Llama Fund: Crowdfund AI Models
Key topics
The Llama Fund is shaking up the AI landscape by crowdfunding large-scale AI model training, sparking a lively debate about the project's potential and pitfalls. While some commenters are excited about democratizing AI research, others are sounding alarm bells over the project's name, which is trademarked by Meta and closely associated with the company's AI efforts. The project's backers argue that crowd-sourcing could help researchers scale up models and stay competitive, but skeptics are questioning the team's expertise and the project's overall viability. As the discussion unfolds, it becomes clear that the Llama Fund is tackling a pressing issue: the concentration of AI power in the hands of a few big labs.
Snapshot generated from the HN discussion
Discussion Activity
Very active discussionFirst comment
N/A
Peak period
38
0-3h
Avg / period
8.3
Based on 75 loaded comments
Key moments
- 01Story posted
Aug 25, 2025 at 4:40 PM EDT
4 months ago
Step 01 - 02First comment
Aug 25, 2025 at 4:40 PM EDT
0s after posting
Step 02 - 03Peak activity
38 comments in 0-3h
Hottest window of the conversation
Step 03 - 04Latest activity
Aug 27, 2025 at 11:14 AM EDT
4 months ago
Step 04
Generating AI Summary...
Analyzing up to 500 comments to identify key contributors and discussion patterns
Want the full context?
Jump to the original sources
Read the primary article or dive into the live Hacker News thread when you're ready.
I'm sick and tired of the big labs controlling most of the generative AI landscape. We are launching Llama Fund as a means of democratizing large scale AI model training through crowd funding.
Our platform will allow researchers to propose a training pipeline, from data curation to the number of GPUs required. Ideally they will already have a toy model working. From there users can crowdfund the effort based on milestones. Researchers can offer incentives, such as providing commercial licenses to contributors.
We hope this will open up a whole new avenue for large scale model work, powering the open source future of AI.
Would love to hear thoughts from the community!
They can pitch it however it makes sense for their work, giving milestones similar to any other crowd funding platform.
It would be interesting to also have a space where the community could propose research directions.
Neat idea, but I'd change the name
There was a lot of llama hype in the starting tbh, and we all know how naming things is the hardest thing sometimes in software.
Model runs take millions. Was really expecting this to have a credible major sponsor or alternatively propose a new distributed torrent-y training model to sidestep the massive pile of money issue
It's not that challenging to spin up a crowdfunding platform, the laws and registrations are simple. It's capped at $5m per project, but if you go further into Reg A or Co-op models you can extend that much further. You can still train amazing models with $5m.
The whole point is I'm not a big company, I just worked on a startup doing heavy ML training and was often dismayed by the state of the OSS ecosystem. Whole cottage industries can form around open sourced models, and I want to help power that.
I do think for critical mass this will need a 3rd party custodian or platform of sorts though. DIY crowdfunding may not be hard to set up legally, but it's still facing the same rug pull risks as other crowdfundings.
To be clear my concern here is structural - not challenging your personal integrity. I'm sure your heart is in the right place
>I want to help power that.
Same - I love the direction. Especially given risk of the big players running away with this and OSS gang being left with no competitive models. Think it would need a bit more guardrails before I drop in cash though
I agree that a backer would definitely help here, one of the reasons for posting this is to see if we can drum up more interest there.
I agree in the guardrails as well, we are thinking checkpoints with clear deliverables to unlock more cash could help. It’s really interesting that the existing crowdfunding platforms don’t enforce a whole lot of this and still work. It seems off to me but I can’t argue much with the outcome, it could be that the smarter crowd that engages here would care more though.
If you are serious, I recommend that you ask the admins (hn@ycombinator.com) to delete this post, then spend at least half an hour thinking about a better name, and what readers might need to know before you (1) create a more substantive landing page and (2) re-announce. That would at least give you a chance at making a positive first impression.
What exactly is wrong about it? I spent a fair amount of time on it and just saying "its bad" isn't really helpful feedback.
- the black text on fog gray background gradient is... not great looking.
- What's with the seemingly arbitrary floating red sun?
- the name Llama.fund is a bit odd given this is for general LLM models.
Thematically it just feels all over the place.
The website can be changed y'know.
That's a red flag for me if true.
> …just saying "its bad" isn't really helpful feedback.
Not sure if serious, but you can start with these:
• Trademark misuse – Legal exposure, identity confusion, invites takedown
• Gmail contact address – Unprofessional, lacks credibility, signals amateur status
• Anonymous "we" – No trust, no track record, only named person using alt email
• Lovable-generated site – Bad looking, substance-free, designed to farm email addresses
Would me having domain email routing somehow make it more legit? No it wouldn’t, I’ve now gotten validation that people like the idea without doing any of that, which is the exact advice that y combinator gives, yet is never followed in HN comments
- responsivity, the site does not scale well on portrait screens, specifically the email input is shifted to the left and partially off-screen
- layout, the header is not vertically aligned. The elements are off-center
- design, it looks nice at a first glance but lacks polish. Consider starting with an existing landing page template (there are lots of free ones) and refining with AI, if you are unable to create one yourself or hire/ask someone.
- contact/ TOS, you provide your (presumably) legal name and email address. If this is a real fund you should at least provide some sort of address and name of the fiscal host. A personal gmail address also strongly indicates that this is not a functional fund with proper organization of fiscal resources.
- name, already mentioned by other posters but the llama part is not good. My first thought was "Meta has new fund?" however the site's content directly squashed this (no legal contact -> no proper fiscal host). Even disregarding trademark concerns, you don't focus on llama but on open-source models, so please try to name it appropriately.
I would also say this is early stages and I don’t think it’s necessary to have every piece of this solved.
This isn’t a fund really it’s crowd funding, and we are looking to gauge interest as we reach out to investors and early clients.
I’m generally surprised by how many people on HN don’t follow their advice. Ship early, get validation
Man, I understand the implementation might have some rough edges but that's besides the point because the idea is cool, not sure why people are almost picking up on this guy.
Maybe I am wrong, I usually am, but I have been on hackernews for almost an year and HN is usually not like this. Most comments here feel like bully comments, literally being too harsh is not necessary and just reflects our personality back imo.
those are my 2 cents atleast.
He "ships" a website with a gmail address and nothing substantial. I could do the same, likely better, in 20 minutes. How could I even tell this from a scam, there is nothing of substance. And the great idea is just obvious and all the painful details to make it work are completely ignored
Like, What do you expect him to add, create a distributed training system, well that is orders of magnitude inefficient than normal training where people donate their money
If you want him to get some reputation, that's fair but I have always believed in building in the public kind approach. Maybe I am wrong, but yes the website definitely might be made better and honestly I might create some checkpoints from this website like never ever use some other persons trademarks,
make the website pleasant to see
just use some mail service, its not worth it showing the gmail sign. The people mocking this is wild
These are things that are easy to do imo.
As I said in the other project, transparency feels like the key to such problem. And honestly the fact that you could ship it in 20 minutes might be valid but I mean :/ cmon man.
What are your thoughts? I also thought of such idea and wanted to build something like this but gave up, Might build it in a year or two but what would you suggest him to do? Instead of giving him harsh responses, lets be productive since I don't care who implements my idea. I just want a place where people crowdfund models. I don't care if some patrick person builds it or I build or you build it. It should be good though
It's fine to develop "in the open" but this is a pitch for several million dollars, and handing that out without any credentials or track record is just not happening.
At least provide rough estimates for what is needed to get this done. What architecture? What training data? Where does the training happen? Who manages and administrates the cluster? Volunteers who try this the first time or paid experts? What solutions to failure recovery, to storage, to tracking and monitoring? Who has the last word on fundamental decisions? How will the legal component be handled? Do they already have a good law firm, how much would that cost? Will the first training be successful right away or how many iterations will be required? Can you even get access to the required GPUs at a reasonable price point? Train on older architecture? How much effort is required and planned to save cost by making training more efficient, ........
And well it turns out that the author hasn't used AI.
And the next comment I see is yours which "suggest" him to use "AI" for "refining"
This felt so weird on why are you suggesting him AI when the gp accused him of using AI in the first place as an almost deregotary term when he in fact didn't use it, like huh???
Note: the website has been updated since my original comment, notably the points on layout/design have been mostly fixed. The lovable tag has stayed through both iterations.
Man wants to crowdfund AI models to create open source AI models
He should've instead shared the prompt that he entered into lovable instead.
Makes a shit ton of sense now. Okayyyy wtf man, I won't really ever have the guts to push cheap AI slop like lovable directly to HN. I am okay with claude code / gemini / heck chatgpt or deepseek etc. but lovable etc. just makes it feel as if the person writing it doesn't even know how to read code or be comfortable with code itself, let alone writing the code.
They are the peak AI bubble things, slap AI on crypto and boom your chain now has attention when the AI thing might be very shitty and doesn't have much to do with AI anyways but just say/mention AI
It’s not as efficient as raising capital though and collocating it
Might create something similar in the future or maybe contribute it to you. Not competition, since the goal is to bring change I suppose.
Transparency really matters in such things I suppose.
What are your thoughts on supporting cryptocurrencies as donation method too? I feel like that there are cryptocurrencies like monero which are truly private and so that might help a little too making the donations more private imo
(Genuine questions, seeing this made me realize I don't even have a ballpark idea.)
I like the idea of crowdfunding models. Absolutely. Yes. However, I want to see how that translates to progress. I’m ok even if the model fails at a catastrophic level, but I want to see the progress towards that end for every dollar sourced.
The training logs, the assumptions, the cloud spend, the markers, and the fit to the end goal.
Do that, and you’ll have something.
Existing crowdfunding platforms don’t do a whole lot of this and still work which is interesting.
Nice job taking a go at something you feel is important.
Apart from that, I agree with the other comments that the website looks unprofessional and llama is a bad name to use for this.
If I would want to give this a shot, I would first get engineers committed with a plan to start as soon as there's funding, set up a non-profit to handle the operation, and make sure that potential investors get the impression I knew what I was doing by providing a full plan and timeline, including addressing the legal challenges (among those, make it clear that the resulting model will be commercially usable and not sued to death. Are you planning on guaranteeing indemnification or do you want to release the model as-is? Etc)