“Cubic saves us maybe 50% of the hours of technical debt that a PR will incur by merging it.”

— Nick Sweeting, founding engineer at Browser Use

Browser Use is the leading agentic browser automation company today - and are used by teams at Tesla, Netflix, IBM, Adobe, and NVIDIA.

Their AI browser agents allow developers to create AIs that book flights, run tests, and perform research online. These workloads touch low-level browser internals and demand extremely careful review.

With a large, fast-moving open-source repo and a high volume of AI-authored contributions, the team needed a way to deliver fast, accurate feedback.

The challenge

The team faced a flood of large contributor PRs, often arriving without descriptions, sometimes “like 5000 lines of changes and no PR description”.

This review load concentrated on a few maintainers, creating delays. Their initial attempts to use AI code reviewers only made the problem worse.

Choosing cubic over the competition

Before adopting cubic, the team trialled a suite of other AI review bots, including Coderabbit and Github Copilot to manage their PR backlog. The experiment was a failure.

“I very quickly messaged the founders and I was like, my God, turn all of this off.,” Nick recalls. “None of these bots are actually helping.” They flooded PRs with useless comments on code style, extra spaces, and variable names, while missing substantive bugs.

The team turned them all off. Then, a week later, a new bot appeared: cubic.

"I was about to message the founders and be like, my god, yet another bot... Except it found a bug in my code. And it was succinct. It was only like two sentences where it’s like, ‘hey, you missed this edge case.’ It was correct. And it was super easy to fix. It was actionable."

Nick Sweeting

This was the turning point. cubic delivered immediate, high-signal feedback. “It just kept happening,” Nick says. “It just kept finding real legit problems and almost never commenting on useless code style.”

cubic in action

cubic now automatically reviewers every PR, uncovering concrete bugs, missing edge cases, security exposures, and inconsistent API ergonomics. Contributors are asked to clear cubic’s comments before a human review is initiated.

This allows maintainers to focus on architecture and long-term design rather than triaging drafts. Senior engineers treat cubic as a zero-ego first reviewer, while junior contributors iterate directly on its concise, actionable comments.

"Over time, I just started relying on it more and more to help with the contributor PRs especially. Like I wouldn't even look at a PR until it had the cubic review. And then my first comment would be, please fix all the cubic comments."

Nick Sweeting

Results

• PR merge time slashed by 85%; average time to merge is now about three hours.

• cubic routinely prevents subtle bugs that tests and linters don’t reliably catch.

• Maintainers reclaim hours of attention for high-level API design and roadmap planning.

• Contributors iterate faster and learn the codebase by resolving precise, actionable feedback without waiting for a human.

Ready to see faster merges with higher confidence? Install cubic now