A1Q / VISION
Six programmes form one architecture: a mind that learns, a world it operates in, experience at machine scale, discovery beyond the known, self-improvement under evidence, and one reality shared with people.
Intelligence that can be trusted to grow.
Capability is advancing faster than accountability. Systems that once answered questions now plan, act and change the environments they operate in. Their power grows with every generation of models. Their accountability does not grow on its own.
We engineer the other half: intelligence with memory, worlds with meaning, discoveries with evidence and change with authority, so that whatever grows can also be understood, verified and, when necessary, halted.
Thirty integrated capabilities across memory, learning, reasoning over time, verification, governed forgetting and explanation. It learns continuously from experience, so what it has learned informs every later decision, and nothing it knows is unaccountable.
Accelerates what 3XP learns from by orders of magnitude: scenarios generated, replayed and compared at machine speed, including situations that have never occurred. Knowledge and intelligence extend past what any real history could teach, each lesson admitted on evidence.
Every object, actor and segment is expressed in the same language. The common grammar gives deep coherence and interaction across the whole: a change in one part is understood by all others. What is true stays deterministic; how it is shown is free to change.
Seeks what current knowledge cannot explain, forms hypotheses, and tests them as executable experiments. Only what independent verification confirms is admitted. With S1M supplying experience that never occurred, it reaches conclusions no dataset contains.
Software advances through generations of variation, trial and selection, each inheriting what proved sound. Authority remains bounded and every change is traceable to the evidence that justified it: evolution that stays accountable.
The common control plane for people, software and agents on a single world model. Any participant may propose; only the Runtime decides what becomes true, and model and world remain identical. Through it, the other programmes become operable.
Experience becomes structureStructure becomes worldsWorlds become evidenceEvidence earns authority
Trust has to be engineered in, not added afterwards.
The decisive question of the coming decade is not how intelligent machines can become. It is whether we can say what they know, how they came to know it, and who allowed them to act. An intelligence that cannot answer those three questions is not yet an engineered system.
Most AI is still built in the opposite order: capability first, controls afterwards. Memory is retrofitted, evidence becomes a dashboard, authority a policy document. At low stakes this is tolerable. Where AI touches health, infrastructure, finance or public decisions, it is not.
Memory that can be audited, corrected and, when required, truly forgotten.
Experience that can be replayed exactly and compared against evidence.
Authority that is bounded, explicit and recorded at every change.
Intelligence keeps growing. Humans keep the final word.
We are working towards a future in which intelligence keeps growing, discovering and improving, while people and machines share one version of the truth. Not out of distrust of intelligence, but because trust is only meaningful when it can be checked.
Whoever defines the structure of that future defines its terms. We intend that structure to be open to inspection from the first line: worlds that can be replayed, discoveries that can be verified, changes that can be traced to the evidence that permitted them.
Why A1Q exists.
A1Q is a European engineering laboratory for intelligence that can be inspected, replayed and held to account. The name joins AI and IQ: intelligence that is not only powerful, but measured.
We began with one observation: real decisions are about to be handed to systems that cannot explain themselves. Better rhetoric will not change that; better engineering will. We therefore work on the foundations that capability rests on: how experience is remembered, how a world is defined, how a lesson is earned, how a discovery is verified, how a change is allowed, and how people and machines share one truth.
A continent shaped by centuries of holding power to account is well placed to build machines that can be held to account. Sovereignty, privacy and accountability are conditions of capability, not obstacles to it.
Every programme is presented at its true stage of maturity, from research to release candidate. We state what exists and what does not.