"Good to merge? Test suite passes locally?" is perfectly valid English. You need to make sure that the bot is configured to not insist on arbitrary prescriptivist style guides that nobody cares about.
And NGI like a human would be given this task and consider the spirit of the law, the goal we want to acheive, and enforce it to reach that. The next token AI doesnt model that. It just predicts the next token, and understanding the spirit of the law does not seem to be in the emergent capabilities of that
I think LLMs are actually great for catching things like this, and you generally don't need some higher-level understanding about the goals involved to notice the ambiguity. My point wasn't that bots shouldn't be used like this, just that they need to be given the right instructions.
this is the part im talking about. i also think LLMs are very capable at detecting different types and levels or grammar, but they cant decide which ones should be filtered out to meet a certain goal. they need detailed instructions for that, and thats somewhat inefficient and causes issues like this right here.
we have done this song and dance many times with AI. its the bitter lesson: you need a system that learns these things, you cant just give it a rule based engine to patch the things it cant learn. that works in the short term, but leads to a dead end. we need something that has the "common sense" to see when grammar is fine versus hindering communication, and this just isnt there yet. so it needs to be given detailed instructions to do so. which may or may not be sustainable
It sounds like you mean that curating messages purely to conform to a particular style guide requires context-dependent information and can never be accomplished reliably with some unmodified generic prompt across many different projects.
I'm saying that while this is true, if you ignore grammar guidelines and just look for cases of ambiguous and confusing wording specifically, then this can actually be accomplished reliably with a generic prompt, if you get that prompt right. Not 100% accurately of course, but good enough that it would be beneficial overall in the long run.
But yeah the bot needs to loosen the boundary a bit there
https://www.reddit.com/r/memes/comments/7lkgdk/video_game_hu...
Our Slack channel is also completely slopped to the point it's mostly bot conversations constantly spamming about review reminders, pull request statuses, and other useless info you can just look up yourself if you need it.
The signal to noise ratio is to the point where I just ignore everything.
There's no utility in reading them, which means asking people more questions directly in slack instead, which genuinely makes the entire process slower for everyone. I know for a fact that no developer in my team would struggle to write 4-5 bullet points for any given PR in less than 2 minutes. My boss just genuinely thinks the slop is better (probably because he's not usually the one reading them)
Our Confluence is also rapidly turning into nothing but totally uninformative Claude-generated noise.
It's annoying to be repeatedly called a Luddite for questioning this stuff when the end results can be so obviously wasteful and unhelpful.
I've been employing the strategy of pushing on the customer and product delivery expectations. Ask some variants of the following in your next team meeting:
"Where are we at with Customer X?"
"How does Bob (the customer's CEO) feel about the project right now?"
"How will this sell us more copies or make the customer pay more money?"
These kinds of questions can rapidly reorient a team if a member with the power to terminate other employees is present. If the most senior management seems apathetic to these questions and you are interested in building an actual career, it might be time to move on. At some point, the person(s) paying for everything need to ask themselves if they intend to conduct a serious business or simply run a daycare for neurotic tech enthusiasts.AI code review really does help but only when used correctly
I'm building the tool to fix that
I kept reading and reading and the "violation of our guidelines" phrase wasn't appearing, so I got bored
In all honesty, it works better for me in a vanilla browser on mobile. GitHub app is less predictable IMHO.
A goddamn staff engineer candidate with 20 years experience submitted an auto generated, broken app, and then tried to pass it off as though it didn't work because our API was down. No, the AI hard coded the wrong URL and he didn't notice (the correct URL was specified in the email containing the test instructions)
I want off Mr Altman's wild ride lol
I basically just stop reading on sight
I don't know this particular project but seeing threads like this kill any motivation to contribute.
"We are a fast moving start up (even at 3-4 years old), we believe in moving fast and breaking things ... That's why we don't do code reviews or unit tests.. we just edit live running code and restart the server"
Vs
"This one line change needs 4 -5 commits to add feature flags,unit tests, integration tests - all to be reviewed by different teams and wait 1-2 months to be deployed properly to production"
I feel this is a needlessly obtuse statement. I'll explain you why, as I've worked professionally with frontend development. From your comment it seems you don't have that type of context.
The text that is expected to feature in a UI element is a critical factor in cross cutting concerns such as product management and UX design, and it involves things like internationalization and accessibility support. This means that if you change a line of text, it needs to be translated to all supported languages, and the translation needs to meet usability and GUI requirements. This check needs to be done in each and every single locale supported.
I can give you a very concrete example. Once I was tasked with changing a single line of text in a button featured in a dialog. It turns out the french translation ended up being too long that forced line breaks. The UI framework didn't handled those line breaks well and reflowed the whole UI, causing a huge mess. This required new translation requests, but it turned out that the new translations were too vague and ambiguous. Product Managers got involved because the french translation resulted in poor user experience. Ultimately the whole dialog was redesigned.
But to you it's just a text on a single button, isn't it?
I'm not saying that developer should also be UI specialist, that's nonsense. But a developer should know who to talk to regarding advice, and the contact should be easy. The model where a dev doesn't even know that French translation exists is wrong. The correct model is having a dev think "wait, this might affect translation, better send a Slack message to some UI guy and let him know".
Actually, what I said is a pipe dream. The reality is, most devs are average devs, and companies need to optimize processes for that. This results in very official communication that takes years to get anything done, because you can't trust average developer that they'll contact the UI guy. This leads to a lot of frustration from above-average developers, who simply need a different environment to shine (more freedom but also more responsibility)
Shouldn't automation be somewhat useful? All these bot comments — do they really bring more value than they create distractions?
> Antiwork emerged from Gumroad's mission to automate repetitive tasks. In 2025, we're taking a bold step by open-sourcing our entire suite of tools that helped run and scale Gumroad. We believe in making powerful automation accessible to everyone.
So yeah, it does look like they bring value for genuinely more reliable code.
Some of the commit descriptions: "fix", "fixes", "clean up".
How is that self-service?
+ const safeToken = typeof token === "string" && token.length > 0 && token.length < 256 ? token : "";
Of arguable quality I would say. The length size limit is arbitrary and the > 0 is ridiculous.> …
> Your expertise about the system's constraints helps provide important context that static analysis tools can't capture.
So much fawning bullshit bloating the message and the token count. I think this might be the thing with LLMs I dislike most.
Suggestion for prompt writers: “Don’t waste tokens. Keep messages succinct and direct.”
Fuck off.
Dear humans who advocated for installing the bot, let me use anodyne, US corporate bullshit language so you'll understand:
Your bot does not add value. Get rid of it, before it drives out all voluntary contributors.
Mathematicians use calculators, and so too do elementary school students, and grocery store clerks, and civil engineers. What each person needs from a calculator can be similar, but would you give a graphing calculator to the store clerk and expect them to be more "productive?"
Admittedly, my metaphor is leaky—and I also can't comment on the participants of the PR—but after reading the comments and the code itself, I'm getting a lot of "here’s a new calculator with a bunch of graphing functions, trigonometric menus, and poem generators—now go do the basic arithmetic you were already doing, but you work for the calculator now" vibes.
Said another way, it took me a lot more time and effort to understand what the bots were saying and if I agreed, than it did for me to formulate my own thoughts and questions.
Like the saying goes, "the best calculator is the one you have with you," and I'd much rather just use my own.
It's just slop and useless context
Code review should only help with the repetitive and tedious parts of review
I'm building a tool to fix that
slacktivism123•6mo ago
When
https://github.com/antiwork/flexile/pull/427#issuecomment-30...
results in
https://github.com/antiwork/flexile/pull/427#issuecomment-30...
More incredible examples where a LLM flags contributors' pull requests because their comments contain minor grammar errors:
https://github.com/antiwork/flexile/pulls?q="our+contributin...
samrus•6mo ago
WesolyKubeczek•6mo ago