A design system is no longer just a contract. It's an interface for machines.
Every token, every component, every naming convention is read, interpreted, and reproduced by AI — at a scale and speed no human review can match. When an agent generates a screen, it doesn't reason about your design intent. It pattern-matches against what your codebase actually says. If your system drifts — a hardcoded hex here, an undocumented component there — the machine doesn't catch it. It amplifies it.
We believe this changes what a design system is for. It's no longer enough for it to be beautiful, or even consistent. It has to be legible to machines: deterministic, self-describing, structurally honest. We call this AI readiness — and today, almost no one can measure it.
So we decided a score should be a fact, not a feeling.
Most quality tools hand you a number and ask for your trust. We refuse that. Every claim Lyse makes is falsifiable: same input, same score, byte-for-byte. Every rule's false-positive rate is published. Our entire corpus and methodology are open. We built a continuous validation infrastructure — The Bench — that runs Lyse against hundreds of real-world codebases, every day, in public. Without it, a score is an opinion. With it, it's a measurement.
And we believe this infrastructure should belong to everyone.
Not just teams with a dedicated design-systems org. Not behind a black box, not gated by a sales call. The engine is open source. Your code never leaves your machine. The drift you can't see today shouldn't require a vendor's permission to measure.
This is our deep conviction: the gap between what a design system intends and what its code actually does is the most expensive, least measured problem in software — and it's about to get worse. We're here to make that gap visible, measurable, and closeable, for every person who maintains a design system, and for every machine that builds on one.
That's the infrastructure we're building. In the open. Starting now.
Maxime, Thomas and Noe ❤️