Standing on the Shoulders of Repos
In 1980, a printer jammed at MIT and it changed the world.
Richard Stallman was a programmer at MIT’s Artificial Intelligence Lab. Their old printer jammed constantly, so he did what any good engineer would do—he modified the software to alert users when it got stuck. Problem solved. When Xerox donated a new laser printer, he went to do the same thing. But this time, Xerox wouldn’t release the source code. Fine. Stallman tracked down a researcher at Carnegie Mellon who had access to it. The guy refused to share. He’d signed an NDA.
Stallman was livid. Not because of the printer. Because of what the printer represented.
The Nobility of Tools
For as long as we’ve been a species, progress has worked the same way: someone makes something, someone else makes it better. Tools carry that progress forward in a way that nothing else does. Knowledge shifts with context and truth can bend with perspective, but a sharper blade cuts regardless of who holds it. Either a tool works or it doesn’t.
And when they work, they multiply. A wheel is useful. Put it on a cart and it moves civilizations. Strap an engine to that cart and it reshapes the planet. Tools compound on top of each other with each one amplifying the last.
For most of history, this was automatic. Tools were inherently open. You could take apart a plow, study how it worked, and build a better one. A blacksmith could examine any blade and improve it. Mechanical tools wore their logic on the outside—gears, levers, springs. The design was the tool. Software changed that. For the first time, a tool could be sealed shut. You could use it but never understand it, never improve it, never build on it. That’s what Stallman walked into at MIT. Open source didn’t invent a new principle—it restored the oldest one we have.
Locking the source code didn’t just prevent Stallman from fixing a printer. It broke the chain. It stopped the next person from building on what the last person built.
So he spent the rest of his life making sure it couldn’t happen again. He launched the GNU Project in 1983 and the Free Software Foundation in 1985. His work laid the foundation for Linux, which today runs the majority of the world’s servers, phones, and supercomputers. Git, React, Python, Kubernetes—the infrastructure we all build on exists because people kept improving each other’s work. Open source turned software into the most powerful compounding system ever created. One repo builds on another builds on another.
Compounding in the Wild
There’s plenty of debate about what AI will or won’t become. But in terms of raw progress, the test is simple: can we build on what others have done before us?
The answer is, of course, yes, and open source is a big reason why.
Take OpenClaw, the open-source AI agent created by Peter Steinberger that became the fastest-growing repository in GitHub history. Steinberger’s own aha moment came when he absent-mindedly sent a voice note to his agent over WhatsApp, not knowing if anything would happen. The agent received the audio file, checked the file headers, found ffmpeg on the machine to convert it, realized Whisper wasn’t installed locally, found an OpenAI API key sitting on the system, called the transcription API, and replied. All in seconds. All without being told how.
The agent stacked tools—ffmpeg, Whisper, OpenAI’s API—each one built by different people, across years, across continents. Steinberger didn’t build any of them, he may have not even known of them. His agent found their work and assembled it on the fly.
This is happening everywhere now. Developers point Claude Code at a codebase and say “figure it out.” The agent reads the repo, understands the patterns, installs dependencies, and starts building on top. Someone publishes an open source skill for voice transcription and within days it’s running inside thousands of other people’s agents. A researcher releases a dataset and before the week is out, three teams have built tools on top of it.
Open Source Is the Multiplier
Stallman understood something in 1980 that feels more relevant now than ever. Right now the conversation is all about agent intelligence—reasoning, autonomy, tool use. That matters. But it’s like praising a chef’s technique without mentioning the ingredients. Memory gives an AI statefulness over knowledge—it can remember what you said yesterday. Open source gives AI statefulness over progress—it can pick up where thousands of developers left off. That’s the part we need to celebrate far more.
The agents that handled Steinberger’s voice note didn’t invent ffmpeg or Whisper. Thousands of contributors built those tools and left them open for anyone—or anything—to use. That decision, repeated across millions of repos, is what turns an AI agent into a genuine multiplier.
Every repo is a rung. Every tool is a shoulder. And for the first time, we have machines that know how to climb.