This hit me hard. I contributed thousands of hours of labor into this library and for what? To be kicked out after somebody else starts charging admission for it? Jobs are extremely difficult to come by now?
What is "The Library"?
The aggregate collection of text contributed by any given single individual is known as The Library. This Library that became AI through a large language model was a gathering of each The Library from the collective mind — mine included — yours included perhaps — ingested into their algorithms for predictive text generation.
Many times I would find myself reading StackOverflow™ in a bind between wanting to figure out my thing and correct the record when a post didn't quite get what I needed correct. That back and forth, contributed by me — and by you perhaps — was The Library that these AI companies possessed (nicest way to describe legal theft) and then turned into a subscription. A subscription based on usage now, not even time. How much do you need The Library and how much are you willing to pay? That's what Dario is looking for. That's what Sam is looking for.
My figtree package is in The Library in a way that is both seen and unseen. I didn't consent to that work being uploaded, so why do I pay a subscription to access The Library that contains my works *and their predictive derivatives? Those predictive derivatives are where the intellectual property is captured beyond the raw code itself — the *between the lines meaning from project A to project B, what a competitor or bad actor could make against my own efforts.
The Problem With Open Source
I have contributed 99+ open source packages on GitHub. I'm a 5-digit user ID from 2009. And I see a problem with open source in 2026.
Look at what happened in June. Anthropic launched Fable 5, and within three days the US government invoked national security export controls and forced it offline worldwide. Think about what that means: models trained on our freely-given contributions are now considered so dangerous they get handled like weapons. Knowledge went in open and unconsented. It came out closed, priced, and government-controlled.
So why would I release anything open source anymore? Sam, Dario and Elon — along with thousands of others — will just take the work, incorporate it into the LLM itself, and use my own knowledge against me. I would be foolish to think open source could survive this. Any engineer who runs the numbers arrives at the same place: closed source is the future. Private repos. Stop training your replacement for free.
I ran those numbers too. I arrived at the same place.
Romania
And then I remembered that I have already lived in a closed-source world.
I've been involved in the internet since the day I arrived in the United States from Romania — the era when the University of Maine ran BITNET, before the good people of Maine had access to TCP/IP. My programming career began at 9 years old because of the University. By 11 I had the internet. In 1994 I was doing computer work beside my mother in Neville Hall in Orono — the same building that ran the state of Maine's connection. Nobody charged us admission. Thousands contributed willingly because the incentive was to learn and better yourself, and an orphan from Ceaușescu's Romania became an engineer at Cisco, Oracle, and WB Games because that door was open.
The Internet is going through a revolution like Romania did after Ceaușescu's Christmas Present 1989, and AI caused it. But I know what grows behind closed doors. I grew up in it.
So No
Everything above this line says I should close my source. I'm not going to, and you should know it's not because the theft isn't real — it is, and I want it remedied: disclosure of training data, consent for future ingestion, compensation for what was taken.
But somewhere there is another 11-year-old at a public terminal, and my code costs them nothing.
Don't close-source your contributions for anyone's bottom line. Be like me — Andrei — and continue giving selflessly, expecting nothing in return, more useful open source utilities that the greater good actually needs.
Why do you still publish open source — or why did you stop? I'll respond to every comment.
Top comments (7)
"Any engineer who runs the numbers arrives at the same place: closed source is the future." That line hit hard because it's true for anyone who's been doing this long enough to see the pattern repeat. The thing that keeps me on the open side though isn't idealism â it's the observation that the most resilient open-source projects I've seen are infrastructure-level, not application-level. When the value is in the protocol or the embedded runtime rather than the hosted service, enclosure just doesn't stick the same way. The incentive mismatch is real but not total â what's your take on where in the stack open source can still survive without needing to be a charity case?
You've located the fault line. Enclosure sticks where the value is the knowledge inside the code — and slips where the value is the code's position: the protocol everyone speaks, the runtime everyone embeds, the dependency everyone trusts. AI is knowledge-enclosure, not service-enclosure.
Counterintuitive angle worth sitting with. AI trained on open source could quietly reduce the incentive to publish it, if the model becomes the interface to the knowledge instead of the code. The thing that erodes isn't the license, it's the why: the discussions, the commit history, the decisions behind the code that a generated snippet strips out. Open source was never just the source; it was the reasoning around it. That's the part a model flattens.
This was a thoughtful read. I can understand the frustration, especially for people who've spent years contributing to open source. I still hope we find a better balance between protecting creators and keeping knowledge accessible for the next generation.
By the way, there are also lots of great open-source projects collected here: webcurate.co/c/open-source.
The part that travels is the interface shift: the model became the read path to The Library, so the payoff for publishing moved from 'be findable' to 'be the source the model treats as canonical.' That is a real loss for the reputation economy that made contributing worth it. Attribution, not the code itself, is what got laundered.
But I would push back on closed-source-as-the-answer. Open source was never that the code was secret; it was the right to fork, read, and run, and training does not touch that right. What it breaks is provenance. So the fix that actually addresses your grievance is attribution at the model layer, not withdrawing the code. Closing your source mostly denies the next builder the door you walked through, without pulling your work back out of what is already trained.
You've named the two halves of it. The right to fork, read, and run survives training untouched — the code was never the secret. What got laundered was attribution, and the reputation economy that made contributing worth it went with it. So your remedy sits at the model layer. So does mine. Credit is half the fix; the other half is that The Library cannot remain the source of truth while the read path pays the write path nothing. That's what x402 addresses on the agentic web: the machine that treats my work as canonical pays per read, while the human who forks it still pays nothing. Open source leveled my abilities inside closed-source shops — Cisco, Oracle — so I won't withdraw the repos and shut the door I walked through. Keep it open for people. Charge the machines admission. Attribution and compensation, both at the read path — neither requires closing the source.
Open source still feels like the fastest way to level up as a builder.
If that door closes, the next wave of devs loses the same path a lot of us had.