irys · Pre-Seed Memo

For most of human history, meaning was not encoded through letters. Before alphabets, humans thought, remembered, and transmitted culture through Symbols. Symbols were denser than language. They held meaning, memory, mythology, status, ritual, and aesthetic judgment all at once. A Symbol was never just representation. It was compressed intelligence.

Then language became dominant. Letters became the interface through which humans encoded thought. Modern AI inherited that interface. Large language models reason through words, tokens, and prompts. The entire generative era is built on one assumption: if you want AI to create, you must tell it what to do through language. irys breaks that assumption.

Build the world's first symbolic language of taste.

A Symbol is a curated aesthetic concept that AI can interpret consistently: users select Symbols that already represent an aesthetic instead of describing it through prompts. Today, the archive contains more than 6,000 manually curated Symbols, organized into a proprietary taxonomy developed from decades of experience across fashion, design, architecture, branding, and visual culture. Symbols compose into Patterns: discoverable aesthetic identities that can be applied consistently across creative work. Rather than organizing images, irys organizes the logic behind them. We call the decoding system built on top of this archive Taste Intelligence (TI).

We're building a symbolic knowledge system for taste. AI is the interface that makes it possible.

Over the past twenty years, Guido Callarelli has built brands, products, and creative systems across fashion, luxury, and technology, working with companies including Supreme, Versace, Nike, Off-White, The Attico, Valentino, Kelly Wearstler, and Zendaya. He also worked as an Art Director within Kanye West's DONDA, alongside Virgil Abloh and Matthew Williams, contributing to one of the most influential creative organizations of the last decade.

Throughout that work, one observation remained constant. Creative execution continuously evolved. Taste never did. It remained implicit, living inside people rather than systems. The symbolic archive behind irys is the result of converting two decades of accumulated creative experience into a structured representation before asking AI to reason over it.

The founding team includes:

• Taylor Flynn · Founding Full-Stack Engineer (LOI signed)
• Mahnoor Rafi · Founding Designer, ex-Apple (LOI signed)
• Mikhail Vasilkovskii · Founding Technical Advisor, The Bot Company (ex-Snap), author of 15+ publications across CVPR, SIGGRAPH, AAAI, and ICCV.

Recruitment is underway for a Founding AI Engineer and a Founding GTM Engineer. Alongside the core team, irys is assembling an advisory board of academics, historians, and domain experts to validate the symbolic taxonomy, and has signed a founding cohort of industry ambassadors helping shape the archive and its early adoption across the creative industries.

Taste in AI is trapped behind prompts.

Taste is already inside today's models: they contain every major aesthetic ever learned, but it remains buried behind prompts, statistical averaging, and trial-and-error with no structured way to access it. Creatives fight to bring their taste in: they prompt, iterate, and get something close. Never the thing. Never consistently. Everyone else struggles to get taste out: without structure it can only be learned through years of observation, references, and proximity.

Those with taste can't bring it in. Those who want it can't get it out. irys surfaces it, making taste accessible through Symbols for the creatives who work with AI and anyone who wants to learn it.

Symbols are the building blocks through which irys represents taste.

Built on principles from Neurosymbolic AI, irys pairs the fluency of large language models with structured, human-readable Symbols. Each Symbol is manually curated, categorized, and visualized through a proprietary symbolic language, creating a consistent representation that both people and machines can understand. Together, these Symbols form nos, irys's proprietary symbolic archive and core intellectual property.

Unlike prompts, Symbols are persistent. Unlike images, Symbols are reusable. Unlike embeddings, Symbols remain human-readable. Symbols are organized within a proprietary, evolving taxonomy consisting of twelve Symbol typologies, divided into two complementary layers.

The symbolic language continuously evolves as new Symbols are curated, expanding both its expressive power and the universe of discoverable Patterns.

A Pattern is a discoverable aesthetic identity — a persistent, reusable expression of taste that can be applied consistently across different subjects and creative disciplines.

Built from a consistent symbolic language rather than natural language, Patterns become reusable, searchable, comparable, and machine-readable. Instead of reconstructing taste through prompts for every project, designers compose with the same symbolic building blocks, creating aesthetic identities that remain persistent across time, people, and creative disciplines. Patterns can be discovered, modified, applied, and continuously refined while preserving their underlying logic.

For example, the discovered Pattern Marble Quattrocento (Florence, 15th century) composes six Symbols — Italian Rationalism (Style), Champagne Silk and Gilt (Palette), Bianco Carrara Marble and Brass (Materials), Rectangle (Form), Honed Matte (Texture) — into a persistent aesthetic identity: Carrara marble held in north-lit studio calm, a single point of warm gold interrupting the cold field. The same Pattern applies consistently across furniture, fashion, and product.

Patterns are the fundamental knowledge and data asset inside irys.

When a Pattern is discovered for the first time, irys recognizes it as unique, assigns it a canonical identity, and attributes the discovery to its first discoverer. Every identical composition always resolves to the original Pattern. Discovery is first come and irreversible.

Built on today's Symbol taxonomy, there are over 1 quadrillion discoverable Patterns. At 10 million discoveries a day, exhausting the map would take 270,000 years. The map is effectively infinite. Every discovered Pattern reveals another part of it. irys curates Symbols. The community discovers Patterns.

One language. Two modules.

irys launches as a single platform that expands through modules. Every module is built on the same symbolic language. Every module compounds the same Pattern dataset.

One Language. Many Products. Discover and Design are the first two modules on the same symbolic foundation. Every future module — from brand governance to team collaboration to industry-specific verticals — builds on the same archive and compounds the same Pattern dataset. The platform grows without fracturing the language.

Foundation models will keep improving. irys is model agnostic. The advantage is three compounding assets.

Better models execute Symbols more faithfully, so model progress raises the moat rather than eroding it.

irys launches in industries where taste is a strategic advantage, including fashion, interiors, furniture, product design, architecture, and branding, before expanding into adjacent creative disciplines. The creative-software layer these tools sit in is worth roughly $15B and growing about 10% a year — but the addressable audience is far larger. Adobe's Future of Creativity Study counts more than 303M creators globally, most still building without structured access to taste. That's the base irys grows into.

The platform is Creator-first, with Professional and Custom tiers unlocking as usage matures.

Creator-first. Professional and enterprise tier up.

irys follows a tiered SaaS and credit-based model on one archive. Both Discover and Design ship in every tier from day one — creators use them lightly, professionals use them fully, and enterprises use them across teams and institutions. As usage matures, Professional and Custom tiers unlock more powerful workflows while remaining on the same symbolic language.

Subscriptions provide predictable recurring revenue, while usage-based credits monetize compute-intensive workflows such as Pattern decoding, fragment board generation, and AI-assisted creation.

For the first time, irys makes culturally grounded taste accessible to anyone, not only trained professionals. Discovery itself is the growth engine: every Pattern a user reaches is permanently named and credited to them, a shareable claim that markets the platform with every discovery and bridges exploring in Discover to composing in Design. Organizations then adopt irys to preserve and scale taste across teams.

ARR shown is subscription-only and assumes a ~$15K average annual Custom contract; usage-based credits are additional upside. Reaching under 1% of the world's 303M+ creators gets irys to $1B ARR. There is room for orders of magnitude more.

Give Intelligence an Eye.

Language gave intelligence a voice. Symbols give intelligence an eye. For the first time, taste can exist as structured knowledge rather than intuition. Every Symbol curated expands the language. Every Pattern discovered expands the map.

What begins as a platform for discovering and designing with taste becomes an ever-growing symbolic representation of humanity's aesthetic knowledge. Just as language became the foundation for modern AI, we believe symbolic representations of taste can become the foundation for a new generation of culturally aware creative intelligence, fluent across disciplines, and capable of carrying our history forward.

To make that possible, we're building the world's first symbolic language of taste.

$1M Pre-Seed on a SAFE

This round is designed to transform the symbolic archive into a production product. Target runway: ~12 months.