in 2 Weeks With Codex, Cutting a $6 Million Plan to $12,000

Summary

Sources

Insights

Related Stories

Sources

Related Stories