Comment by riazrizvi
4 months ago
The purpose of Stallman’s open source movement was to redistribute power back into the hands of creators who were getting walled out of anything but proprietary work for an employer. If they were fired, they had nothing to show for years of work except a reference, since their deep expertise was essentially meaningless. (An experience I’m sure almost everyone here is familiar with, since we’ve all spent some years on proprietary systems).
Now, with LLMs, exposing your source code essentially hands over a large chunk of your hard won expertise for free to whoever wants to use it. That old model of 100% open source is broken, to my mind.
The new approach I think should be open source stubs with demos of what is capable with your additional proprietary piece.
Stallman’s open source movement
do you want to give RMS a heart attack?
RMS founded the Free Software movement to protect the users of software.
to redistribute power back into the hands of creators who were getting walled out of anything but proprietary work for an employer. If they were fired, they had nothing to show for years of work except a reference, since their deep expertise was essentially meaningless
ignoring the fact the big philosophical different between Free Software and Open Source, neither had the above as a goal. for the first decade or so of the movement, all Free Software and Open Source development was done by people in their free time. practically none of it was done at work. the exceptions are MIT and BSD projects which both predate the Free Software and Open Source movements.
on other words, developers always had the ability to do stuff in their free time regardless of the license. those that live in countries that allow employers to own everything had to fight their employers to be allowed to do so, and they still have to do that. the cases where employees are getting paid to work on Free Software or Open Source are rare, although they are less rare today than in the past because more companies release their sources. but again, this was not the goal at the founding. at least not that this should help the developers. the goal was always to support and protect the users, to allow them to share and modify the software they use.
The GPL he wrote is the basis of the reciprocity agreement that drove the open source movement, it is the legal mechanism that prevents commercial actors from taking over shared works, and locking other creators out of continued participation in their collective creations.
Stallman explicitly warned about working on proprietary software for an employer:
> “If I sign a nondisclosure agreement to work on a proprietary program, I am agreeing not to help you. I am agreeing to withhold information from you, and to refuse to give you a copy so you can learn from it.” This isn’t just about ethics toward the public — it’s about how such arrangements strip a developer of the ability to show, reuse, or build on their own work.
GNU Manifesto (1985).
The GNU GPL is in no way reciprocal, under it, code flows downstream to users, not back upstream to developers/maintainers. Its only if downstream devs/users are inclined to send code back does it reach upstream devs/maintainers.
I don't think you understand the passage you've quoted (without a link), and you seem to have accidentally added your own words to it (there's a close quote, but then more words.) That being said, I can't find the quote at all in the essay; did AI make it up?
https://www.gnu.org/gnu/manifesto.html
The closest thing I could find was this:
> Software sellers want to divide the users and conquer them, making each user agree not to share with others. I refuse to break solidarity with other users in this way. I cannot in good conscience sign a nondisclosure agreement or a software license agreement.
So were users or "creators" his concern? I don't remember him ever giving too much of a shit about the happiness of creators, I wouldn't have approved. I don't (particularly) care about programmer's problems.
-----
edit:
I can't find the quote "If I sign a nondisclosure agreement to work on a proprietary program" on the entire internet.
[dead]
Two things immediately wrong: Stallman had nothing to do with Open Source; his movement is Free Software, which is at most a precursor to the separate, but sometimes overlapping, ideas of Open Source. Stallman also did not start Free Software so that people could make their creations available as evidence in résumés. He started the movement to empower software users after he felt powerless when confronted by a proprietary printer driver.
> The purpose of Stallman’s open source movement
My understanding is that the purpose of Stallman's free software movement is "that the users have the freedom to run, edit, contribute to, and share the software." The FSF is focused on "defending the rights of all software users." Its about the users, not the developers.
I see what you mean, but this knife cuts both ways. It makes proprietary software easier to write by extracting knowledge from open codebases, but it also makes open source software easier to write by extracting knowledge from those same open codebases.
That's just the main idea, but also:
1. LLMs make existing software (even obscure stuff, so long it fits in the context window) more intelligible:
- how do you compile this (when you are inexperienced and the ecosystem of that language is a baroque mess, it might seem impossible)?
- what does this error message mean?
- what parameters do I need to use in my invocation to get it to do XYZ?
- what does this function do? why does it use this algorithm?
2. They also make new software easier to write, and existing software easier to modify:
- ask about anything concerning the part of source code that fits in a context window, and you'll get a (probably correct) explanation of what it does, faster than a half-dead IRC channel or StackOverflow would respond
- the above, but also: the LLM has infinite patience and can drill down as deep as you want. You can ask "OK, but why?" for as long as you want, as about anything you want. You might get a hallucinated answer sometimes, but a frustrated human who would be asked the same way, could also just make something up to shut you up.
- for anything in the context window, ask about how to go about making a functionality change to add or modify a feature
- the above, but if it's small enough, just get the LLM to write the change for you. It might be buggy and messy, but you'll be one step ahead if you lack the skill to make the change yourself
- how do I set up the build chain? Why is my compiler not picking up the path properly? Is the project directory structure wrong? This used to be a huge problem before LLMs, and relied on undocumented knowledge.
---
For me, the whole point of open source is ready-made, (hopefully) not too buggy components that I can use and customise as an end user, or plug into the thing I am building as a developer. LLMs make the freedom of FOSS become much more practical, particularly to those sympathetic to the movement but technically less experienced.
Well yes exactly. LLMs have increased the value of open source to users. So by reducing the extent of the open source, value is maintained, but rebalanced slightly back in favor of the creator, with their larger closed source piece.
BTW most business-astute maintainers always managed a closed piece of expertise which is what they charged for. I’m saying that proportion needs to grow now.
Free Software is not for resume padding, it’s for free computing.
So it should be easy to reuse your open source code, but not too easy?
A ‘freemium source’ model, where you’re advertising possibility and promoting human-human partnership.
Industry practices that over commoditize human talent are bad IMO.
Our whole industry needs to bend its collective mind to maintaining economic participation. We’ve possibly put too much of a strain on society with LLMs. Especially as more and more people cotton on to what other services they no longer need, as models get better and better. We can’t survive as a species if too much of our lives are based on self-gratification, we have to maintain the drivers that make us interact and learn to get along.
I get the point but it still sounds to me like "what about the horse maintainers in a car world."
The world is changing, we improvise, adapt, overcome. Where many species have their world model encoded in their DNA, we have a beautiful neural computer that can change its world model and resulting strategies on the fly. Let's use it.
> The new approach
That won't work. The breaking of that model is far more widespread than one thinks because of how it was broken.
The breaking of the model breaks underlying models all the way down to the basis for economic distribution of labor.
Its a phase change where labor and expertise are free, without restriction and the people with that expertise do not receive economic benefit for it anymore. In short, your demand curves goes to 0 in that area. There may be a great need for something, but if the demand is 0 no one will fulfill that need. People aren't slaves. Many people conflate demand with need, Hayek in his economics in one book cover the distinction. TL;DR demand is the group of people where there is a point at which two parties are both willing to exchange something for something. Need is where no such crossection between the S/D curve in exchange can occur for the two parties involved. One is much smaller than the other, and at 0, it doesn't happen or you only get the efforts of slaves.
The trend is inevitably towards stalling the economic cycle, where such experts simply do not create such things, they do not share, the ones that could either abandon that expertise or they withdraw keeping it to themselves.
The vast majority of all action though is done for economic benefit, and when that's no longer the case people don't do it. People aren't slaves.
People, professionals, aren’t so stationary. You’re saying that this line on the asymptote is the threshold where incentives die. But the old axes need to be adjusted for new broader possibility. As long as professionals stay ahead of non-professionals by riding the same tools, to keep their position on the boundary of expertise, they will be in demand.
Better to do that by not sharing how as much (source code), but rather what (interactive demos).
People, professionals, aren't so stationary.
You are right, they will do something, and that something will become violent when the rule of order breaks down as a result of these basic underlying aspects.
People aren't static, but neither are they so dynamic that they can learn exponentially, expertise which took education and years to learn cannot compete with free, and lets remember people age and die, and expertise is only gained by doing the job the first rung on that experience ladder. The nature of the asymptote that people fail to realize is, the asymptote isn't at its given value when you see the whole, it trails off as it approaches certain points exponentially. Sometimes to infinity.
You can't compete with slavery, and in the case of AI, you have a digital facsimile of a person that is a slave being used by people that don't have that knowledge. It may be able to go on just long enough that there won't be any experts left by the time everything starts breaking down (which happens frighteningly quickly, typically within 10 years).
There is a train, its on tracks and those tracks go over a cliff, but that cliff is around the bend, and no one knows someone's switched the route to over that cliff, they didn't notice. The conductor keeps feeding the fuel, full steam ahead, and what happens with mass is you have inertia and between the end of the tracks and the rocky shore below? Trains don't stop on a dime.
Its an asymptote, and the nature of cascading failures is that by the time the average person realizes you have a problem, there isn't anything you can do to fix it, that is the difference between average and intelligent, and likewise the average person can't avoid the outcome because the lag and the time to react don't provide a window to change it; its already passed, its hysteresis, and its a bitch.