Research
Years of research. Tested in the open.
Our products grow out of research into how complex organisations sense change, reason under uncertainty and stay viable. We publish how we test our own systems, including a challenge designed so that Kybern can lose.
Requisite Cognitive Architecture
How perception, memory, reasoning and checking fit together, so intelligence comes from the whole system rather than one giant model.
Explore RCA →
Requisite Agility
How organisations stay viable as their world changes, built from what the frontline experiences and what demand requires, not only from the top.
Explore Requisite Agility →
Reality Challenge leaderboard
259 of 259 internal cases against a naive baseline’s 74, behaviour by behaviour, with the method and its limits published.
See the scores →Complexity & adaptive systems
Many sciences. One question.
Real organisational problems do not respect discipline boundaries. A rising waiting time can be a queueing problem, a network problem and a feedback problem at once. Kybern’s reasoning picks the methods that fit the question, shows why, and combines what each one sees into an option nobody would have found from one angle alone.
- Eleven methods implemented in the prototype, including queueing and flow, network science, feedback and control, Bayesian, causal and resilience analysis.
- A research library of more than sixty scientific frameworks to draw on.
- Every combined option is a hypothesis, tested against evidence before anyone relies on it.
Frontiers
What we are working on next.
Each must beat a simpler alternative before it ships.
A model that questions itself
Finds where its picture is wrong, and knows when to change its mind.
The counterfactual organisation
Causal “what if we changed this?”, proven in one domain first.
A discovery engine
Proposes a new explanation, and the test that could prove it wrong.
Collective intelligence
Helps people think better together: rival views, evidence, one decision.
Federated Kybern
Formations and partners share signals, never raw data.
Kybern domain models
Open and regional models, including Indian ones, adapted per sector and language.
Research blog
From the research to the field.
Why AI near real decisions must know when it is wrong
Confidence is cheap. A system that advises on consequential decisions must say what it does not know and stop where its authority ends.
Read →
Your documents changed last week. Does your AI know?
Most document AI answers from whichever version it finds first. Here is how Change Radar keeps every answer on the latest version, offline.
Read →Inside the Reality Challenge: a test designed so Kybern can lose
259 of 259 internal cases against a naive baseline’s 74. That is not proof, and we explain why.
Read →Research with us.
Bring real cases to the Reality Challenge, co-develop a domain pack, or fund the research.