Yup, I’m posting another this week. Sorry.
This week I’m hoping we can wrangle a solution around AI and our selfhosted community. There are plenty of strong opinions (both pro and con), but one thing is for certain - there needs to be better disclosure in promo posts. Two options (that aren’t mutually exclusive):
- Any posts of an AI focused, AI Developed, etc software gets an [AI] tag. No, a [Not-AI] tag is not needed to accomplish this, thats kind of a “non-golfer” sort of tag.
- Comment requiring an AI disclosure response to every promo post, if its not detailed in the post itself. Specifics (generating docs for commands, translation, whole-boat vibe-coded this app, etc) would be requested.
I will say that having disclosure and/or tagging would mean that comments that just say “slop” or “fuck ai” or whatever would be off topic at that point, that information is already provided, so its just noise (and sometimes pretty uncivil - I’ve been light on that for now due to the need for a rule on this).
The tag [AI] would make it easy to filter out (or search for, if that’s your thing), but there is a wildly different degree of AI use out there, and from the posts with a positive score, its usually due to responsible AI use (translations, a snippet they had to do something obscure with, available to use with AI but doesn’t require it, whatever), which is why I think the disclosure has a place as a benefit to everyone.
Please provide any input or alternative options on this, and I can then put it to a vote like the last one. Comments seem to be the best approach without involving something off-site, but if you have a better idea/option, please share.
Personally. I want an AI tag so I know to look more carefully.
I don’t mind AI speeding up a skilled engineer.
But I do mind a crypto bro, turned AI bro, with little experience, too eagerly advertising their vibe coded app.
Its too exhausting to audit everything I may be interested in and the AI tag would help me to budget and optimize my time.
Sounds like a good solution for me. I think most users dislike completely vibe coded apps, not ai supported apps. So maybe we should be more specific here.
For example: [Vibe-Coded]
This would also support new users finding the right tag.
I want a community where people can use AI to help build a tool and be able to post about it here. But unfortunately, I’m just not seeing that. The AI-generated apps seem to be coupled to a drive-by, AI generated post (and comment replies) all full of em dashes and the standard Claude slop language.
So, yes, mandate an AI tag. Hold posters to it and remove violators, because it seems to always be the same class of “contributors” that are cosplaying as software developers.
Not sure if your rule changes are touching this, but the worst offenders I don’t want to see here are:
- posting and commenting text written entirely by AI
- not open sourcing or giving any visibility into their code
- adopting a paid model
The people doing that remind me of the people who would approach me 20 years ago saying “hey I have an idea for an app I want you to build and I’ll give you 5% of my company. It’s like Facebook for dogs, but I need you to sign an NDA before I say any more”.
i agree with you. i have been working orofessionally as a software developer for over 27 years. i’ll use ai to help research something but i cant atand low effort full ai projects being posted.
i always saw non devs using ai to fully generate something for them personally to fix a very custom need but why do these people post projects thry honestly had no hand in.
Man it is great for creating custom apps to scratch your need. I have custom programs to filter out blocklists for one country and dynamically update firewall rules on unified gateways. Stuff that well never be put in as it would hurt their paid subscription.
I want a community where people can use AI to help build a tool
Sure, that’s github
and be able to post about it here
Fine, but others including myself want that slop as far away from here as possible. Maybe start another community? I suggest calling it c/vibehosted.
Fine, but others including myself want that slop as far away from here as possible
And there are people like me who are fine with moderate AI use and would rather judge the project themselves rather than have them rejected outright.
Maybe there should be a community poll
I don’t have a problem with AI. I have a problem with vibe-coded apps released as a one-shot and then never maintained or supported. That’s slop.
I also have a problem with the trace apps (lifttrace, nutritrace, etc.) because while they’re entirely vibe-coded, they are actively developed, but they’re posted here by a brand promotion account that doesn’t otherwise contribute to the community. If there’s any “x% self-ptomotion” threshold, they fail it, because it’s 100% self-promotion.
I know I also reported another post as slop recently but I don’t remember what it was.
If there’s any “x% self-ptomotion” threshold, they fail it, because it’s 100% self-promotion.
Not with f/loss, just account age and they are above the threshold there.
Yeah. Abandonware isn’t cool generally
Honest question intended to spark discussion.
Does this mean that all “single developer” projects can be considered abandonware (that aren’t open source/forkable)?
Or really “all” non open source software really. Companies “can” die.
IMO, abandonware means software that is a dead-end upon its very release, with no hopes or plans for anyone to every build upon it. Abandonware is generally not extensible, follows no good design philosophy that would let someone else build it up, and embodies essentially nothing.
Even a 100-line throwaway Python script has more utility to someone when it is published on PasteBin or whatever. But something like a binary executable released with no source code, with no support, and with no intent by the developer to ever make anything more of it, that’s abandonware.
Not a fan of a tag, since it’s not transparent enough. Sounds like every minor use of AI would warrant a tag, which seems past the point.
The disclosure comment I feel works well. People that care about if/how AI was used can check it to get a proper impression of the scale of and workflow for AI usage, and those who don’t care can ignore it.
Thanks for grappling with this @[email protected]
i want a community where AI is not tolerated at all. ai is a corporate grift and there’s no room for it in a self built community founded on resistance of the tech status quo
Yes this is needed. Thank you for the proposal here.
I would suggest that this probably needs to be really explicit about any AI involvement, i.e. a minimum if AI is used in any capacity in the coding process, it should require the tag. And ideally an explanation if it was used in other parts of the process.
That last post that came up said they used AI ‘for code review only’. In my mind even that deserves the tag, because these terms are so easy to work around. Someone can ‘code up’ the following:
#include <studio.h> int main(void) { printf(“Program that does X thing”); }
(yes, I know the main arguments are not written correctly. You get the point)
and then have the AI reviewer ‘fix’ their code by doing all the actual work. A strict requirement for this tag, for any AI involvement in the creation of the code seems like the key. The code part is going to be where the security issues crop up, and where it’s really important to know who or what is producing the code you’re about to run on your home server.
I think we’re fairly used to a world where people use templates for their websites, documentation, etc. AI use there bothers me less, but an honest disclaimer saying what the AI did would sure go a long way to reducing the hate comments. I think people will still drive-by downvote, but that can’t (and IMO really shouldn’t be) prevented. But without a rule, people aren’t going to be honest.
The scary part is just how emboldened people feel nowadays to just entirely use an AI for all the coding, documentation, website, and then not even put their name on the project. These to me feel like borderline state actor trojan horses disguised as open-source projects.
Legitimate open source developers can spend years writing code to do something very sinplle but useful, and for them to be drowned out by a bunch of completely AI driven, slop posts really bothers me.
I like the AI tag idea. I’m someone who has what I’d call a noderate approach to AI, not an AI bro but any means but I’m also okay with some things built with AI if they’re done with care. If others don’t want to see it, fine, then that’s what a tag could be useful with. However the fuck AI/slop comments on something that admits to being AI is annoying to me. (We know it’s AI, they literally said it is).
If it becomes too much content, then yes would be okay with bi-forcating the community, buy only after it becomes a problem.
I’m not consistent about it yet, but because of exactly this, I’m trying to differentiate the two when I talk.
Responsible automation? I use ML or machine learning.
The grift consuming the world? A Tech Bro? “AI”
I think one of the saddest things is the conflation between the two, like you can’t even talk about one without invoking the other. Or it opening up that whole ethical debate, when you’re just talking about, like, a 100M transcription model trained by one research in some university on a potato.
Yeah it’s heresy on Lemmy, but I do find it genuinely useful. My only regret is that I have to use Claude/Anthropic more than I’d like, which is why I have a vested interest in selfhosting myself. I’d rather figure out how to run the larger models myself and cut them off completely, but you even begin to mention that here and you’ll get downvoted to hell.
You don’t even need Claude anymore. GLM 5.2 API is good enough for 95% of the same things and vastly cheaper.
MiMo 2.5 Pro and Kimi are also very good. And then there’s Cerebras API if you just want simple things done quick.
The thing with self hosting, while awesome, is that it requires a lot of hardware and considerable time investment for what’s essentially a “base tier model,” or at best one step down for what’s still a very cheap API. I still love it, especially the privacy and control aspect, but you aren’t running Claude at home unless you’ve got a threadripper or server hardware collecting dust.
…Hence I can understand why people don’t pursue it. Especially since a cursory Google search will lead you to trying the Deepseek distillation on Ollama (which is awful).
That’s where I am okay with hardware, but can’t seem to fit the models on my 3090. I have dreams of something like an A100 someday, but not until there’s a ton of used ones that hit the market. What do you use for your hardware?
I have a single 3090!
That’s the dream GPU, these days.
And I have 128GB CPU RAM. So the best model I can run is MiMo 2.5 (a 300B model) at around 10 tokens/sec, using hybrid CPU inference.
…But that’s the worst-case scenario, for speed. It’s an IQ3_KT quant (a high quality “trellis” quantization type, but very slow on CPU), with a gigantic model that barely fits in my RAM+VRAM combined, with no DFlash or any kind of speculative decoding turned on. I could tune it to be much faster, but I mostly just want “max quality, fast enough to read as it streams, barely fits in memory” for this model.
For speed, or prompts with lots of thinking or context (like agenic use), I just run Qwen 3.6 27B now. That would fit in your 3090 no matter how much CPU RAM you have, but you have to be smart about the backend and quantization you pick. If you just use Ollama, it’s gonna tell you it won’t fit, or use some horrible default that spits out garbage.
…This is what I meant to emphasize.
It’s not just the hardware. You kinda have to be part developer, part enthusiast to even follow this stuff, it up optimally, and keep it up-to-date. If you just try to Google “best LLM for 3090,” you will get absolute garbage.
I’ll have to play around with mine then, because I’ve had not great luck with it, or at least very disappointing. The CPU offloading is fairly slow, but maybe I should try tweaking more
Be sure to try the ik_llama.cpp fork. Basically, it specializes in MoE CPU offloading on Nvidia cards, and more efficient quantization types than mainline llama.cpp:
https://github.com/ikawrakow/ik_llama.cpp/
And see this repo for specific 3090 configs: https://github.com/noonghunna/club-3090
Honestly I should just write up my setup in this community too.
I’m still impressed you got any MiMo to work at home, at 10 tok/s.
For those trying to visualise that -
https://mikeveerman.github.io/tokenspeed/?rate=10&mode=agent&think=10
Is it a constant 10 or does it (it must do, right?) drop off as context increases?
I imagine you must have compaction or something to mitigate that.
It’s drops off, but not as much as you’d think.
MiMo uses 5:1 SWA, so its long-context compute doesn’t increase as catastrophically as older models. That, and most of the “slowness” comes from the MoE layers being on CPU (whereas the attention layers that get heavier at high context are all on the 3090).
That’s the beauty of these MoEs: they’re just the right size for the “compute-lite” parts to stay in CPU RAM.
I will measure it tomorrow. It is a constant ~9-10TPS for short queries, but definitely slower near my current max context of 85K.
And do you mean prompt compaction? I don’t automate that; when I use that particular model, I tend to use it in Mikupad, aka “raw” notepad mode, and manipulate the context directly. This is so I can do things like chop out conversations, pick different tokens from the logprobs, or edit its own replies/thinking and continue mid reply.
I like manually handling this because, being a local model, prompts are cached. Streaming starts quickly if most of the prompt stays cached, which is actually a really nice advantage over APIs.
I would still prefer an additional [Non-AI] tag. Even if people are arguing against it - it is not same omitting an [AI] tag and consciously saying “I never used and never will use AI”. And the latter is the thing most users who want the AI-tag are looking for.
I really like having the disclosure comment pinned for a more nuanced explanation of what, if any, AI went into a project or post. I think just a tag can’t capture the levels of AI use.
I’m personally a never-genAI, but, unless we go No AI as a community, I don’t think it makes sense to group all projects that touch AI for documentation with all that use it for testing with all that completely let the AI generate all their code, etc. And I don’t think setting a threshold for which get tagged makes sense either. Basically, a tag is misleading no matter how it’s implemented.
I think this should be a thing. At the same time, I would also want something similar for funding or platforming fascists, but that is unlikely to end up being done. I think a simple tag, the [AI] one would work, is the best current solution. I think extra detail in the post is a good thing to do, for example AI assisted documentation, AI assisted bug finding, AI assisted vibe coding. They are all different and have different effects on the product and community. If someone uses AI to find bugs in their own code I am all for it, that is a great use if it. If they use AI to write their login system I am not keen at all given the likelihood of intense security issues and the low likelihood that they will ever fix it.
I think tags are a good idea. I would change the tag to [AI / LLM], and maybe some subtags like [chatbot], [image processing], etc. AI is here to stay, or a least until the US realize the hole under their entire economy (Or both in worst case scenerio) , so regulation is a good solution to this. (In my humble opinion)
I will say that having disclosure and/or tagging would mean that comments that just say “slop” or “fuck ai” or whatever would be off topic at that point, that information is already provided, so its just noise (and sometimes pretty uncivil - I’ve been light on that for now due to the need for a rule on this).
good idea
it won’t solve the “noise” problem though. I was relatively active on [email protected] and we were constantly nagged by sloppies even though the community is clearly dedicated to generativeArt
it won’t solve the “noise” problem though. I was relatively active on [email protected] and we were constantly nagged by sloppies even though the community is clearly dedicated to generativeArt
Maybe y’all would get less hate if the sub was called “generativeArt” instead of grifterly claiming to be “AI”.
i joined the community, i didn’t create it.
Also:
Anything with an [AI] tag, first thing in the title, will have a drive-by downvote issue.
Not sure how to deal with that, or if its even a concern.
EDIT:
Maybe it should be something else that’s not such a loaded keyword?
[ML] for Machine Learning? [SAI]? [LAI]?
I’ve been messing with ‘AI’ for a decade, and even I hate what the term has come to represent.
Oh I’m sure it will be. I don’t know if its a concern though - TBD I suppose.
TBD indeed. But it will effectively ‘downrank’ posts and their visibility, maybe into the negative vote range. I’ve seen highly negative scores across the board in more machine-learning focused subs, and that’s without a tag that catches the eye so easily.
I think even modifying the acronym could make a difference, though (as I ninja edited).
I do like the idea of a different tag, still easy to filter but less of a target like the ai generated communities out there.
For what it’s worth, I asked my self-hosted LLM (MiMo 2.5, no network access outside my desktop), and it came with [AIT] (AI-Topic).
…I think that’s my favorite so far. [AIP] would work too.
I feel like that “obfuscates” the tag enough to blunt impulse downvotes in /new and feeds, without being deceptive or anything.
Actually both are pretty good - AIT for it as a more discussion oriented, AIP for a project post.
I like it, I think something like that would be a great idea
Where should the line be drawn for what is AI coded vs AS assisted (in some manner) and thus tagged?
XerahS is very openly coded with AI and should be tagged. But what about the recent debate about the use of LLM in cURL?
So thats one of the questions that will need to go to a poll, but if I were to give my opinion it would be that any use would mean a tag, and the disclosure area is where it can be detailed.
Otherwise we’d end up with a bajillion differing tags, so a disclosure section makes more sense.
Not tagging for assisted would also mean setting arbitrary lines, and would be way too subjective imo.
