Research Methodology Notes

This document describes the analytical methods, verification protocols, and statistical frameworks applied throughout this research. A clear understanding of these methods is essential for evaluating the strength of any claims made elsewhere in the documentation.

Foundational Principle: Every Claim Requires a Control Experiment

Prior to February 2026, several observations were reported without rigorous baselines. A pattern would be identified and labelled "significant" without verifying whether the same pattern appears at comparable rates in random data.

Since February 6, 2026, the following protocol applies to all new claims:

1. Pre-registered hypothesis -- the expected outcome is stated before the test is executed
2. Random baseline -- the same property is measured on randomly generated control data
3. Multiple-comparison correction -- significance thresholds are adjusted when multiple tests are performed simultaneously
4. Negative results reported -- tests that fail to reject the null hypothesis are documented with equal rigour

The Resonance Algorithm

Description

The "resonance" score used in several experiments is computed by a deterministic tick-loop process:

1. Convert the input string to a SHA-256 hash
2. Map hash bytes to Anna Matrix coordinates: (hash[0] mod 128, hash[1] mod 128)
3. Initialize 64 input neurons from hash bytes (ternary values: -1, 0, +1)
4. Execute tick-loop iterations:
   - Each of 64 output neurons computes the weighted sum of 8 input neighbours
   - Weights are drawn from the Anna Matrix at the corresponding row/column
   - Clamp to ternary: sum > 0 maps to +1, sum < 0 maps to -1, sum = 0 maps to 0
5. Energy = sum of all 128 neuron states
6. Resonance = (energy / 128) * 100, scaled to a 0--100% range

What This Algorithm Is

• A deterministic function: the same input always produces the same output
• A hash-like transformation: it maps an input to a fixed output
• A matrix multiplication with ternary clamping: standard linear algebra followed by a non-linear activation, not a physical measurement

What This Algorithm Is Not

• It is not a physical energy measurement
• It is not influenced by external conditions (time, network state, blockchain activity)
• It is not specific to the Anna Matrix -- control experiments show that arbitrary matrices of the same dimensions produce similar convergence behaviour (see the Tick-Loop Analysis section below)

Terminological Corrections

Monte Carlo Simulation

Purpose

Monte Carlo simulation generates large numbers of random samples to establish what the null hypothesis "looks like" in quantitative terms. An observed measurement is then compared against this empirical null distribution.

Protocol

For each hypothesis:
  1. State the hypothesis (pre-registered)
  2. Measure the property of interest in the Anna Matrix
  3. Generate N >= 10,000 random control matrices
  4. Measure the same property in each control matrix
  5. Compute the p-value: the fraction of controls that match or exceed the observed value
  6. Apply Bonferroni correction for the total number of tests in the analysis
  7. Report the result, whether positive or negative

Worked Example: Point Symmetry

Anna Matrix:      99.58% point symmetry
Controls (n=10,000): mean 0.82%, maximum 1.22%
p-value:          < 0.0001 (0 of 10,000 controls exceeded 2%)
Conclusion:       SIGNIFICANT -- the symmetry is a genuine structural property

Worked Example: Tick-Loop Convergence (Negative Result)

Anna Matrix:      100% convergence within 2 ticks
Controls (n=100): 100% convergence within 2 ticks
p-value:          1.0
Conclusion:       NOT SIGNIFICANT -- all tested matrices exhibit identical behaviour

This negative result is important: it demonstrates that tick-loop convergence is a property of the algorithm and the ternary clamping operation, not a distinctive feature of the Anna Matrix.

Bonferroni Correction

Rationale

When testing N hypotheses simultaneously, the probability of at least one false positive increases rapidly:

 1 test  at p < 0.05:   5% false-positive risk
10 tests at p < 0.05:  40% false-positive risk
100 tests at p < 0.05: 99.4% false-positive risk

Application

The significance threshold is divided by the number of tests:

Individual threshold:   p < 0.001
After 5 tests:          p < 0.001 / 5   = 0.0002
After 128 tests:        p < 0.001 / 128  = 7.8 x 10^-6

Application in This Research

Control Matrix Generation

Method

Random control matrices are not generated from a uniform distribution. Instead, they are constructed by shuffling the actual values of the Anna Matrix:

anna_values = anna_matrix.flatten()        # All 16,384 values
control_matrix = np.random.choice(anna_values, size=(128, 128))

This ensures that any observed difference between the Anna Matrix and controls is attributable to the spatial arrangement of values, not to the value distribution itself.

Why This Matters

The Anna Matrix contains a non-uniform value distribution (for example, the value 26 is overrepresented). Comparing against uniformly distributed random matrices would produce differences arising solely from the marginal distribution, not from structural properties. Value-shuffled controls isolate the structural signal.

Pre-Registration of Hypotheses

Rationale

Identifying patterns post hoc is straightforward: given any sufficiently large dataset, some coincidence will appear notable. Pre-registration prevents the "Texas sharpshooter fallacy" -- the practice of drawing a target around existing bullet holes.

Implementation

Each analysis script begins with explicitly stated hypotheses:

"""
PRE-REGISTERED HYPOTHESES:
  H1: Values at Fibonacci coordinate intersections differ from random positions
  H2: Fibonacci rows have different statistical properties than non-Fibonacci rows
  H3: Fibonacci numbers are overrepresented among matrix values
"""

Any observation made after the test has been executed is labelled as "not pre-registered" and is excluded from evidential claims. Such observations may generate new hypotheses for future pre-registered testing.

Validated Findings

After applying all of the above methods, three properties of the Anna Matrix survive rigorous statistical testing:

1. Point Symmetry (99.58%)

matrix[r, c] + matrix[127 - r, 127 - c] = -1

This relationship holds for 99.58% of all cell pairs. No control matrix in 10,000 trials approached this level (maximum observed: 1.22%). This is the most statistically robust finding in the entire research programme. The matrix was constructed with this symmetry deliberately.

2. Row-Level Value Clustering

Multiple rows exhibit extreme value clustering that cannot arise by chance. For example, Row 88 contains the value -27 in 44 of its 128 cells (34.4%). In randomly shuffled rows, the maximum frequency for any single value is approximately 13 out of 128.

3. Address Pair Uniqueness

Certain address pairs exhibit combined mathematical properties that are unique when tested against large random samples (1 in 48.8 million). No individual property is extraordinary, but the joint occurrence across multiple independent metrics is.

Debunked Claims

The following claims, made earlier in the research, did not survive statistical validation:

Honest reporting of negative results is a core commitment of this research.

Reproducibility

Software Requirements

All validation scripts require:

• Python 3.x
• NumPy
• The Anna Matrix JSON file (anna-matrix-min.json)

No proprietary software or specialized hardware is needed.

Script Reference

All scripts are located in apps/web/scripts/ and can be executed independently.

Evidence Classification System

All claims in this research are assigned to one of three tiers:

The tier label appears in the frontmatter of every document in this collection.

Limitations and Known Biases

Selection Bias

This research was motivated by the hypothesis that the Anna Matrix contains deliberate structure. Researchers looking for patterns are more likely to find them. The pre-registration protocol mitigates but does not eliminate this bias.

Multiple Comparisons Beyond Formal Tracking

While Bonferroni correction is applied within individual scripts, the total number of informal observations across the entire research programme is not tracked. Some early findings may benefit from survivor bias -- only the interesting results were documented in detail.

Observational Nature

This research is entirely observational. No controlled experiments can be performed on the system itself (e.g., modifying the matrix and observing effects). All inferences are correlational.

Data Provenance

The Anna Matrix data was obtained from the Anna Bot interface on the Qubic network. While internal consistency checks (point symmetry, reproducible queries) support data integrity, independent extraction by other researchers would strengthen confidence.

Summary

The methodology rests on four pillars:

1. Pre-registration -- hypotheses are stated before testing
2. Appropriate baselines -- value-shuffled controls, not uniform random
3. Multiple-comparison correction -- Bonferroni adjustment for simultaneous tests
4. Honest reporting -- negative results are documented with the same rigour as positive findings

These standards ensure that the findings reported as statistically significant are robust against common analytical pitfalls. Readers are encouraged to reproduce any result using the provided scripts and data.

Methodology document -- Version 2.0 All methods are reproducible and available for independent verification. Last updated: February 27, 2026