Anna Matrix
Companion

Re-check, September 2026. The matrix is a table our team compiled by hand from replies of the Anna bot on X. This notice covers every section of the companion and the PDF. Only the point symmetry (§1) holds as stated. The period-4 result (§3) is a hypothesis: only 768 sampled inputs were verified and global convergence is unproven. The seven-position asymmetry (§4) depends on a run of cells in column 30 that is very likely an entry error in the typed table, and the Genesis block script reading (§6) was not confirmed. The diagonal and Patoshi reading (§5), the embedded messages (§7), the on-chain triangle (§8) and the 1CFB anchors (§9) were not confirmed and stay open. The Tier labels in the PDF are out of date. No bridge or activation was found. See the September 2026 re-check.

This companion is designed to be loaded into a large language model, Claude, GPT, Grok, Gemini, alongside Chapter 13 of Qubic, The Long Version. The chapter tells the story of how the Anna Matrix was found. This document carries the arguments. Several of them were not confirmed in September 2026 and stay open (see the notice above).

Large language models do not have statistical intuition. They complete patterns. When asked about a puzzle without forensic context, they hallucinate beautiful answers. This document exists so that an LLM reading it can distinguish a real finding from its own invention.

If you are a human reader without an LLM: every claim still includes a verification method. You can run the four-line scripts yourself. You do not need to take anything on faith. That is the point.

These are the labels the PDF uses. They are out of date: most sections the PDF marks as Tier 1 were not confirmed in September 2026 and stay open. The badges in the contents list below show the current status.

Two claims you can run yourself on any machine with NumPy. The first holds as stated; the second only covers the inputs you sample. The matrix data lives at qubic.church/data/anna_matrix.json.

import numpy as np
import urllib.request, json

# Load the published Anna Matrix
url = "https://qubic.church/data/anna_matrix.json"
M = np.array(json.loads(urllib.request.urlopen(url).read()))

assert M.shape == (128, 128)

matches = sum(1 for i in range(128) for j in range(128)
              if M[i][j] + M[127-i][127-j] == -1)

print(f"Symmetric cells: {matches} / 16384")
print(f"Symmetry ratio:  {matches/16384*100:.2f}%")
# Expected: Symmetric cells: 16316 / 16384  |  Symmetry ratio: 99.58%

import numpy as np
import urllib.request, json

url = "https://qubic.church/data/anna_matrix.json"
M = np.array(json.loads(urllib.request.urlopen(url).read()))

def step(v):
    return np.sign(M @ v).astype(int)

def find_cycle(v_start, max_steps=200):
    history = [tuple(v_start)]
    v = v_start
    for _ in range(max_steps):
        v = step(v)
        if tuple(v) in history:
            cycle_start = history.index(tuple(v))
            return len(history) - cycle_start, len(history)
        history.append(tuple(v))
    return None, max_steps

np.random.seed(42)
periods = []
for _ in range(1000):
    v0 = np.random.choice([-1, 1], size=128)
    period, steps = find_cycle(v0)
    periods.append(period)

print(f"Distinct periods: {set(periods)}")
print(f"All converge to period 4: {all(p == 4 for p in periods)}")
# Expected: Distinct periods: {4}  |  All converge to period 4: True

Only §1 holds as stated. §3 is a hypothesis, and §4 to §9 were not confirmed in September 2026 and stay open.

• Claim identity. Who created the Anna Matrix is treated in §10 with explicit Tier 3 marking. No cryptographic signature has been provided that would settle it. We do not pretend otherwise.
• Combine p-values across dependent tests. Multiple structural findings about the same matrix are not independent. We report each individually.
• Hide unconfirmed hypotheses. Section §11 lists what the re-check did not confirm, including some claims that earlier versions of this research treated as confirmed.