Croft

The value of intelligence compounds through continuity. Every meaningful human relationship becomes more valuable over time because understanding accumulates.

A doctor becomes more effective after years of knowing a patient. A teacher becomes more effective after years of knowing a student. A lifelong friend understands things that never need to be explained.

We believe intelligence should work the same way.

Today’s AI stores memories. Human Models build understanding.

Today’s AI improves incrementally. Human Models compound.

The most valuable intelligence is not the one that answers best today. It is the one that understands you better tomorrow.

Every generation of AI has introduced a new foundational capability. Language Models learned to understand language. Multimodal models extended that understanding across text, image, audio and video. Reasoning models learned to solve problems.

None of them continuously understands people.

A Human Model does. It continuously builds and refines a persistent understanding of an individual — their context, goals, patterns and history — that deepens over a lifetime rather than resetting with every conversation. We believe the next major breakthrough in AI is not greater intelligence alone. It is persistent human understanding.

Understanding, not memory. Storing facts is not the same as understanding a person. A Human Model learns patterns, distinguishes the momentary from the stable, and grows more useful through every shared experience.

Continuous over a lifetime. It holds continuity across years rather than sessions.

The most intimate technology should answer to you. We believe ownership, alignment, and trust are not features but foundational requirements — properties of the architecture, not policies added afterwards. The intelligence a person relies on should be shaped around their needs and values, and never subordinated to anyone else’s incentives.

Shaped around one person. Not a generic assistant shared by millions, but a trusted intelligence built around an individual — their context, their priorities, and the life they are trying to build.

Owned by the individual. An understanding of you, accumulated over years, is too consequential to be locked inside someone else’s platform. You can see it, correct it, delete it.

Sovereign by design. We build in Europe, on European infrastructure, under the world’s strongest privacy standards — because personal AI at this depth requires trust that cannot be retrofitted. Privacy is not our constraint; it is our starting point.

More than powerful models. Persistent human understanding cannot be bolted onto today’s architectures as a feature. It takes a different architecture — systems that maintain continuity over time and run reliably in the background of everyday life: dependable, efficient, and private by default.

Present wherever life happens. People do not live through text boxes. We see natural interaction through voice, vision, and language as essential to an intelligence that genuinely supports a life — and understanding is what makes natural interaction possible.

Research and product are inseparable. We believe the strongest research comes from solving real problems for real people, and the best products are informed by deep technical understanding. Each sharpens the other.

We measure what truly matters. We judge ourselves by whether our systems understand the people they serve better over months and years — not by attention, engagement, or metrics that happen to be easy to optimise.

Built for the long term. We aim to build carefully and responsibly: technology that expands human capability while remaining firmly on the side of the individual.

The relationship between a person and their intelligence will become one of the defining relationships of this century. We believe it should be personal, continuous, trustworthy, and ultimately belong to the person it serves.

That is why we are building Human Models. That is the future we are building at Croft.

We’re building AI that pushes technical boundaries while remaining unambiguously on the side of the person using it. Our team combines rigorous engineering with creative exploration, and we’re looking for collaborators to help shape this vision.