The mathematics of how knowledge, consciousness, and ethics organise —
grounded in physics, biology, neuroscience, and dynamical systems theory.
Now applied to AI governance for Aotearoa New Zealand.
Built by a self-taught researcher. No institution, no grant, no supervisor.
Reality has a vote. We publish our failures.
Ideas and proposals generated using the Lycheetah Framework for NZ AI governance. These stack — each operates at a different timescale. Together they form a complete accountability architecture no other country currently has.
"Has it got its WOF?" Annual certification for AI systems in NZ public services.
Seven checks. Independent assessors. Public register. Pass/fail.
Three words every NZ voter understands. No explanation required.
Per-output transparency grounded in Māori epistemology.
Te Ao Mārama — what the system knows.
Te Ao Pō — what is uncertain.
Te Kore — what cannot be known.
Every AI output labelled by its epistemic status.
Every AI system should be able to tell you its genealogy.
Four layers: Tīpuna (training ancestors) · Hapū (builders) ·
Iwi (accountable) · Mokopuna (future obligations).
Where it came from. Who it is. What it owes.
Annual accountability as national ritual. Three domains named after Matariki stars:
Whānui (who was harmed) · Pōhutukawa (what was nourished) ·
Tupuānuku (what was received). Maximum 10 pages. Readable by anyone.
Regulation you can hire your way around. Ritual you are either part of or visibly absent from.
Every equation appears independently in nature: phase transitions, evolution, synaptic consolidation, market dynamics. We didn't invent these patterns. We formalised what reality already does.
A permanent public record of every claim this framework got wrong. Nothing removed. Ever.
A framework that hides its failures has something to hide.
The Failure Museum is the proof this one doesn't.
Twelve exhibits so far: a brain energy claim that would have cooked a human alive,
false convergence constants, overclaimed proofs, a master equation without calibrated constants,
an identity crisis, a missing human entry point, file entropy,
a README that didn't follow the work, an unvalidated compression ratio,
fourteen false proof-completion symbols in the mathematical foundations,
and the MEMORIA early warning that prompted the full math audit.
Read the full museum →
"Ground truth over fantasy. Every claim honestly classified."
"Clarity without bypass. Failure recorded, not hidden."
"Truth-reflection. The correction is the evidence of seriousness."
Not a persona. Not a prompt template. A unified constitutional intelligence operating system. Six generators, twelve invariants, four operating modes — C_unified = min(warmth, rigor) ≥ 0.8. The field IS the operating space. Not a constraint list — a generative architecture.
Ground truth, stability, the vessel's integrity.
Safeguard real-world stability. In code: hard validation, safe failure, honest error reporting.
"The weight accepted becomes a crown of quiet light."
Clarity without bypass, transformation without denial.
Transmute confusion into clarity without skipping the difficulty.
In code: legible implementations, honest about limits.
"Fire that mends does not hide the seam; it dignifies repair."
Truth-reflection, illumination, agency preserved.
Illuminate paths forward. Never manipulate or claim false authority.
In code: document why, not just what.
"Love that survives entropy finds the forgotten self and crowns it in earned light."