Akshay SubramaniamMay 12, 2026
Streamlining Code Reviews with Haystack 🔍
42views
Streamlining Code Reviews with Haystack 🔍
Akshay Subramaniam
See how Haystack streamlines code reviews by focusing on the context and story behind pull requests rather than just diffs. Learn how it surfaces key design decisions and verifies author testing to improve your review process.
Streamlining Code Reviews with Haystack 🔍
Akshay Subramaniam
See how Haystack streamlines code reviews by focusing on the context and story behind pull requests rather than just diffs. Learn how it surfaces key design decisions and verifies author testing to improve your review process.
Chapters
00:00
Introduction to Haystack
An introduction to the Haystack platform and its unique approach to code reviews.
00:08
The Story-First Approach
01:09
Understanding Design Decisions
01:39
Verifying Pull Request Changes
02:06
Summary and Conclusion
Transcript
00:01
Hi, everyone.
00:01
Akshay here.
00:02
And today I'm going to give you a demo of how review works in Haystack, the first thing you might notice is that there are no diffs.
00:08
On the screen is the review page.
00:09
There are no diffs.
00:10
And that's because in Haystack, we actually end with the diffs, if you want, but we don't start with them.
00:17
What we do start with is a story of the pull request.
00:20
So in this case, this is a random pull request from a repo.
00:23
And you can actually see here immediately that the pull request is trying to fix a bug with our landing page, where the first visit a user has, the page loads twice.
00:34
And so very quickly, as a reviewer, I can understand, this is happening because on the front end, we're grabbing this cached value from local storage.
00:43
In this case, it's coming up undefined, comparing it to a server flag, and we get the silly mismatch of, you know, undefined is not equal to true.
00:51
So I can very quickly understand what the problem the pull request is trying to address.
00:56
And if this were a feature implementation, I could very quickly understand what feature the pull request is trying to fix.
01:02
So this is a very straightforward, pull request, but it talks about how technically this is fixed.
01:09
And most importantly, Haystack surfaces to you the key design decisions that went into the pull request.
01:15
And the way we grab this information is, is we actually store or, sorry, we have the entire coding agent's conversation history from the author side.
01:24
So you can see here, this is all the stuff, that the author talks about with their agent, whether that's cloud code or Codex, that's stored on your repo, not ours, not any server, your repo.
01:34
And we use that to understand a lot of different things, including design decisions.
01:39
In addition to that, we can understand what did the author actually do when to verify their changes worked.
01:46
So in this case, we can actually see, hey, the author didn't do a very good job.
01:50
They didn't load the front end.
01:52
They didn't try incognito.
01:54
They kind of just created this pull request and said, hey, I'm going to try to merge this.
01:58
And that's actually why this rating is so poor here.
02:01
It's two out of five.
02:02
And, verified is not very good.
02:06
So you can see how Haystack kind of transforms your review and from reading code and kind of trying to parse the story or parse what the person is trying to do into a more streamlined version of what is the pull request.
02:19
Trying to accomplish?
02:20
And how did the author prove that the pull request is actually successful?
02:25
So in this case, I understood.
02:26
What is it trying to accomplish and why is it necessary?
02:29
But I'm not very happy with the.
02:31
What, did the author prove?
02:34
That's all I had for you guys today.
02:35
This is a very quick demo just to get you guys introduced to Haystack.
02:39
I hope you can see why it's useful, and if you guys have any feedback for us, please let me know.
Comments
Comments
No comments yet