Google Stitch Validates What We've Been Building
Expression infrastructure for the Agentic Era. We built it before the name existed.
On March 18, 2026, Google updated Stitch with a feature it called vibe design. The idea: describe a feeling, a business objective, or a design inspiration, and the tool generates multiple design directions that match it. They also shipped design.md – a portable markdown file that captures brand guidelines in a format AI agents can read.
We have been building the mature version of exactly this category since 2025.
When a platform the size of Google enters your category, it does not replace your work. It confirms that the problem is real and that the market is ready.
What they named, and what we formalized
Vibe design names a real phenomenon: the shift from specifying components to describing intent. Instead of "a blue button, 40px, border-radius 8" you say "something that feels decisive but not aggressive" and the AI interprets it. This is genuinely useful for early exploration. It is the right direction.
What Google calls vibe design, Designless formalized as expression infrastructure. The distinction matters not as a competitive claim but as a definition of what the category requires at production scale.
Vibe design generates directions. Expression infrastructure resolves them – deterministically, production-ready, available at runtime to any agent building on your behalf. The sketch and the blueprint are not the same thing. They are sequential.
The gap that both approaches address
The AI design landscape – Stitch, every vibe coding tool, every AI builder – keeps bumping into the same wall. Agents build fast. They generate output at a rate no design team can review. But the output is brand-less by default. It is generic unless someone manually intervenes.
Design systems were built for human designers. They give vocabulary: tokens, components, spacing scales. A designer takes that vocabulary and applies judgment. They know when a card needs more breathing room because the content is dense. They know when the hierarchy should be flatter because the audience is in a hurry.
Hand that same vocabulary to an agent and it gets the words without the judgment. It gets the dictionary without the understanding of what the poem is trying to say.
This is the gap that vibe design begins to address – and that expression infrastructure is built to fill.
Design.md is documentation. Expression infrastructure is a foundation.
Google's design.md is a meaningful step. A markdown file capturing colors, typography, spacing, and component patterns in a format that AI tools can read. It solves the portability problem for early-stage work: describe your brand once, carry it into your coding tools.
The limitation is what any static file has: it describes the system without enforcing it. An agent reading design.md learns what the brand looks like. It does not receive the judgment to apply that knowledge correctly at runtime, for every output, across every surface it builds.
Expression infrastructure is not a description of the brand. It is a foundation that agents can query at the moment of decision – receiving guidance that is specific to the context, enforced by the system, and consistent with every other output the brand has ever produced.
What Google calls vibe design, we formalized as expression infrastructure. The category exists. It now has witnesses.
What this means for the Agentic Era
The AI design landscape is now moving in the direction Designless has been pointing since 2025. That is good for everyone building in this space. Category validation is not zero-sum. When the problem becomes visible to a larger audience, the demand for real solutions increases.
What comes after the vibe is the infrastructure that makes the vibe scale. Not the canvas where design happens once, but the foundation that governs design wherever agents build – which is everywhere, at a rate and volume no human team can review by hand.
Designless is expression infrastructure. Not a design canvas. Not a prompting guide. Not a one-time export. The foundation that gives AI agents the ability to invoke design judgment at runtime – for your brand, with your expression, at production scale.