Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Show HN: I canceled my AI code reviewer and wrote a free local one

github.com

Post reply on HN
Print view View original

11–20 of 39 posts

Re: Show HN: I canceled my AI code reviewer and wrote a free local one

#11

Post by StableAlkyne »

gchamonlive wrote: post #9
StableAlkyne wrote: post #8

I've never seen this idiom before: try: AVOUCH_VERSION = importlib.metadata.version("avouch") except importlib.metadata.PackageNotFoundError: AVOUCH_VERSION = "0.3.2" What's the intended purpose? It seems to be protecting against a scenario where the package is not installed but the CLI is somehow running... ...But that seems like a very odd thing to be concerned about. Plus, unless I'm missing something, the program…

Could be for unit testing, maybe it's something that makes sense for the developer, or it's AI slop. In all cases it's a minor nitpick in my opinion.

I'm assuming it's vibe coded, but it's a code smell I've noticed is specific to vibe coded projects.

AIs tend to have a debilitating phobia of any kind of exception. Claude especially will throw try/except and weird coersion everywhere in the code.

This bites you because it hides legitimate errors, even when the program is in such an unrecoverable state that it doesn't make logical sense to continue. It becomes impossible to debug, and requires the test suite to be absolutely perfect because you end up throwing away the only way for the language to warn you that something went wrong.

So while yes, it's relatively minor, like many code smells it is indicative that other things may be wrong.

Re: Show HN: I canceled my AI code reviewer and wrote a free local one

#13

Post by gchamonlive »

StableAlkyne wrote: post #11
gchamonlive wrote: post #9

Earlier quoted context omitted.

Could be for unit testing, maybe it's something that makes sense for the developer, or it's AI slop. In all cases it's a minor nitpick in my opinion.

I'm assuming it's vibe coded, but it's a code smell I've noticed is specific to vibe coded projects. AIs tend to have a debilitating phobia of any kind of exception. Claude especially will throw try/except and weird coersion everywhere in the code. This bites you because it hides legitimate errors, even when the program is in such an unrecoverable state that it doesn't make logical sense to continue. It becomes impos…

Are other things wrong? I'd be wary to shame people's work without doing legwork to getting evidence to support your claim, otherwise you risk passing as someone just seeking attention and upvotes, riding on the tides of HN anti-AI sentiment. Which is valid for vote farming just to be clear, I'd just respect you less if this was the case.

Re: Show HN: I canceled my AI code reviewer and wrote a free local one

#15

Post by mukundzha6 »

ramon156 wrote: post #10

how is tjis better than (OpenCode|Pi|CC|Codex) that had a Reviewer agent? the whole idea of e.g. CodeRabbit is that it allows to review PRs autonomously. making it local defeats the purpose. it doesnt help that these projects have AI generated README's. Not AI-assisted, or AI-curated, but fully generated. it's a bunch of sentences that say nothing. "the diff is whats reviewed". okay, what about per-commit reviews? do…

I just formatted and fixed it using ai. I didnot knew its looking fully ai generated. I will fix it. Till day after tommorow readme is going to be updated. And we are not competing with ai tools. I am aiming it to make it an standard for developer before making commit. Later its gonna have ai intergartion as well. THANKS BUDDY FOR THAT SUCH GOOD FEEDBACK!.

Re: Show HN: I canceled my AI code reviewer and wrote a free local one

#17

Post by mukundzha6 »

DemolitionMan wrote: post #14

I can't quite figure out what distinguishes it from other similar projects—and, above all, what problem it solves.

I wanna make it standard for developers before making pre commit. We are local focused , no data stored of anyone and provide lightning fast analysis for now. Later we are gonna make ai integartion for better output. THANKS BUDDY FOR THAT FEEDBACK.

Re: Show HN: I canceled my AI code reviewer and wrote a free local one

#18

Post by mukundzha6 »

StableAlkyne wrote: post #11
gchamonlive wrote: post #9

Earlier quoted context omitted.

Could be for unit testing, maybe it's something that makes sense for the developer, or it's AI slop. In all cases it's a minor nitpick in my opinion.

I'm assuming it's vibe coded, but it's a code smell I've noticed is specific to vibe coded projects. AIs tend to have a debilitating phobia of any kind of exception. Claude especially will throw try/except and weird coersion everywhere in the code. This bites you because it hides legitimate errors, even when the program is in such an unrecoverable state that it doesn't make logical sense to continue. It becomes impos…

Bro. I didn't had money fo claude code to review my codebase so i build this. And you are saying that i use claude code or anyother tools fot this. NOT GONNA LIE. I FORMATTED THE CODE FROM AI TO BE BREAKPROOF SO IN SOME CASES IT MAY LOOK AI GENERATED BUT ITS MOSTLY MY CODE WITH SOME AI TWEAKS.NOTHING MORE.

Re: Show HN: I canceled my AI code reviewer and wrote a free local one

#19

Post by StableAlkyne »

gchamonlive wrote: post #13
StableAlkyne wrote: post #11

Earlier quoted context omitted.

I'm assuming it's vibe coded, but it's a code smell I've noticed is specific to vibe coded projects. AIs tend to have a debilitating phobia of any kind of exception. Claude especially will throw try/except and weird coersion everywhere in the code. This bites you because it hides legitimate errors, even when the program is in such an unrecoverable state that it doesn't make logical sense to continue. It becomes impos…

Are other things wrong? I'd be wary to shame people's work without doing legwork to getting evidence to support your claim, otherwise you risk passing as someone just seeking attention and upvotes, riding on the tides of HN anti-AI sentiment. Which is valid for vote farming just to be clear, I'd just respect you less if this was the case.

I mean, my original post is just asking why it's in there. Maybe I'm missing something, so I asked. That isn't "shaming" their work.

> otherwise you risk passing as someone just seeking attention and upvotes

This isn't Reddit.

Re: Show HN: I canceled my AI code reviewer and wrote a free local one

#20

Post by doc_ick »

mukundzha6 wrote: post #17
DemolitionMan wrote: post #14

I can't quite figure out what distinguishes it from other similar projects—and, above all, what problem it solves.

I wanna make it standard for developers before making pre commit. We are local focused , no data stored of anyone and provide lightning fast analysis for now. Later we are gonna make ai integartion for better output. THANKS BUDDY FOR THAT FEEDBACK.

However other tools like gitlab sast, python black, and others have differential flags that only scan for changed files. Worst case on a precommit that can be wired to pass only git changed files to be reviewed.

Not for criticism, but is this targeting that “air gapped” git diff scanning area?

Post reply on HN