Applied AI research lab · Aotearoa New Zealand

The model can be wrong. The world keeps the record.

Taniwha is an applied AI research lab. We investigate how AI can remain useful inside systems where truth, authority and consequence exist independently of the model.

Kano world · still frame building the live world in your browser…

The latest result

MEASURED 27 Aug 2026 · Ground · Mind · Assurance

Where reliability lives

Cognitive and institutional reliability were experimentally separable. Five pre-declared properties held across every tested trajectory while the cognition was ablated, killed, reset and wholesale-substituted.

What this is not

One world, one constitution, one comparator model. It does not show cognition is unimportant — and the pre-registered cell where the substituted model panel beat our own architecture is published at full strength.

How it works

Three planes. One thesis.

We investigate how artificial intelligence can remain useful inside systems where truth, authority and consequence exist independently of the model.

How the three planes fit togetherA mind proposes an act. Rules permit or refuse it before it executes. The world resolves what happened and keeps the record. The record returns to the mind as evidence. Refused acts return to the mind as a typed refusal without ever reaching the world.MindAssuranceGroundbelieves, and can be wrongpermits or refuseskeeps the recordEngineĀrai · KeteKano · Ponoproposes an actpermittedrefused — typed, before executionthe record — what actually happened, not what the model claimed
Three planes, four paths. The model proposes and acts; it does not own permission, and it does not own the record.

Ground

EVIDENCED

A world the model cannot overwrite.

In the programme this is epistemic containment.

Agents that believe, trust — and can be wrong.

In the programme this is architectural composition.

Assurance

EVIDENCED

Rules that fire whether the model remembered them or not.

In the programme this is consequential containment.

The full programme, methods and results →

Apparatus · Copperhollow / Hīkoi

Walk the ground. Watch the settlement.

Hīkoi is our first-person viewer for Kano worlds — and its window on Copperhollow, the live settlement where thirty-seven minds work, wander, and get refused on an authoritative world. The apparatus runs in public; nothing here is a mock-up.

Hīkoi — a lakeside stand of trees at midday, every tree grown by the world
Still · Hīkoi alpha · rendered from seed
Hīkoi — looking across clear water to generated islands, shoreline and vegetation from the same seed
Still · Hīkoi alpha · same seed, same world
Download · runs offline

One zip: a generated planet, a first-person walk, and an MCP server your own AI can interrogate — no account, no network.

Get Hīkoi on GitHub → Explore Kano →

What ships today

Two you can run now. One that isn’t ready.

All three came out of the grounding and enforcement experiments — transferred research, not a separate product line.

Ārai

Open-source enforcement for coding-agent rules

Your agent read the rules and did it anyway. Ārai enforces them before execution, whether the model remembered or not.

Ships now — open source

Hīkoi

A first-person viewer and window into Kano worlds

An offline planet and an MCP server. Walk the world yourself, then let your AI query the same ground.

Download now

Kete

Organisation-level policy enforcement and audit

Ārai across a whole organisation, with an audit trail that survives the session and Pono as the ground under it.

Closed alpha — not taking signups

Everything else — research systems, instruments and demonstrations — is on the status map, honestly labelled.

The lab

Three people in Aotearoa New Zealand.

Independent, with no outside investors and no reason to overstate a result. Questions get pre-registered before the run; findings get published with the misses included.

The name is not metaphorical — a taniwha is shaped by the territory it inhabits, and so is a mind. The full account of the name →

Who