Open a pull request with most AI code review tools and you get a wall of comments. One per finding. One per nit. One per “consider extracting this into a helper.” Push a fix, rerun the review, and you get a second wall stacked on top of the first. Now your teammate is scrolling two walls to find the comment that actually matters.
Surmado Code Review works differently. One PR, one comment. Rerun it and the same comment gets edited in place.
How it works
When Scout reviews a pull request, every finding goes into a single structured comment. Severity bands at the top. The issues themselves below. A short brief at the end so the human reviewer knows where to look first.
When you push a fix and comment /rerun-review, Scout reads two things:
- The new diff.
- The previous review it posted.
Then it edits the same comment in place. Issues you patched get marked resolved. Issues you didn’t get re-flagged with a note that they were called out before. New issues from the new code land at the bottom.
The comment thread on your PR doesn’t grow. The single review just gets more accurate every time you push.
Why this matters
A PR comment thread is supposed to be a conversation, not a billboard. Every extra comment a bot posts is a comment a teammate has to scroll past to find the one that matters. After a few cycles, nobody reads any of them.
This is the same trade-off that came up in our CodeRabbit alternatives writeup and our Greptile breakdown. Catch rate isn’t the only metric. Signal-to-noise is. A reviewer that catches everything but buries it in 40 comments has the same practical value as a reviewer that catches nothing.
The “aware of what it said before” part
Edit-in-place is one thing. The harder part is making the rerun useful, not just shorter.
Scout’s rerun is grounded in the previous review. If it told you on the first pass that a function was missing PII redaction, the rerun looks specifically at whether that function now redacts PII. Not “is there any PII issue somewhere,” but “did the thing I flagged get patched.” If yes, that finding moves to a resolved section. If no, it stays open with a marker that this is the second time it has been called out.
That feedback loop is why “one comment” works. The comment carries its own history. You don’t need a chronological thread of bot comments to remember what was flagged before. The current comment already knows.
What it looks like in practice
Push a PR. Get one comment back from Scout, anchored to the rules in your STANDARDS.MD file. Read it. Fix what matters. Push again. Comment /rerun-review. The same comment updates with what’s resolved, what’s still open, and what’s new. No fresh thread. No “you’ve been notified 17 times.”
If you want the longer story on how the tool was built, our vibe-coding writeup covers the parts we hand-crafted and the parts we let Scout build.
The Bottom Line
PR comment spam isn’t a feature. It’s a habit other tools picked up because posting more comments is easier than knowing which ones matter.
Surmado Code Review skips it. One comment. Edited on rerun. Aware of what it already said.
Try Surmado Code Review for $15 a month. 100 PRs. Your STANDARDS.MD. Every PR, before your teammates open it.