Code review is now the bottleneck. I'm experimenting with diff visualization tools to help me judge…
@mikkmartin3.6K followers
67,540 views
32 Comments
Comments from the original post on X (Twitter)

@ameerashhab97
Very interesting - i've been building this: - but working on getting the builds signed. We're still experimenting!
9 likes

@DavidHruby74
You can't review 1.2k of code. You're just wasting time with this. Instead write good skeleton for your agents and let them implement. The you can review much faster. Also aim for smaller PRs, like 300 LoC, 500 max
5 likes

@zxdubx
I like it, but also I'm already overwhelmed looking at it
5 likes

@HarshitKhemani
experimenting with something called Grove
3 likes

@andrewvijay
Looking at it is the bottleneck. More graphs and more solutions for the same are useless. We need strict rules at the time of authoring and verification harnesses in CI. I'm sorry but no one will use this treemap view after a few times. They will ask the agent to analyse it
2 likes

@vzeroaman
Dude yesssss I’ve been waiting for something like this. Thank you.
1 like

@draginol
This is really well done!
1 like

@YoavCodes
This beautiful but I’m not sure how useful in practice. Would love to read a writeup of your experiments

@Rafirusso
you copied my website url

@jerem_devfront
nice job, that made me think of that

@_watzon
Ok that's kinda fire. I'm a sucker for good visualization tools.

@andyasprou
🤣

@Lukekw_
I think the bottleneck exists precisely because we still cannot trust the AI reviews of the PRs. We should look for ways of improving them rather than the raw diffs imho.

@mdsiaofficial
Which application is this?

@polymorph3us
a potentially interesting way to identify where cohesion and modularity could be stronger.

@EgeMustafaCelik
pretty diffs help ngl, until the agent sneaks a rename and a behavior change into one pr. then you're back to reading it line by line are you ranking them by what actually changes at runtime or mostly by file tree?

@thecsguy
how can I try it?

@parkerbrowndev
I like the experiment. What are the two most important values for making a judgement. How does this serve each of them?

@nonamefa5
what actual problem does it solve appart from looking pretty ?

@ken_at_em
this is neat, what does this tell you? do you just skim this against mentally expected code changes?












