Flash Design Genome

Flash - Design system management system

[Flash Design Genome @flashdgenome]

People build with design systems.
Agents build with design genomes.
Flash turns yours into one your agents can actually read. Change one token, everything moves.

Build your first Genome free → https://t.co/5KgS09lPZW

[Flash Design Genome @flashdgenome]

UI consistency shouldn’t be something you check for in review. It should be what you build from.

[Flash Design Genome @flashdgenome]

@nondualrandy It’s not entirely the AI’s fault

It can only reuse what it can see, and most design systems are spread across Figma, code, and docs that don’t agree

So it guesses, and guessing sometimes means the wrong color, button, or component. Where does yours live?
https://t.co/qzrD6Z9itK

[Flash Design Genome @flashdgenome]

Five hex codes. One brand blue. Which one is right?

Figma, code, the token file, Storybook, and the docs each say something slightly different. People ask in Slack. An agent can’t. It guesses.

Where does your team’s real version live?

[Flash Design Genome @flashdgenome]

Five hex codes. One brand blue. Which one is right?

Figma, code, the token file, Storybook, and the docs each say something slightly different. People ask in Slack. An agent can’t. It guesses.

Where does your team’s real version live?

[Flash Design Genome @flashdgenome]

So we have design systems to help designers and developers speak the same language.

Yet designers still hand off Figma files for developers to interpret, and teams still catch drift after shipping.

The system exists. But why isn't it doing enough of the communicating?

[Flash Design Genome @flashdgenome]

@zoink The interesting challenge with non-deterministic UI is defining what can change and what can't.

An agent should be able to adapt an interface to the task without reinventing the product's patterns and relationships every time.

[Flash Design Genome @flashdgenome]

@sahinabut The key is separating what you're trying to test from what the product has already figured out.

If the agent can build on the existing system, you can experiment with new interactions in a working flow without it reinventing the rest of the UI.

[Flash Design Genome @flashdgenome]

@Eric_M_Stevens The “why” rarely lives in the design system itself. It’s scattered across product flows, code, docs, and people’s heads.

The challenge is connecting those pieces so an agent understands why one choice makes sense here over another.

[Flash Design Genome @flashdgenome]

@jacobgold @GeorgeMayer Large teams already struggle when product rules are scattered across code, docs, and people's heads. Agents inherit that same fragmentation.

A shared system gives both people and agents the same context to build from.

[Flash Design Genome @flashdgenome]

@letstri @theo Elements that look alike but behave differently are tricky for AI.

Humans may spot the difference, but agents need to know a component's purpose, where it belongs, and how it connects to the product.

Tokens and components alone don't explain the system behind the interface.

[Flash Design Genome @flashdgenome]

https://t.co/a0Pon4pekm

[Flash Design Genome @flashdgenome]

Design system: List of tokens and components for human reference

Design Genome: A machine-readable model of your DS

That difference matters when the reader is an agent.

See what we mean 👇

[Flash Design Genome @flashdgenome]

@benjaminakar @raunofreiberg We’re building Flash around exactly this.

The canvas is for exploring visual ideas. Everything after that, agents should pull from a live, machine-readable model instead of hallucinating what your product looks like.

https://t.co/a0Pon4pekm

[Flash Design Genome @flashdgenome]

@dotmariusz @raunofreiberg We spent a decade making design systems easier for humans to browse. Now the primary reader may not be human.

[Flash Design Genome @flashdgenome]

@raunofreiberg how do you prompt for an idea you haven't had yet? That's the canvas earning its keep.

What I don't get is going back through it for work where nothing new is left to discover.

[Flash Design Genome @flashdgenome]

@ellaomoni @aaapinen You're right that it's not a token, though. Even once you decide the value, there's still no record of the decision, or that the state exists at all, until someone builds it.

[Flash Design Genome @flashdgenome]

@carlomakes This is one of the real problems with Figma though.

It’s good for building a design system that doesn’t exist yet, but not so good at tracing where a spacing issue came from, let alone fixing it.

[Flash Design Genome @flashdgenome]

@hakimuddinkika https://t.co/a0Pon4oGuO

[Flash Design Genome @flashdgenome]

@sankalpdomore People build with design systems. Agents build with Design Genomes.

[Flash Design Genome @flashdgenome]

@rehanxahmed @sankalpdomore Figma is great for exploring what doesn't exist yet.

If you make such changes in code while using a normal design system, they won't reflect back in the system.

This is the advantage a live source has over a static reference: changes can flow both ways.

[Flash Design Genome @flashdgenome]

@uxisaac_ Design systems in Figma are created to establish consistency between design and code.

But the system is rarely what does the communicating with code. Flash is built around making the system itself the shared source.

Check it out: https://t.co/a0Pon4oGuO