Independent AI Research Lab

AI should be designed
around humans,
not in spite of them.

We explore what becomes possible when human cooperation is not the goal of AI design — but its foundation.

Read the research → About Horizons
Scroll

Research pillars

01

Alignment &
Explainability

AI that works with humans must first be understood by them. We research the conditions that make AI systems transparent, predictable, and trustworthy.

02

Human-AI
Cooperation

Cooperation is not a feature — it's a design principle. We explore what systems look like when human agency is preserved and amplified at every layer, including the cognitive foundations that make genuine cooperation possible: memory, perception, planning, and the architectures that emerge when these are designed around the human, not around the task.

03

Human, Society &
Environmental Impact

Every AI system exists in a world with real costs and consequences. We evaluate progress not just technically, but in terms of civilizational benefit.

Active research

All writing →
Can a language model become a genuine strategic collaborator? A Catan experiment.
Exploring shared mental models between humans and AI under conditions of uncertainty, resource scarcity, and social negotiation. Entry 01: setup, hypothesis, and first observations.
On cooperation as a design principle: why the framing matters.
Most AI systems are built to complete tasks. We argue that the unit of design should not be the task, but the human-AI system as a whole.

A lab with
a point
of view.

Horizons was born from a frustration and a conviction. Most of the conversation around AI centers on productivity, on flashy features, on headlines designed to capture attention rather than provoke thought. AI is sold as a product — one that replaces workers, inflates valuations, and scales at any cost — without stopping to consider that we are building the cognitive infrastructure of our civilization.

The conviction: the same way humans have always built great things by cooperating with each other, artificial cognition should be no different. Not as a tool to extract more output, but as a genuine collaborator designed around human life — its complexity, its costs, and its potential.

This is not a startup. It is not a consultancy. It is a laboratory with a point of view.

Meet the researcher →