Free as in Freedom: Why Open-Weight AI Needs the Idealism We Had in the Nineties

My thoughts on the recent letter published by Nvidia advocating for open-weight models.

Sometime in the late nineties, during my university years, I fell in love with an idea. I was spending all of my time with Linux back then, installing it, breaking it, reinstalling it, reading everything the Free Software Foundation was about. Not only Linux, but the ideal behind the Free Software Movement excited me to no end.

Richard Stallman and the Free Software Foundation were not only making software free, they were making a moral argument that software freedom was a right, that being able to run, read, modify, and share the code you use was a matter of principle, not convenience. Free as in freedom, not free as in beer. That idealism was intoxicating, and I drank it like kool-aid completely. I wrote about it. I argued for it every chance I got. I did talks about it.

That was one of the founding beliefs that shaped everything since for me. I’ve spent the better part of my career building on top of open source, and I have never stopped being grateful for what it gave me. You could open the source, read how someone smarter than you had solved a problem, understand it, and then bend it to your own purpose. That was the whole joy of it. Hacking, in the real sense of the word, was about the freedom to look underneath, to learn, and to make something new out of what you found. Take that freedom away and you lose the thing that made the work feel like it mattered. In a way, you lose your individualism.

I spend most of my time with AI now. I follow the open-weight models and I follow the closed frontier ones. I now strongly feel we need the same movement we had in the nineties, the same kind of idealism, but for AI and intelligence. If software was worth fighting for back then, intelligence is worth it more. If intelligence becomes ubiquitous, and I believe it will and it should, then the question of who controls it stops being a business question and becomes a question about who gets to learn, build, and think with the most powerful tools of the age.

Closed, proprietary models don’t give you that freedom. They give you a service. You can rent access to intelligence, but you can’t see underneath it, can’t learn from how it was made, can’t modify it to fit the shape of your own problem. And above all, can’t run it the way you want to. I don’t think intelligence is something people should only be allowed to rent. It’s something everyone would want to get their hands on. It’s something everyone should get their hands on.

With software, the four freedoms were complete. You had the source and a cheap computer, and you could actually exercise every one of them: reading it, running it, changing it, sharing it. With models, this isn’t so straightforward. Model weights are not source code you can sit and read like a program. Actually reproducing a model means having the training data, the pipeline, and millions of dollars of compute that almost no individual has. So even total openness leaves a gap the nineties movement never had to face.

But I don’t think that weakens the case. I think it changes it. In the nineties, the fight was simply will they give us the source? With AI, even full openness isn’t enough on its own, because exercising the freedom also depends on access to compute, to data, to infrastructure. The old movement assumed the hardware was already democratized and fought for the software. A movement for open intelligence has to fight for both at once. That’s harder and is also more important.

There is one freedom that survives the analogy completely intact. It’s the freedom to run it the way you want. To run a model locally, privately, on your own machine, on your own terms, without asking anyone’s permission and without sending your every thought to someone else’s API. That maps perfectly onto the old free software ideal.

Yesterday, Jensen Huang used his first-ever post on X to share a letter, “Open Weights and American AI Leadership”, signed by twenty-five companies including NVIDIA, Microsoft, Meta, Hugging Face, Dell, IBM, and Palantir. Satya Nadella backed it as well. It’s a serious coalition making the case that open-weight models strengthen security, accelerate innovation, and are foundational to keeping the ecosystem healthy.

Reading the letter brought back memories of an older life. Of the ideals I stuck to fervently. Of the Free Software Movement, and all it meant.

However, reading it closely, I couldn’t help but notice who wasn’t in the signatory list. The frontier labs — OpenAI, Anthropic, Google — were absent. The companies with the most powerful models and the most to gain from keeping them closed did not sign it.

The letter, in addition to the above, makes the pragmatic case for openness. That is, it is good for security, good for competition, good for American leadership. It is being made loudly by some of the largest companies on earth. But the moral case, the one the Free Software Foundation made, the one that moved me at twenty, is not there. Almost nobody is making it anymore. We seem to have forgotten how.

There’s an uncomfortable geopolitical fact underneath all of it. There is an observable lack of focus on open-weight models in the US, and the country genuinely pushing the open-weight frontier right now is China, while several leading American labs have moved the other way. That should be a wake-up call.

The letter makes pragmatic case. What’s missing is the conviction. We need people willing to believe, fervently and a little unreasonably, in what open weights and open-source AI could mean for society. The same kind of belief that built the free software movement out of almost nothing and went on to quietly power most of the modern world.

I strongly believe the new world is going to be built through intelligence. And if that intelligence is going to be everywhere, then people deserve to see what’s underneath it, to learn from it, and to use what they learn to make something of their own. That was true for software. It must be true for intelligence.

Free as in freedom. It meant something in the nineties. It’s time we remembered it means even more today.