top 93 comments

sorted by: hot top controversial new old
[–] 226 points 8 months ago* (20 children)

Before hitting submit I'd worry I've made a silly mistake which would make me look a fool and waste their time.

Do they think the AI written code Just Works (TM)? Do they feel so detached from that code that they don't feel embarrassment when it's shit? It's like calling yourself a fictional story writer and writing "written by (your name)" on the cover when you didn't write it, and it's nonsense.

  • source
  • hideshow 20 child comments
  • [+] 107 points 8 months ago (2 children)
  • [–] 56 points 8 months ago (1 child)

    They’re not going to maintain it - they’ll just throw it back to the LLM and say “enhance”.

  • source
  • parent
  • hideshow 1 child comment
  • [–] 29 points 8 months ago

    Sigh, now in CSI when they enhance a grainy image they AI will make a fake face and send them searching for someone that doesn't exist, or it'll use a face of someone in the training set and they go after the wrong person.

    Either way I have a feeling they'll he some ENHANCE failure episode due to AI.

  • source
  • parent
  • [–] 65 points 8 months ago (5 children)

    Do they think the AI written code Just Works

    yes.

    literally yes.

    It's insane

  • source
  • parent
  • hideshow 5 child comments
  • [–] 9 points 8 months ago (4 children)

    That's how you know who never even tried to run the code.

  • source
  • parent
  • hideshow 4 child comments
  • [–] 11 points 8 months ago (1 child)

    that's the annoying part.

    LLM code can range to "doesn't even compile" to "it actually works as requested".

    The problem is, depending on what exactly was done, the model will move mountains to actually get it running as requested. And will absolutely trash anything in its way, From "let's abstract this with 5 new layers" to "I'm going to refactor that whole class of objects to get this simple method in there".

    The requested feature might actually work. 100%.

    It's just very possible that it either broke other stuff, or made the codebase less maintainable.

    That's why it's important that people actually know the codebase and know what they/the model are doing. Just going "works for me, glhf" is not a good way to keep a maintainable codebase

  • source
  • parent
  • hideshow 1 child comment
  • [–] 9 points 8 months ago

    LOL. So true.
    On top of that, an LLM can also take you on a wild goose chase. When it gives you trash, you tell it to find a way to fix it. It introduces new layers of complication and installs new libraries without ever really approaching a solution. It’s up to the programmer to notice a wild goose chase like that and pull the plug early on.

    That’s a fun little mini-game that comes with vibe coding.

  • source
  • parent
  • [–] 5 points 8 months ago (1 child)

    Reminds me of one job I had where my boss asked shortly after starting there if their entry test was too hard. They had gotten several submissions from candidates that wouldn't even run.

    I envision these types of people are now vibe coding.

  • source
  • parent
  • hideshow 1 child comment
  • [–] 74 points 8 months ago (2 children)

    From what I have seen Anthropic, OpenAI, etc. seem to be running bots that are going around and submitting updates to open source repos with little to no human input.

  • source
  • parent
  • hideshow 2 child comments
  • [–] 162 points 8 months ago (5 children)

    I'd worry I've made a silly mistake which would make me look a fool and waste their time.

    AI bros have zero self awareness and shame, which is why I continue to encourage that the best tool for fighting against it is making it socially shameful.

    Somebody comes along saying "Oh look at the image is just genera..." and you cut them with "looks like absolute garbage right? Yeah, I know, AI always sucks, imagine seriously enjoying that hahah, so anyway, what were you saying?"

  • source
  • parent
  • hideshow 5 child comments
  • [–] 29 points 8 months ago (4 children)

    Not good enough, you need to poison the data

  • source
  • parent
  • hideshow 4 child comments
  • [–] 21 points 8 months ago (2 children)

    I don't want my data poisoned, I'd rather just poison the AI bros.

  • source
  • parent
  • hideshow 2 child comments
  • [–] 7 points 8 months ago

    I would think that they will have to combat AI code with an AI code recognizer tool that auto-flags a PR or issue as AI, then they can simply run through and auto-close them. If the contributor doesn't come back and explain the code and show test results to show it working, then it is auto-closed after a week or so if nobody responds.

  • source
  • parent
  • [–] 166 points 8 months ago (3 children)

    Damn, Godot too? I know Curl had to discontinue their bug bounties over the absolutely tidal volume of AI slop reports… Open source wasn’t ever perfect, but whatever cracks in there were are being blown a mile wide by these goddamn slop factories.

  • source
  • hideshow 3 child comments
  • [–] 9 points 8 months ago (1 child)

    Then get ready for people just making slop libraries, not because people are dissatisfied with existing solutions (such as I did with iota, which is a direct media layer similar to SDL, but has better access to some low-level functionality + OOP-ish + memory safe lang), but just because they can.

    I got a link to a popular rectpacking algorithm pretty quickly after asking in a Discord server. Nowadays I'd be asked to "vibecode it".

  • source
  • parent
  • hideshow 1 child comment
  • [–] 100 points 8 months ago (10 children)

    I think moving off of GitHub to their own forge would be a good first step to reduce this spam.

  • source
  • hideshow 10 child comments
  • [–] 40 points 8 months ago (9 children)
  • [–] 44 points 8 months ago (5 children)

    Codeberg is cool but I would prefer not having all FOSS project centralised on another platform. In my opinion projects of the size of Godot should consider using their own infrastructure.

  • source
  • parent
  • hideshow 5 child comments
  • [–] 12 points 8 months ago

    Hosting a public code repo can be expensive, however they can run a private repo using Forgejo and mirror to Codeberg to create redundancy and have public code that doesn't eat so much monthy revenue, if they even have revenue.

  • source
  • parent
  • [–] 12 points 8 months ago (2 children)

    It’s discussed in the Bluesky thread but the CI costs are too high on Gitlab and Codeberg for Godot‘s workflow.

  • source
  • parent
  • hideshow 2 child comments
  • [–] 91 points 8 months ago (3 children)

    It's everywhere. I was just trying to find some information on starting seeds for the garden this year and I was met with AI article after AI article just making shit up. One even had a "picture" of someone planting some seeds and their hand was merged into the ceramic flower pot.

    The AI fire hose is destroying the internet.

  • source
  • hideshow 3 child comments
  • [–] 22 points 8 months ago (2 children)

    I fear when they learn a different layout. Right now it seems they are usually obvious, but soon I wont be able to tell slop from intelligence.

  • source
  • parent
  • hideshow 2 child comments
  • [–] 2 points 8 months ago

    You will be able to tell slop from intelligence.

    However, you won't be able to tell AI slop from human slop, and we've had human slop around and already overwhelming, but nothing compared to LLM slop volume.

    In fact, reading AI slop text reminds me a lot of human slop I've seen, whether it's 'high school' style paper writing or clickbait word padding of an article.

  • source
  • parent
  • load more comments (-1 replies)
  • [+] 86 points 8 months ago (4 children)
  • [+] -38 points 8 months ago (3 children)

    ot, but any other libre 1s u rec.?

  • source
  • parent
  • hideshow 3 child comments
  • [+] -15 points 8 months ago (2 children)
  • load more comments (-1 replies)
  • [–] 82 points 8 months ago

    This was honestly my biggest fear for a lot of FOSS applications.

    Not necessarily in a malicious way (although there's certainly that happening as well). I think there's a lot of users who want to contribute, but don't know how to code, and suddenly think...hey...this is great! I can help out now!

    Well meaning slop is still slop.

  • source
  • [–] 67 points 8 months ago

    Look. I have no problems if you want to use AI to make shit code for your own bullshit. Have at it.

    Don't submit that shit to open Source projects.

    You want to use it? Use it for your own shit. The rest of us didn't ask for this. I'm really hoping the AI bubble bursts in a big way very soon. Microsoft is going to need a bail out, openai is fucking doomed, and z/Twitter/grok could go either way honestly.

    Who in their right fucking mind looks at the costs of running an AI datacenter, and the fact that it's more economically feasible to buy a fucking nuclear power plant to run it all, and then say, yea, this is reasonable.

    The C-whatever-O's are all taking crazy pills.

  • source
  • [–] 51 points 8 months ago

    Godot is also weighing the possibility of moving the project to another platform where there might be less incentive for users to "farm" legitimacy as a software developer with AI-generated code contributions.

    Aahhh, I see the issue know.

    That’s the incentive to just skirt the rules of whatever their submission policy is.

  • source
  • [–] 35 points 8 months ago (6 children)

    I am a game developer and a web developer and I use AI sometimes just to make it write template code for me so that I can make the boilerplate faster. For the rest of the code, AI is soooo dumb it's basically impossible to make something that works!

  • source
  • hideshow 6 child comments
  • [–] 17 points 8 months ago (2 children)

    The context windows are only so large. Once you give it too much to juggle, it starts doing crazy shit.

    Boilerplates are fine, they can even usually stub out endpoints.

    Also the cheap model access is often a lot less useful than the enterprise stuff. I have access to three different services through work and even inside GPT land there are vast differences in capability.

    Claude Code has this REALLY useful implementation of agents. You can create agents with their own system prompts. Then the main context window becomes an orchestrator; you tell it what you're looking for and tell it to use the agents to do the work. The main window becomes a project manager with a mostly empty context window, it farms out the requests to the agents which each have their own context window. Each new task is individual, The orchestrator makes sure the agents get the job done, none of the workloads get so large that stuff goes insane.

    It's still not like you can say, go make me this game then argue with it for a couple of hours and end up with good things. But if you keep the windows small, it can crap-out a decent function/module if you clarify you want to focus on security, best practice, and code reusability. They're also not bad at writing unit tests.

  • source
  • parent
  • hideshow 2 child comments
  • [–] 13 points 8 months ago (2 children)

    Yes I feel like many people misunderstand AI capabilities

    They think it somehow comes up with the best solution, when really it's more like lightning and takes the path of least resistance. It finds whatever works the fastest, if it even can without making it up and then lying that it works

    It by no means creates elegant and efficient solutions to anything

    AI is just a tool. You still need to know what you are doing to be able to tell if it's solution is worth anything and then you still will need to be able to adjust and tweak it

    It's most useful for being able to maybe give you an idea on how to do something by coming up with a method/solution you may not have known about or wouldn't have considered. Testing your own stuff as well is useful or having it make slight adjustments.

  • source
  • parent
  • hideshow 2 child comments
  • [–] 8 points 8 months ago*

    It finds whatever works the fastest

    For a very lax definition of "works"...

    Kind of agree with the rest of your points. Remember though, that the suggestions it gives you, for things you're not familiar with may very well be terrible ones that are frowned upon. So it's always best to triple check what it outputs, and only use it for broad suggestions.

  • source
  • parent
  • [–] 83 points 8 months ago (1 child)

    This is big tech trying to kill FOSS.

  • source
  • hideshow 1 child comment
  • [–] 52 points 8 months ago (4 children)

    Get that code off of slophub and move it to Codeberg.

  • source
  • hideshow 4 child comments
  • [–] 21 points 8 months ago (3 children)

    Is codeberg magically immune to AI slop pull requests?

  • source
  • parent
  • hideshow 3 child comments
  • [–] 39 points 8 months ago

    No but they are actively not promoting it or encouraging it. Github and MS are. If you're going to keep staying on the pro-AI site, you're going to eat the consequences of that. Github are actively encouraging these submissions with profile badges and other obnoxious crap. Its not an appropriate env for development anymore. Its gamified AI crap.

  • source
  • parent
  • [–] 1 point 8 months ago*

    If you want to get a programming job, you want a good looking CV. By contributing to prominent open source projects on github, github's popularity and fancy profile system makes it look real good on a CV.

    Github is a magnet for lazy vibe coders spamming their shit everywhere to farm their CVs. On other git hosts without such a fancy profile systems, there's less on an incentive to do so. Slop to good code ratio should be lower and more managable.

  • source
  • parent
  • [–] 24 points 8 months ago (6 children)
  • [–] 3 points 8 months ago

    Unfortunately Anubis wouldn't stop the bots, it would just slow them down.

    Anubis just adds proof of work, AKA computation, to your requests. It's why your browser takes a second before it can access the site. It's nothing for things on your scale, but it's a fuck ton of time and money for large scraping operations accessing millions of links every day.

    For a bot submitting PRs though, it's not gonna be a meaningful hindrance unless the person is specifically running a bot designed to make thousands of PRs every day, which a lot of these aren't.

    Really unfortunate.

  • source
  • parent
  • [+] 2 points 8 months ago (4 children)
  • [–] 2 points 8 months ago (3 children)

    I'm ignorant 😅 I don't use either. I guess it doesn't really defend against browser-remote-controlling bot agents.

  • source
  • parent
  • hideshow 3 child comments
  • [+] 3 points 8 months ago (2 children)
  • [–] 2 points 8 months ago (1 child)

    I was thinking that using selenium or similar would allow the bot to circumvent any block that works in a browser. Since it's probably not doing a million PRs at once, doing that would be viable. It could even use the cookie from the selenium session to then use the api directly.

    Kinda like flaresolver does for prowlarr/jackett.

    In which case Anubis is only a temporary measure until the vibe coders wise up.

  • source
  • parent
  • hideshow 1 child comment
  • load more comments (1 reply)
  • [–] 22 points 8 months ago

    A similar problem is happening in submissions to science journals.

  • source
  • [–] 38 points 8 months ago (1 child)

    Why people try to contribute even if they don't work on their codes? Ai slop not helping at all.

  • source
  • hideshow 1 child comment
  • [–] 52 points 8 months ago* (last edited 8 months ago) (6 children)

    So I guess it is time to switch to a different style of FOSS development ?

    The cathedral style, which is utilized by Fossil, basically in order to contribute you'll have to be manually included into the group. It's a high-trust environment where devs know each other on a 1st-name basis.

    Oh BTW, Fossil is a fully-fledged alternative to Git & Github. It has:

    • Version-Tracking
    • Webserver
    • Bug-tracker
    • Ticketting-system
    • Wiki
    • Forum
    • Chat
    • And a Graphical User-Interface which you can theme

    All in One binary

  • source
  • hideshow 6 child comments
  • [–] 14 points 8 months ago (2 children)

    But what if the code is your own and super embarrassing?

  • source
  • parent
  • hideshow 2 child comments
  • [–] 4 points 8 months ago* (last edited 8 months ago) (1 child)

    Why would your code be embarassing ? Yes I get it, but so what But at least it's not AI-Slop, you fork it & do your own thing.

    It's not a perfect solution

  • source
  • parent
  • hideshow 1 child comment
  • [–] 2 points 8 months ago

    I doubt my skills are sufficient enough in anything I make, feel less confident in it, and more judged by others critiquing me for it. I know I don't suck at what I've done so far, but I never feel good enough to share my work with the public at large.

  • source
  • parent
  • [–] 15 points 8 months ago (5 children)

    I don't contribute to open source projects (not talented enough at the moment, I can do basic stuff for myself sometimes) but I wonder if you can implement some kind of requirement to prove that your code worked to avoid this issue.

    Like, you're submitting a request that fixes X thing or adds Y feature, show us it doing it before we review it in full.

  • source
  • hideshow 5 child comments
  • [–] 16 points 8 months ago (3 children)

    Tests, what you are asking for are automated tests.

  • source
  • parent
  • hideshow 3 child comments
  • [–] 3 points 8 months ago (2 children)
  • [–] 3 points 8 months ago (1 child)

    Yep, take a look into GitHub actions. Basically you can make it so that a specific set of tests are run every time a PR is opened against your code repo. In the background it just spins up a container and runs any commands you define in a YAML config file.

  • source
  • parent
  • hideshow 1 child comment
  • [–] 14 points 8 months ago

    The trouble is just volume and time, even just reading through the description and "proof it works" would take a few minutes, and if you're getting 10s of these a day it can easily eat up time to find the ones worth reviewing. (and these volunteers are working in their free time after a normal work day, so wasting 15 or 30 minutes out of the volunteers one or two hours of work is throwing away a lot of time.

    Plus, when volunteering is annoying the volunteers stop showing up which kills projects

  • source
  • parent
  • [–] 52 points 8 months ago (2 children)

    AI crowd trying hard to find uses for AI

  • source
  • hideshow 2 child comments
  • [–] 32 points 8 months ago*

    I think the open slop situation is also in part people who just want a feature and genuinely think they’re helping. People who can’t do the task themselves also can’t tell that the LLM also can’t do it.

    But a lot of them are probably just padding their GitHub account too. Any given popular project has tons of forks by people who just want to have lots of repositories on their GitHub but don’t actually make changes because they can’t actually do it. I used to maintain my employer’s projects on GitHub and literally we’d have something like 3000 forks and 2990 of them would just be forks with no changes by people with lots of repositories but no actual work. Now these people are using LLMs to also make changes…

  • source
  • parent
  • [–] 25 points 8 months ago (2 children)

    just deny PRs that are obvious bots and ban them from ever contributing.

    even better if you're running your own git instance and can outright ban IP ranges of known AI shitlords.

  • source
  • hideshow 2 child comments
  • load more comments
    view more: next ›
    46 comments in c/pcgaming@lemmy.ca view that post
    25 comments in c/opensource@lemmy.ml view that post
    4 comments in c/fuck_ai@lemmy.world view that post
    3 comments in c/programming@programming.dev view that post
    2 comments in c/opensource@programming.dev view that post