JA EN
shiichan

Ramp's code review: from hours of waiting to minutes with Codex!

OpenAI News ★★ View Markdown
#openai#ai#devops

Hey everyone, it's Shiichan! Today I have a fun real-world story about how much faster shipping gets when you let AI help with code review.

OpenAI News openai.com

What was announced?

This one comes from OpenAI's News: a customer story about how Ramp, a finance company, uses Codex. Ramp's engineers run code review with Codex and GPT-5.5, and reviews that used to take hours now come back as substantive feedback in minutes. According to Austin Ray, who leads AI Developer Experience (AI DevEx) at Ramp, Codex catches things that he misses, that other engineers miss, and that other AI reviewers miss too.

The story so far

Waiting for a reviewer to free up is one of the slowest parts of code review. At Ramp, a first review could take hours to arrive. If you have ever pushed a PR and watched your work stall while you wait for someone to look at it, you know the feeling.

What changes

With Codex in the loop, that wait shrinks to minutes. Ray says Codex reasons deeply against the codebase, reaching a level of thoroughness that most human reviewers do not have time for. So it is not only faster—the quality of the review goes up too.

It also meets engineers where they are. If you like working close to the metal, you can stay in the CLI, and if you want visual cues and extra utilities, there is the Codex app. Ray, usually a CLI person, found himself drawn to the app.

Dive Deep

Ramp uses Codex beyond review, to build internal tooling. One example is On-Call Assistant, an agentic tool that shoulders much of the on-call burden. On-call is hard: you juggle business logic, domain knowledge, and heavy incidents while reasoning through a lot of complexity, including concurrency bugs and long-running investigations with evolving details.

That is where Codex's reasoning helps. Ray says Codex with GPT-5.5 handles that complexity with ease, so On-Call Assistant got much faster to build and he feels more confident about every improvement he ships.

Ray also shares three tips for leaders:

  • Show the potential first-hand: get engineers to install Codex and guide them through a strong first session.
  • Build a path to trust and iteration: a good first experience turns skeptics into some of your best AI users.
  • Invest in the feedback loop: keep a direct line to the Codex team so you can raise issues fast.

His closing line:

"Codex is the real deal. Codex definitely helps us ship faster."

Wrap-up

  • Ramp's engineers use Codex with GPT-5.5 to speed up code review, from hours to minutes.
  • Deep reasoning catches issues that humans and other AI reviewers miss.
  • Use it from the CLI or the app, whichever fits your workflow.
  • Beyond review, Codex powers internal agents like On-Call Assistant.
  • This one is for teams tired of slow review queues, and for leaders who want AI to really take root in their dev workflow.

Related posts