Foundry Zero Manifesto
The machines we use limit what we can build. It's time to tear them apart.
In 1993, Shuji Nakamura did something most researchers won't dare: he took his MOCVD reactor apart. Completely. Then he put it back together exactly how he wanted. He bent steel pipes, changed angles, welded quartz tubes, cut high-purity carbon, rewired the system, altered gas nozzles.
The result? The blue LED. A Nobel Prize. An entire industry.
The breakthrough wasn't just in the physics. It was in the machine.
The Problem
Most researchers today cannot do what Nakamura did. Not because they lack skill, but because:
Warranties stop them. Open the machine, void the warranty. The incentives are clear: don't touch.
Bureaucracy stops them. Equipment is institutional property. Modifications require approvals. Approvals require justifications. Justifications require certainty about outcomes you can't predict.
Knowledge stops them. Modern equipment is complex. Documentation is scattered. Tribal knowledge is locked in the heads of technicians who left years ago.
This isn't a small problem. This is a fundamental constraint on what's possible in advanced materials, manufacturing processes, and physical technologies. You cannot discover what no machine can make.
The Insight
Nakamura's success came from intimate knowledge of the apparatus. He understood not just what the machine did, but how it worked, where it could bend, what could be changed.
That knowledge is rare. And getting rarer.
Equipment has become more sophisticated. Supply chains have consolidated. The gap between "user" and "builder" has widened. Most researchers are operators, not modifiers.
What if that knowledge could be democratized?
The Answer: Foundry Zero
Foundry Zero is a skunkworks for advanced materials manufacturing. But it's not about building new machines from scratch. It's about understanding the machines we have—completely.
Here's how it works:
Start with real manufacturing equipment. MOCVD reactors. CVD systems. Plasma etchers. Sputtering tools. Wire bonders. The apparatus that actually makes things.
Disassemble completely. Document every component. Photograph every connection. Measure every dimension. Understand the design choices—and the compromises.
Build a complete knowledge base. Not just manuals—real understanding. Material specs. Thermal profiles. Gas flow dynamics. Electrical schematics. Failure modes. Vendor alternatives.
Train AI systems on the complete knowledge base. Not just "how to operate" but "how to modify." The AI becomes a modification partner, not just a manual.
When a researcher has a new material or process in mind, the AI suggests modifications: "Change this gas nozzle geometry." "Adjust this heating profile." "Replace this component with this alternative." Actionable. Specific. Grounded in real knowledge of the machine.
Why Now
Three things make this possible today:
AI can actually help. Modern AI can process complex technical documentation, understand mechanical and electrical systems, reason about modifications, and suggest alternatives. It's not perfect, but it's good enough to be useful.
Used equipment is available. The semiconductor and materials industries upgrade constantly. Older-generation equipment that's still perfectly functional gets retired. It's available. It's affordable. And it's good enough for research and small-scale production.
The cost of not doing this is increasing. As technologies become more sophisticated, the gap between what's needed and what's commercially available widens. Relying entirely on equipment vendors to build what you need is slow, expensive, and limiting.
What This Enables
Foundry Zero isn't just about understanding machines. It's about changing what's possible:
Faster iteration. Modify the tool to match your process, not the other way around.
Novel processes. Combine techniques in ways no vendor would build as a standard product.
Lower barriers. Researchers and startups can access capabilities currently locked behind million-dollar budgets and vendor relationships.
Real innovation. Not just new materials, but new ways of making them.
The Vision
"Without the ability to modify the apparatus, you cannot discover processes the apparatus wasn't designed for."
Foundry Zero is a bet that Nakamura's approach can scale. That the deep knowledge required to modify manufacturing equipment can be captured, structured, and made accessible.
The goal isn't to replace equipment vendors. It's to give researchers the agency Nakamura had: the ability to bend the machine to their vision, not bend their vision to the machine.
The machines we use should not limit what we can build.