my take on the code review tools that are coming out: they’re solving real problems, but not the biggest problem: i need a way to review code effectively before it ever gets to github

Jan 22, 2026 · 1:32 PM UTC

88
14
532
77,296
Sort replies: Relevant Recent Liked
Replying to @dillon_mulroy
i just shipped `amp review` about 30 min ago. it has zero integration with github or PRs whatsoever and never will because i 100% agree with you. it's experimental, and it's also 5:45am and i need to get to sleep, but couldn't not reply when i saw this post.
15
5
142
16,561
👀👀👀
8
3,202
Replying to @dillon_mulroy
I shipped deep code reviews in @RepoPrompt this week. Imo it provides the deepest codereviews because of how the app manages context for you. It supports git & jj, and you can invoke it from any agent either using mcp or the cli. It even uses claude code or codex to run it
2
1
21
1,856
the problem is i don't want more agent code review - i have that problem solved and feel good about the reviews i get from them - i need MY code reviews solved
2
2
844
Replying to @dillon_mulroy
I have a gift for you 🎁. Shipping Sunday I guess.
2
16
1,741
👀👀👀 lets gooooo!
1
3
873
Replying to @dillon_mulroy
Agreed! Iterating through GitHub kills your flow - we shipped the cubic CLI last week: docs.cubic.dev/ai-review/cli… Run cubic reviews locally - syncs all your memory, confirm, and custom agents
2
5
1,342
my issue is not agent/ai reviewing my code, its me reviewing its code
1
3
837
Replying to @dillon_mulroy
What specifically do you want to be better? Relative to, say, cursor review mode?
1
3
2,411
maybe cursor review mode is the play idk ive never installed it - can I oauth to cloudflares ai gateway
2
3
1,558
Replying to @dillon_mulroy
difftastic fixes this
1
3
2,141
it does not
4
4
1,337
Replying to @dillon_mulroy
the tricky part is knowing when to trigger review in local flow. after every function? after a feature? continuous feels noisy but batch feels like you lose context
1
1
881
yup, lots of tension
607
Replying to @dillon_mulroy
Assuming you are using Claude Code you can run their PR review toolkit agents on your local branch changes before opening a PR github.com/anthropics/claude…
1
838
ai reviewing my code is not the problem, that's a solved problem. me reviewing ai code is the problem
1
4
714
Replying to @dillon_mulroy
So pre-commit tools like pre-commit.com/ and then something that gets a LLM opinion on the commit itself?
2
1,961
human opinion
1
2
1,798
Replying to @dillon_mulroy
why do you feel the need to review code before it gets to github? curious because I often only review code once done in github for smaller tasks larger features?
1
413
Replying to @dillon_mulroy
You can run CodeRabbit via CLI and also through VS Code extension, so that is solved.
1
361