A different way of thinking about artificial intelligence
Taniwha, Explained
Most AI systems today are built to answer questions, predict outcomes, or optimise a task.
Taniwha is built to explore something else entirely:
What happens when minds have to live in a world they don't control.
An agent can believe something that isn't true. It can misunderstand what happened, trust the wrong person, learn from bad evidence, forget something important, or act confidently on a false belief.
None of that changes reality.
So in Taniwha, the mind, the world, and the rules are separate things. The mind carries beliefs and history. The world determines what actually happened. The rules determine which actions are valid and what consequences follow.
That separation lets us ask a different class of question about AI: not simply can it produce the right answer?, but what happens when a fallible mind has to learn, act, and persist inside a reality that pushes back?
Looking for the product, evidence, and current status? See the Taniwha Engine page →
Taniwha is a cognitive architecture for simulating long-lived agents — entities that form beliefs, develop identity, build trust, refuse information, forget, adapt socially, and sometimes break.
It is not a prompt-response system. It is designed for persistence, inspectability, and accountable interaction with a world whose state does not depend on what the agent believes. Taniwha treats cognition as a process, not an output.
A note on generations. This page describes the enduring architecture. Most of it runs in Taniwha’s current-generation engine today; sections marked Research line were demonstrated in the first-generation research line and are being migrated. The full status picture lives on the Engine generations section.
The mind is not reality
An agent sees only part of the world. What it sees becomes evidence. Evidence can become belief — and belief, right or wrong, is what the agent acts on.
Action goes back into the world, where reality — not belief — determines what actually happens.
Then it compounds. Other agents observe the consequence. They can misunderstand it, lie about it, believe each other, build institutions around it — and change what happens next.
A false belief doesn't move the rock.
But it changes what an agent does near the rock, what it tells others, and what they come to believe — and therefore what happens tomorrow. In Taniwha, that whole loop is real, recorded, and inspectable.
You can watch this loop running live in Copperhollow, a settlement of agents who see, believe, act, and get things wrong in a world that keeps its own record.
The core idea: the graph is the mind
In Taniwha, an agent's mind is represented as a living graph.
- Nodes represent beliefs, motivations, memories, identities, and experiences.
- Edges represent relationships: support, contradiction, influence, trust, resemblance.
Understanding is not stored as isolated facts. It emerges from how strongly ideas connect to one another.
An agent can accumulate many concepts and still understand very little — or hold few ideas that are deeply integrated. What matters is structure, not volume.
The graph is persistent. It changes gradually. And it carries history.
But the graph is the mind — not the world, and not the rulebook. What actually happened is resolved by the world, and shared reasoning rules live outside any one agent's beliefs. An agent cannot make something true by believing it.
How Taniwha agents perceive the world
Each agent operates in cycles of perception followed by periods of consolidation.
When something new happens, the agent does not simply "record" it. Instead, the information is filtered through several layers:
- What the agent already believes
- Who the information is coming from
- How familiar or trusted the source is
- What the agent cares about
- Whether the information fits the agent's identity
In the current generation, that filtering runs on the agent's existing beliefs and its trust in the source. In the first-generation research line it went further, through an identity gate — a cognitive firewall — described next.
Research line · migrating
Refusal, identity, and "trauma without memory"
Sometimes information is too incompatible with an agent's identity to absorb.
When that happens, the identity gate rejects the information entirely. The agent doesn't absorb the concept — it's rejected before it becomes part of what they know.
But something still changes.
The rejection leaves a structural scar on the gate itself. Future encounters with similar information provoke faster resistance, even though the agent cannot explain why.
This is intentional.
Taniwha models a phenomenon common in human cognition: defensive refusal that protects identity without preserving the rejected belief. We sometimes call this trauma without memory.
Learning, forgetting, and consolidation
Taniwha agents do not learn instantly.
Learning happens during consolidation phases, where the agent:
- Commits selected short-term experiences into longer-term structure
- Detects recurring patterns and crystallises them into stable beliefs
- Weakens or removes beliefs that are no longer reinforced
- Rebalances motivation and exploratory pressure
Forgetting is not a failure mode. It is a core feature.
Beliefs that are not used or reinforced gradually decay. Understanding only survives active integration.
Analogy and inference
Agents can form analogies.
They can detect structural similarities between situations, generate speculative bridges, and transfer expectations or rules from one domain to another when alignment is strong enough.
These inferences are imperfect by design — agents can over-generalise and must correct course through experience. This allows them to reason beyond what they've directly encountered, while still being grounded by what actually happens.
Inference in Taniwha is exploratory and abductive, not formally optimal. Agents can generate hypotheses that feel plausible, act on them, and revise over time.
Identity, trust, and social understanding
In the current generation, trust is directional: each agent holds a stance toward each other agent, from distrust to trust, shaped by observed outcomes — and that stance gates what testimony and instruction it admits.
Social life is recorded alongside it: bonds accrue through shared acts, slights land in a single breach, hearsay carries at reduced confidence, apology can discharge a slight, and cultural norms shape how the same act is read by different agents.
Research line · migrating
The first-generation research line went further: identity as a structure of lineage paths influencing what an agent accepts, rejects, and shares; trust that was layered and contextual; and disclosure decisions made before communication — an agent might simply never offer certain knowledge to another. That model is being reconsidered as it migrates.
Research line · migrating
Understanding other agents: portraits
Taniwha includes a built-in theory of mind system.
Agents observe the behaviour of others over time. As evidence accumulates, they form portraits — persistent beliefs about another agent's tendencies, intentions, or reliability.
These portraits influence:
- Interpretation of future actions
- Expectations during interaction
- Willingness to cooperate or resist
Social understanding is not assumed. It is constructed.
The first-generation portrait behaviour has not yet been revalidated in the current generation — but its foundations are being rebuilt around associative recognition, prediction, and surprise: watching what another agent does, and noticing when they act out of character.
Research line · migrating
Planning and action
Agents act through a tactical planner.
Rather than pursuing rigid goal hierarchies, they balance:
- Long-term background pressures
- Short-term tactical concerns
- Immediate operational actions
They can navigate, gather information, react to events, or act directly.
When actions are blocked — by trust, identity, or context — the refusal is defensive. It shapes future behaviour through motivation and trust, not curiosity by default.
Research line · migrating
Time, development, and mortality
Taniwha agents exist in time.
They can be configured with:
- Developmental stages
- Shifting motivations
- Finite lifespans (when mortality is enabled)
Stress, isolation, and repeated defensive responses can shorten an agent's life. Stable relationships and reinforcement can extend it.
When an agent dies, it leaves behind structured traces of what it learned and how it changed.
Ending is treated as a first-class phenomenon.
How Taniwha is used
Taniwha runs in two primary modes:
Persistent worlds
Multi-agent simulations with space, time, and social interaction, where agents form relationships and shared history.
Cognitive test harness
A controlled single-agent environment for probing internal dynamics such as identity gating, trust evolution, consolidation, and forgetting.
Both modes descend from the same research lineage.
A note on which engine. Briarwatch, the long-run validation simulation, runs on Taniwha’s first-generation research engine; Copperhollow runs on the current-generation substrate grounded in Kano. Capabilities and validation results are labelled by engine generation while migration and revalidation continue.
And minds need somewhere to stand. Beliefs in Taniwha are validated against a world that pushes back — increasingly, that world is Kano: deterministic, causal ground truth an agent can observe, remember, and be wrong about. The graph is the mind; the world is the ground.
What could this architecture be useful for?
This is the speculative part of the page. The Taniwha Engine is a general-purpose cognitive engine, and these are the kinds of domains where persistent, belief-driven agents could offer something that stateless AI cannot — prospects, not products, and not validated claims.
Training and simulation
Persistent agents could populate training environments — emergency response, corporate negotiation, medical triage. Agents that remember prior interactions and develop opinions would give trainees evolving scenarios rather than scripted ones.
Long-running digital assistants
An assistant built on this architecture could learn your preferences over weeks and months — not by storing a list of rules, but by forming beliefs about what you value and revising them when they turn out to be wrong.
Interactive storytelling and games
NPCs that could hold grudges, form alliances, and change their minds. Characters in interactive fiction that could remember what you said three chapters ago and act on it. The engine would provide the mind — your game provides the world.
Research and behavioural modelling
Model how beliefs spread through populations, how trust erodes under misinformation, or how institutions change what a community does next. Every cognitive step is logged and auditable — and of everything on this list, this is the use closest to how the engine already runs in our own experiments.
Autonomous monitoring and decision-making
Agents could watch data streams, build beliefs about what is normal, and flag anomalies from evolving understanding rather than static rules — and explain why they flagged something, because the reasoning is inspectable.
Education and adaptive tutoring
A tutoring agent could track where a student is struggling — not from a test score, but from a picture of how the student's beliefs about a subject are evolving — and adapt to what seems to stick and what keeps fading.
What Taniwha is — and is not
Taniwha is:
- A cognitive engine for building agents that develop over time
- A system that keeps mind, world, and rules separate — and studies what happens when they disagree
- A system that treats failure, refusal, and forgetting as meaningful
- A system where identity shapes how agents learn, trust, and decide
Taniwha is not:
- A prompt-response system — agents carry persistent state between every interaction
- A general problem solver
- An optimisation engine
- A claim about human-level intelligence
It is an attempt to take cognition seriously — including its limits.
Want to see the technical detail?
Explore the Taniwha Engine