From the source
Runway has handed successive stages of its software delivery line to AI agents, starting with writing and CI and moving through review and approval, after agent-written code tripled the share of human review time in its core API repository.
- Cloud agents drove 38% of the pull requests Runway merged in October, up from 0.1% in June, and the median wait for a first approval fell from over two hours in September to about five minutes in mid-November.
- An internal service runs coding agents in the cloud and stays with each pull request until it is green; in a sample of 100 of its pull requests merged in the first half of November, 78 had no commits from a person.
- A separate AI reviewer shares the writing agents' context, and 93% of its comment threads were resolved or led to a code change in the first half of November, up from 84% in September, while a late-September rebuild cut the cost of each review to about a third.
- Risk-based approval rates each pull request low, medium or high, rounding up when unsure, and decides who can approve it; high-risk changes go to a person, and code owners still sign off on critical areas such as database migrations, secrets and authentication.
- Runway says a person objected to only 15 of the 977 pull requests the AI approved across September and October, about 1.5%, and that of all pull requests merged on AI approval alone from October to November, only one in a thousand was reverted.






