AI research & development — everything open source

Building the substrate of machine intelligence.

Glyphh is an AI research and development company. We work at the frontier of vector-symbolic architectures, hyperdimensional computing, memory systems, and agentic runtimes — and everything we build ships in the open.

Explore the research github.com/glyphh-ai ↗

{ THE FRONTIER }

A search for a cognitive primitive

Deep learning got us here. We believe the next layer is memory, structure, and the machinery agents run on. The frontier is a mathematical primitive models can leverage as a cognitive substrate.

01vsa

Vector-Symbolic Architectures

Structure you can compute with. We represent knowledge as compositional high-dimensional vectors — binding, bundling, and permuting symbols algebraically instead of approximating them.

02hdc

Hyperdimensional Computing

Inference in microseconds, not GPUs. Pure-HDC encoders classify, parse, and route language with character-level precision — no LLM in the loop.

03memory

Memory Systems

Memory that never confabulates. Ada writes every fact into a universal schema at write time, so recall is a deterministic scan — exact, versioned, auditable.

04runtime

Agentic Runtimes

The machinery agents run on. Runtimes that give models tools, state, and compute — from the desktop to the cloud — over open protocols like MCP.

{ PRODUCTS }

Research you can run

We're a software company too. Our products are built directly on the research — and stay open source, all the way down.

Glyphh Desktop

available now

The workspace your AI drives. A native desktop app for Mac and Windows — your apps, windows, layouts, and 3,500+ connected services behind one secure MCP relay.

Explore Glyphh Desktop →

{ open source }

Every product is built on the research — and open source, all the way down.

github.com/glyphh-ai →

One pricing model. Every product.

All Glyphh products share the same credit system: 1 credit = 30 seconds of compute. Each product draws credits differently, but the meter never changes — and your AI is always your own, never marked up.

See pricing

{ WHY “ADA”? }

Named for the first programmer

Our memory research is named for Ada Lovelace — because she saw, before anyone, what we're still building.

The Analytical Engine weaves algebraic patterns, just as the Jacquard loom weaves flowers and leaves.

Ada LovelaceNotes on the Analytical Engine · 1843

In 1843, Ada Lovelace published the first computer program — and then went further than the machine itself. She argued that an engine could act on symbols, not just numbers: that if relationships could be represented, a machine could compose music, reason over language, weave structure.

That is the oldest statement of the bet we're making. Vector-symbolic architectures treat knowledge as symbols you can compute with. Our memory system stores facts as structure, not approximation — so it recalls exactly, the way Ada imagined machines would: patterns woven, never guessed.

1815Augusta Ada Byron is born in London.
1843Note G — the first published computer program, written for a machine that didn’t exist yet.
1852Ada dies at 36, a century before her program could run.
todayAda — our memory system — carries her name and her idea forward.