Technology rarely advances because we build larger systems.
It advances because someone discovers a better primitive.
On primitives
A primitive is a simple idea that everything else composes from.
- Files.
- Functions.
- Packets.
- Containers.
Each of these looks obvious in hindsight. None of them were obvious at the time. They reshaped entire generations of software for the same reason: they removed complexity rather than adding to it.
A good primitive is not a feature. It is a smaller vocabulary that says more.
Why now
We believe the next generation of AI-native software will be built the same way.
Not by layering increasingly complicated frameworks on top of today's tools, but by finding the small set of primitives that make entirely new systems possible.
Most of what we call AI infrastructure today is scaffolding. It exists because the ideas underneath have not settled yet. Scaffolding is useful, and it is temporary. What survives is the vocabulary beneath it.
The Primitive Works exists to search for that vocabulary.
How we work
We approach engineering from first principles. When a system becomes too complicated, we do not ask how to manage the complexity. We ask a different question:
What is the primitive?
The answer is usually smaller than expected.
Sometimes it is a new abstraction.
Sometimes it is removing one.
Sometimes it is recognizing that two different problems are the same problem viewed from different angles.
We build the smallest thing that proves the idea, then let real use decide what it becomes. A primitive earns its place by being composed, not by being announced.
What we build
Our work spans research, open-source software, and products.
Research is how we test whether an idea is true. Open source is how we find out whether it is useful to anyone besides us. Products are how an idea reaches people who should never have to care about its internals.
We build in public because the most useful primitives rarely belong to a single company. They become shared infrastructure that thousands of builders improve and compose in ways nobody originally imagined.
What we value
- Simplicity over cleverness.
- Composition over configuration.
- Understanding over imitation.
- Enduring ideas over temporary trends.
Technology changes quickly. Principles rarely do.
Our ambition is not simply to build successful software.
It is to leave behind better primitives than the ones we inherited.