Why people try to contribute even if they don't work on their codes? Ai slop not helping at all.
this post was submitted on 18 Feb 2026
881 points (99.3% liked)
Technology
81451 readers
4413 users here now
This is a most excellent place for technology news and articles.
Our Rules
- Follow the lemmy.world rules.
- Only tech related news or articles.
- Be excellent to each other!
- Mod approved content bots can post up to 10 articles per day.
- Threads asking for personal tech support may be deleted.
- Politics threads may be removed.
- No memes allowed as posts, OK to post as comments.
- Only approved bots from the list below, this includes using AI responses and summaries. To ask if your bot can be added please contact a mod.
- Check for duplicates before posting, duplicates may be removed
- Accounts 7 days and younger will have their posts automatically removed.
Approved Bots
founded 2 years ago
MODERATORS
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.
The bots, they don't like when you do that.
If my own mother can't shame me, a glorified sex bot has a snowballs chance in hell of doing it.
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.
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.
One could argue that if the AI response is not distinguishable from a human one at all, then they are equivalent and it doesn't matter.
That said, the current LLM designs have no ability to do that, and so far all efforts to improve them beyond where they are today has made them worse at it. So, I don't think that any tweaking or fiddling with the model will ever be able to do anything toward what you're describing, except possibly using a different, but equally cookie-cutter way of responding that may look different from the old output, but will be much like other new output. It will still be obvious and predictable in a short time after we learn its new obvious tells.
The reason they can't make it better anymore is because they are trying to do so by giving it ever more information to consume in a misguided notion that once it has enough data, it will be overall smarter, but that is not true because it doesn't have any way to distinguish good data from garbage, and they have read and consumed the whole Internet already.
Now, when they try to consume more new data, a ton of it was actually already generated by an LLM, maybe even the same one, so contains no new data, but still takes more CPU to read and process. That redundant data also reinforces what it thinks it knows, counting its own repetition of a piece of information as another corroboration that the data is accurate. It thinks conjecture might be a fact because it saw a lot of "people" say the same thing. It could have been one crackpot talking nonsense that was then repeated as gospel on Reddit by 400 LLM bots. 401 people said the same thing; it MUST be true!
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.
Tests, what you are asking for are automated tests.
Can that be done on github?
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.
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
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.
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.
AI crowd trying hard to find uses for AI
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…
A similar problem is happening in submissions to science journals.
This is big tech trying to kill FOSS.
Which is funny because most of them rely on it
Get that code off of slophub and move it to Codeberg.
Is codeberg magically immune to AI slop pull requests?
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.
No (just like Lemmy isn't immune against AI comments) but Github is actively working towards AI slop
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.
Maybe we need a way to generate checksums during version creation (like file version history) and during test runs of code that would be submitted along side the code as a sort of proof of work that AI couldn't easily recreate. It would make code creation harder for actual developers as well but it may reduce people trying to quickly contribute code the LLMs shit out.
A lightweight plugin that runs in your IDE maybe. So anytime you are writing code and testing it, the plugin is modifying a validation file that shows what you were doing and the results of your tests and debugging. Could then write an algorithm that gives a confidence score to the validation file and either triggers manual review or submits obviously bespoke code.
What exactly would you checksum? All intermediate states that weren't committed, and all test run parameters and outputs? If so, how would you use that to detect an LLM? The current agentic LLM tools also do several edits and run tests for the thing they're writing, then edit more until their tests work.
So the presence of test runs and intermediate states isn't really indicative of a human writing code and I'm skeptical that distinguishing between steps a human would do and steps an LLM would do is any easier or quicker than distinguishing based on the end result.
This could, in theory, also be used by universities to validate submitted papers to weed out AI essays.
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!
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.
load more comments
(2 replies)
load more comments
(3 replies)
I think moving off of GitHub to their own forge would be a good first step to reduce this spam.
load more comments
(9 replies)
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
What if I want to contribue to a FoSS project because I'm using it but I don't want to make new friends?
load more comments
(1 replies)
load more comments
(3 replies)
gzdoom just simply banned ai code, and made a new fork that tries to stay clean. why cant they do the same?
Is all AI code tagged "hey, Claude made this puddle of piss code"?
This is a real "just catch all the criminals" type comment.
gzdoom just simply banned ai code
You got that wrong. Graf Zahl added AI code and all the other contributors left to fork it to create UZDoom.
view more: next ›