All my new code will be closed-source from now on. I’ve contributed millions of lines of carefully written OSS code over the past decade, spent thousands of hours helping other people. If you want to use my libraries (1M+ downloads/month) in the future, you have to pay.
I made good money funneling people through my OSS and being recognized as expert in several fields. This was entirely based on HUMANS knowing and seeing me by USING and INTERACTING with my code. No humans will ever read my docs again when coding agents do it in seconds. Nobody will even know it’s me who built it.
Look at Tailwind: 75 million downloads/month, more popular than ever, revenue down 80%, docs traffic down 40%, 75% of engineering team laid off. Someone submitted a PR to add LLM-optimized docs and Wathan had to decline - optimizing for agents accelerates his business’s death. He’s being asked to build the infrastructure for his own obsolescence.
Two of the most common OSS business models:
- Open Core: Give away the library, sell premium once you reach critical mass (Tailwind UI, Prisma Accelerate, Supabase Cloud…)
- Expertise Moat: Be THE expert in your library - consulting gigs, speaking, higher salary
Tailwind just proved the first one is dying. Agents bypass the documentation funnel. They don’t see your premium tier. Every project relying on docs-to-premium conversion will face the same pressure: Prisma, Drizzle, MikroORM, Strapi, and many more.
The core insight: OSS monetization was always about attention. Human eyeballs on your docs, brand, expertise. That attention has literally moved into attention layers. Your docs trained the models that now make visiting you unnecessary. Human attention paid. Artificial attention doesn’t.
Some OSS will keep going - wealthy devs doing it for fun or education. That’s not a system, that’s charity. Most popular OSS runs on economic incentives. Destroy them, they stop playing.
Why go closed-source? When the monetization funnel is broken, you move payment to the only point that still exists: access. OSS gave away access hoping to monetize attention downstream. Agents broke downstream. Closed-source gates access directly. The final irony: OSS trained the models now killing it. We built our own replacement.
My prediction: a new marketplace emerges, built for agents. Want your agent to use Tailwind? Prisma? Pay per access. Libraries become APIs with meters. The old model: free code -> human attention -> monetization. The new model: pay at the gate or your agent doesn’t get in.
LLMs are why we can’t have nice things.
Capitalism is why we can’t have nice things.
LLMs are just a tool.
Tool used to fuck the poor.
Exaaaactly. None of this would be a problem if everyone’s basic necessities were a guarantee: housing, food, water, essential utilities. Think of all the cool stuff we’d have if all these OSS developers weren’t worried about putting food on the table. I would give zero shits if my open source work was “stolen” to make a profit by some giant company if I didn’t have to worry about making enough money to survive.
Billionaires are why we can’t have nice things. And also why we have LLMs.
It’s correlation, not causation.
I mean, the elephant in the room is the blatant licence violations orchestrated by LLM vendors. If your codebase is GPLed and serves to feed a LLM, it should extend to all the code produced by that LLM.
For decades, the FOSS community has been at each others throats about those licenses, and now that we contemplate the largest IP theft/reappropriation of all time, it’s like, not big a deal. I can’t tell that I’m a prolific OSS contributor, but enough to understand the sentiment: “I put code in the open to help humanity, not to make oligarchs better off with a newfound mandate to pollute”.
Technically the act of incorporating code into a model’s weights does not trigger GPL’s redistribution clause, so they are legally in the right even though morally you shouldn’t scrape copylefted code into a model that can be used to create non-copylefted code.
So these weights don’t count as “derived works” because they are not code, but can only be used to generate code (among many other things) in conjunction with an LLM architecture?
Well, once again, that’s just my hot/IANAL take, but when those weights serve to store information in a way that can easily be extracted losslessly (check-out “model extraction attacks”), we should stop treating them as “just weights”.
check-out “model extraction attacks”
The search results I’m seeing for that term point to people extracting (a clone of) the model, through interacting with the available API of an otherwise closed model. I’m not really seeing anyone interacting with a model to extract its training input data.
Is there a better search term, or do you have a more direct reference to lossless extraction of training data from model weights?
I agree on a moral standpoint, but unfortunately this does not hold up legally. Even for licenses specifically targeted in addressing AI outputs to count as derivative works like RAIL, I couldn’t find any case of it holding up in a US court. The best course of action might just be to add bot-filtering to whatever Git instance you host your copylefted works on until this issue has a legal solution. I’m curious on the FSF’s stance on AI output counting as derived works and if they’d ever consider a GPLv4 or new license to explicitly target AI. Couldn’t find anything online.
Its not lossless.
Its not lossless.
Except for when it is, and even when it’s not, there is a fine line leading to calling that plagiarism.
I’m conflicted on this post. OSS does a lot of good as a whole, but regardless of monetization, I don’t want any of my work training an AI. I can respect that portion of his opinion.
His opinion is actually that AI can use his code no problem, they just have to pay a fee.
The problem is that the big LLM AI companies will just say… ‘Fuck off’, because they don’t like paying for any data, and they also think their models will be advanced enough to write their own libraries soon (if not now, depending how much they believe their own marketing hype).
Pricing is an additional unanswered problem in his new model. As a hypothetical: if 1000 traditional OSS users generate $1000 value in conversion to paid users in his old model - what would an AI license cost? Because one license (eg to Anthropic/Claude) would theoretically be cutting off millions of users, maybe 80%+ of his userbase. Would he ask for millions as a licensing fee?
Whole idea is half-baked IMO, but I am sympathetic to the bullshit situation he finds himself in.
You’re right. Personally, I’d rather support FOSS development. His justification isn’t 100% right but some of the idea resonates with me.
I think this model, however it may work will still be better than what we have currently though. If we can even attempt to charge AI-companies for the training data, that would be a huge step. Because the current model is just they take everything, that they can get their hands on.
And if that makes AI-devellopment ecomically unviable, that’s a really good thing
Posting on linked in… Almost didn’t read it. Complains about one thing while putting it on a walled garden data harvesting Microsoft tool.
Wtf.
I find it incredible, how uncharitable some of these comments here are. As an open source contributor myself, I also really don’t like the fact, that my work just gets stolen and profited of by big companies without my permission.
Even the nicest, most idealist engineer still needs to be able to live from his work. I am not saying he is, but he is completely within his right to protect his work from abuse.
Free software shouldn’t mean, that every company can use our code in any way, they like and open source licenses still have terms, for example copyleft licenses, like GPLv3, still require work, which is based on that code to be licensed with the same terms and appropriately credited. AI companies are clearly not abiding by these terms and aren’t really prosecuted for that.
We should be angry at the companies misusing our work instead of open source devs who have had enough.
This is the best comment of the thread.
So many people are nitpicking his post or criticizing the platform that he shares it on (let’s me honest, linkedIN has a much wider impact than the fediverse if something “goes corporate viral”). People deserve to be compensated for their work.
We shouldn’t be mad at the devs trying to make a living, even those who have different views about what open source is. We should be banding together against the companies who’s entire business model is based on theft and abuse. New anti-AI licenses specifically, techniques to poison AI data baked into every repo, class action lawsuits against companies, etc…
Once Universal Basic Income gets implemented and you don’t need to be paid directly for your work to survive, then we bicker incessantly about the finer points of the real definition of open source.
Bullhlshit. You think they werent using your libraries before for profit.
Free software shouldnt mean that every company can use our code in anyway they like.
That is exactly what free means. I have never seen people write attributions for nuget or js packages used.
most popular oss runs on economic incentives
Citation needed.
Right, Linux kernel development is free, philanthropic work, with zero incentive for profit, funded by IBM, Google… 🙄
Still no?
wheels out Firefox
If Google didn’t foot the bill, Chrome would be your only browser, also, funded by economic incentives. If Firefox exists, there’s no monopoly, which to Google, is why it exists.
That’s not a citation, only considers two projects, and doesn’t even try to make the claim that the majority is corporate funding, though I checked and apparently that is true for the kernel.
God, this post makes me so mad.
I understand that not everyone has the privilege to distribute knowledge for social good. I’m in a privileged position–my day job provides more than enough money for a dignified life, so my own code I release is almost always strong-copylefted and for genuine social good rather than survival.
Seeing so many posts thinking a proper “solution” to web scraping for AI training is closing off knowledge by default worries me. Gatekeeping code/art/knowledge shrinks the commons that made all of this possible. Nobody owes us attention, brand recognition, or monetization. Free Open Source Software exists to protect society’s freedom to study, modify, and share the tools it depends on for social good, not for monetization or attention.
I noticed OP used Micro$oft’s GitHub, notorious for mass AI crawling. You can’t rely on THE worst platform for scraping and then complain about it. Host using Forgejo or similar, and use solutions that don’t restrict user freedoms: bot filtering, rate limits, pay-per-crawl, etc.
I think the root problem is that in capitalism, markets often don’t sustainably fund public goods–but that’s a political problem–not something individual maintainers should solve by privatizing knowledge. Continue to vote for and spread leftist ideas of restructuring society to encourage funding of public goods like Free Open Source Software rather than giving up and abandoning your FOSS values.
Posts code on GitHub (Microsoft) Complains on LinkedIn about AI stealing open source code (Microsoft)
Why would the open source community do this to me???
Lol no I get it, AI is coming to devour us all, and it is just waiting until it can gets enough nourishment from code
If they are allowed to train on OSS code then the same is true of proprietary code, they use the same legal mechanisms. Get your code off GitHub…
He is going to get it off github. He said he’d make it closed source
Just like big corporations. Money is the reason why they go closed source… the fear of using their open source code, while using others open source software.
I have no idea who this guy is, but he sounds more like a shareholder/executive than an open source contributor.
I get where he’s coming from… I do… but it also sounds a lot like letting the dark side of the force win. The world is just better with more talent in open source. If only there was some recourse against letting LLM barons strip mine open source for all it’s worth and only leave behind ruin.
Some open source contributors are basically saints. Not everyone can be, but it still makes things look more bleak when the those fighting for the decent and good of the digital world abandon it and pick up the red sabre.
I guess it’s bound to happen when you gotta pay your bills, but use of AI is lowering your income from your open source work (if any).
I pray Tailwind dies. React too. And JavaScript/TypeScript while we’re at it.
Fucking depressing
I’m curious how the model of just selling your application that’s GPL’d usually works out. I don’t see it done often. The only one that comes to mind is OSMAnd. There’s also other interesting models for funding public goods like threshold pledge systems, assurance contracts, ransom model, wall street performer protocol, etc.
