Ultrareview | Launch week 03, Day 3

Most PRs need fast, accurate feedback. A smaller share, around 15% of our customers' PRs, are high risk: changes where a missed bug would be severe, the kind you'd ask your CTO to review.

Our review agents rank #1 on Martian's Code Review benchmark: they're accurate and built for speed, and a standard review covers up to 200 of a PR's highest-priority files. You get feedback in minutes, and that's the right tradeoff for most changes. But not for all of them.

That's why we came up with ultrareview. An ultrareview is a deeper, multi-pass review that uses cubic's most capable review models, giving our agents time and resources to research, trace, and find more complex bugs. They're the same reviews you already know, on steroids.

Ultrareview is available on paid and trial plans, and each run uses your team's reviewed-lines allowance at 3× the standard rate.

How to use

After opening a PR, mention @cubic-dev-ai ultrareview in a new comment in the PR:

cubic posts a comment when the ultrareview starts and posts its findings when it finishes, usually in about 9 minutes. Anything you write after the command goes to the reviewer as guidance for that run, for example @cubic-dev-ai ultrareview: focus on auth edge cases.

How is this different from codebase scans?

A codebase scan is a deep analysis of your entire repository that runs for hours, often 6 to 12.

Ultrareviews sit in between standard AI code review and codebase scans. They're deeper than a regular review, but less open-ended than a codebase scan (they're bound to the changes in your PR), and still fast enough to keep you shipping.

You can also have cubic start ultrareviews automatically for high-risk PRs or for files that match your rules. The Ultrareview docs cover setup.