METHODS · KEPT ON RECORD
THE BOOK OF CARDS
CHAPTER ONE · THE HOUSE RULE
A method you cannot repeat is a rumour.
the back of each card carries the rest
PREFACE
The cookbook of the research arms. Each recipe states its method, its inputs, its expected return, its risk, its audit trail. A claim that cannot survive the card is not published.
THE SHELF →In an internal 200-question test drawn from six public AI benchmarks, our Hecate method raised Gemini 4 Argon's score from 58% to 85% on the same questions. Internal and not independently audited; some questions were reformatted, so these figures are not comparable to official leaderboards.
In a follow-up open-ended math test (HMMT, BRUMO and CMIMC 2025), the Hecate method scored 84% versus 72% for Gemini 4 Argon answering without code tools. Internal, not independently audited.
THE CARD
THE FIRST SHELF
DRAWER I · FOUNDATION PAPERS
Fifty methods in real estate. High return, low risk.
One hundred methods to lift people out of poverty.
The flagship output of the network.
Research from the suicide reduction arm.
Poverty as a mispricing problem. The poverty-reduction arm's research record.
The variance thesis: neurodivergence as variance, not superiority - structured.
An evidentiary instrument built to survive its own cross-examination.
Filed in order. Press RE-ORDER to change the sequence.
DRAWER II · CONSCIOUSNESS & SINGULARITY
Reply to Bodea (2026): consciousness as a process, not a number. Under submission, Cognitive Processing.
The exhaustive list of known consciousness. 555 rows, one formula, one challenge.
Prologue, threshold, epilogue, applied augmentation. The whole silicon shelf.
Attribution economics: what the public pays for machine minds. In preparation.
First reading of MoM-BH*-1, priced against the primary literature.
Formation, accretion, signature, cosmology. The long unpacking.
Seven held, six fell. One object audited two decades of theory.
Strategic architecture and the anti-mainstream voice layer.
The paper that is not quite a paper. Read it and decide.
100.00% exact genomic interval agreement (UCSC BED <-> SO GFF3 v1.26) across 3,522 evaluated LLM episodes & 5,050 round-trips (+11.11 pp ITT lift on Gemini 3.8 Flash, p=1.62e-27; -98.4% to -99.4% output tokens; MIT & CC BY 4.0).
Updated 2 Nov 2026. Eighteen papers & sealed benchmarks on record. Reproducible, or it did not happen.
Fourteen public preprints & sealed benchmarks, freely available.
SUBSCRIBE
New cards as they are audited. Nothing else.
THE LIST →