alt Hacker NewsI've been helping a bit with OWASP documentation lately and there's been a surge of Indian students eagerly opening nonsensical issues and PRs and all of the communication and code is clearly 100% LLMs. They'll even talk back and forth with each other. It's a huge headache for the maintainers.
I suggested following what Ghostty does where everything starts as discussions - only maintainers create issues, and PRs can only come from issues. It seems like this would deter these sorts of lazy efforts.
Replies
>Indian students
Resume glorification and LinkedIn / GitHub profile attention do that.
I am seeing a lot of people coming up with perceived knowledge that's just LLM echo chambers. Code they contribute comes straight out of LLMs. This is generally fine as long as they know what it does. But when you ask them to make some changes, some are as lost as ever.
Torvalds was right, code maintenance is going to be a headache thanks to LLMs.
I'd argue that the or at least one major reason for the downfall of Stackoverflow (and not just a catalyst) has been a surge of Indian IT people triggering an avalanche of extremely low quality questions and answers. I've been a big fan of SO since about 2010. Not just didn't mind the harsh moderation but actually attribute to it learning how to properly ask a question. But at some point round about 2019/2020 it stopped being fun due to it going from knowledge base to garbage dump.
I've seen this - it's tiring even at low volume. Goes something like:
Someone creates a garbage issue. Someone else asks to be assigned. Someone from the project may say "we don't assign issues" (this step has zero effect over later steps). Someone else submits a PR. Maybe someone else will submit another PR. Maintainers then agonise how they can close issues and PR(s) without being rude or discouraging to genuine efforts.
I contribute regularly to some major open source projects and it’s happening here too. So many issues that aren’t issues. Constant “fixing” of documentation that doesn’t need to be fixed. Bug reports that aren’t bugs, followed by a bad PR “fixing” the “bug.” Or YOLOing an LLM PR to change major behavior that users are relying on. And I click and the authors are always brand new, with only vibe coded or examples projects in their history, and have some truly awful LLM generated GitHub “about me” page complete with emojis and links to their GitHub “projects.”
My suspicion is somehow the perception became that if you’re brand new and land a PR in a major open source repo (even as simple as rewording a phrase in a doc that doesn’t need to be reworded), that would help them get a job (they’re always Open to Work on their GitHub about me page).
It’s so much noise that it’s hard to find the real issues.
Heh, reminds me of that free T-shirt contest thing... Submit crap PRs to random FOSS projects for a chance of winning a shirt, what could go wrong?
Reminds me of this Indian GitHub tutorial on how to open a PR on GitHub. The video got millions of views and has flooded a specific repo with countless README update PRs of people (mostly Indian) trying to append their name to the README.
Article about it here: https://socket.dev/blog/express-js-spam-prs-commoditization-...
Usually the protection against such spam is social shame but the internet is now full of people who have no shame because shame was never part of their culture. It would be more effective to use GeoIP in this case.
Little bit offtopic; is anyone getting more reports of site vulnerabilities I wonder? There are how many AI tools claiming to find those automatically and we get a lot of reports, some even have the tool name on. Thing is, most, well, all so far in the past months, are not true. They seem hallucinations or false positives. We have closed source SaaS products, so these are external scans making these reports.
Stuff like this video is a huge source https://www.youtube.com/watch?v=Ez8F0nW6S
Check the closed PR's on express https://github.com/expressjs/express
Yikes!
I seem to remember there was a large (indian?) educational YouTuber who did a tutorial on how to use Git where they forked a FOSS repo, made a change to the README.md and then made a PR. This caused a huge influx of garbage PRs for that particular repo and other FOSS repos.
Noticed it in corporate context too. About 40% of the performance feedbacks I saw this year were AI written. India and USA crowd. Everything from Europe looked pretty organic but imagine that’ll change too next cycle
I was going to comment something similar.
I was reading some GitHub comments earlier and the AI tone and structure in the comments posted by some users made me feel really uneasy for some reason.
I know a Brazilian who puts lots of emails through ChatGPT because they aren't confident in their English, but this seemed to be AI generating the majority of the content of the message too.
I've noticed this also. But I didn't ascribe it to LLMs, rather figured there is some sort of rogue educator in India who's instructing students to do this on public repos and they just don't know better. But the prof should.
It's likely because of Google Summer of Code. OWASP has participated as an org several times and it's highly likely that they'll participate this time too.
Students often start making PRs around this time to get more familiar with projects before they can put in a proposal when the time comes.
As someone who's been a programmer for a while now, I feel it's pretty easy to identify slop code and when someone is using an LLM to communicate on issues. I'm not against using LLMs for writing code or even for using it to improve your communication, but it cannot be a substitute for critical thinking.
If I was a maintainer of an OSS project, I'd be more likely to _not_ select students who put out slop PRs, proposals, or messages without thought. And also make this clear in the contributing guidelines so contributors know what they're getting into.
> everything starts as discussions - only maintainers create issues, and PRs can only come from issues.
would be so awesome if github supported all that. probably kills their business model though
The other side of the coin is that many real bug reports are dismissed out of hand. That is frustrating if you have spent hours or days triaging an issue and have submitted a well-written bug report. It would be useful if projects advertised what their de facto bug report policy is. If it involves snide remarks and pointless bureaucracy ("you did not check this box") then that should be stated to help others avoid wasting time. Perhaps an LLM could help with that: "The likelihood of an external bug report being acted upon is X%, given analysis of past interactions on bug tracker."
It's Hacktoberfest all year long, baby!
I mean, if people adopt, I guess they can also flood the discussions with LLM nonsense. But for now it seems like the better solution.
What issue added GPU acceleration to Ghostty? It seems silly to add that to a terminal.
Yet again poor communal behavior ruining it for the rest of society, and why we can't have nice things and colonize the stars.
i f8cking hate being born here.
volume of low quality content, dsa/leetcode, etc. is so high, good people/content gets left out. networking, connections, nepotism so much high. getting job based on actual talent very rare.
MNCs which are good outside are so much sh8t here; well capitalism doesn't give a f8ck anyways.
Why you don’t just put an AI guardian to close or to ask them to change the story. Or shadow ban
> Indian students
Is this cultural? I ran a small business some years ago (later failed) and was paying for contract work to various people. At the I perceived the pattern that Indian contractors would never ever ask for clarifications, would never say they didn't know something, would never say they didn't understand something, etc. Instead they just ran with whatever they happened to have in their mind, until I called them out. And if they did something poorly and I didn't call them out they'd never do back as far as I can tell and wonder "did I get it right? Could I have done better?". I don't get this attitude - at my day job I sometimes "run with it" but I periodically check with my manager to make sure "hey this is what you wanted right?". There's little downside to this.
Your comment reminded me of my experience, in the sense that they're both a sort of "fake it till you make it".