What AI can achieve today is only the beginning

The foundations we build now must enable it to work alongside us, expanding our capabilities while preserving our independence.

Our ResearchSend your ideas

Forlais’ Five Futures

Intelligence is still an open question

The approaches we use today should not define the limits of what intelligence can become tomorrow.

A timeline from early computing machines and Alan Turing, through mainframes and the first personal computers, to a network of light in the shape of a brain.

Forlais pursues original research starting from the foundations of AI.

Our ResearchSend your ideas

We want to expand both what intelligent systems can do and our understanding of how they do it. That requires a willingness to question established assumptions, investigate different approaches and let results determine the direction.

There are deeper questions to investigate: how understanding develops, how relationships become knowledge, and whether intelligence can be represented in fundamentally different ways.

AI should work alongside people, not in place of them

The measure of an intelligent system is what people can do with it, and what they keep for themselves.

A woman works at a desk with a laptop and a large monitor of research notes, beside a window overlooking a city skyline at sunset.

Forlais builds AI to extend human capability, with people kept in the decision.

Our ResearchSend your ideas

What AI used, what it changed and what happened along the way should be kept as a record people can examine.

Persistent memory and clearer accountability are there so that a person can follow the work, question it and decide.

The most useful questions are often the ones nobody asks

The thoughts most people dismiss out of hand are often the ones worth chasing hardest.

A desk of books, a notebook and a laptop before a glass wall chalked with a diagram from ideas and research to real-world impact, a city skyline beyond.

We chase ideas wherever they lead.

Our ResearchSend your ideas

We watch for the anomaly that does not fit. The detail most would tidy away is often the door to something greater.

We do not overclaim or hide. A result that looks miraculous is a reason to test harder, not to celebrate.

Advanced AI should be dependable, and within reach

Frontier capability means little if it cannot be trusted, examined and afforded.

Three colleagues work late around screens showing network diagrams in a dark office, with city lights beyond the windows.

Forlais designs for reliability, auditability and lower operating cost from the architecture up.

Our ResearchSend your ideas

CNA, our floatless AI architecture, is central to model efficiency, reliability, auditability and operating cost.

Talos is our work on making advanced AI practical in local and offline environments.

People and organisations should choose how they use AI

Where it runs, what it can reach and what it keeps should be decisions for the people who rely on it.

A hiker with a backpack stands on a rocky ridge, looking out over a mountain lake and cloud-filled valleys at sunrise.

Forlais works across every layer so that the choice is real.

Our ResearchSend your ideas

From model architecture and memory to platforms, infrastructure and specialised hardware, working across the whole stack gives greater control over how technology is designed, deployed and improved.

EvaEsi reaches the web, desktop, mobile, enterprise and the API, and Talos is our work on bringing advanced AI to local and offline environments.