# O0-PHYS-001: Physical nonseparability and the O/0 inference

Date: 2026-11-11. Retrospective public-data reproduction and mathematical audit.
This is not preregistered discovery. The published result (245 trials, 196 wins,
S about 2.42, p about 0.039) was known before analysis.

## Exact propositions

Q1: Under measurement independence and Bell-local factorization, binary
two-setting models obey |S| <= 2, even when the two sites share a common source.
Prove the deterministic case by exhausting its 16 assignments; extend by
convexity to stochastic mixtures. State all assumptions explicitly.

Q2: Does an independently written parser and analysis reproduce the authors'
Bell-test summary from the original physical event log? Use their fixed
event-ready/quality windows, readout mapping and CHSH sign convention without
optimizing thresholds or outcomes. This is numerical reproduction of an
existing physical experiment, not a new experiment or new discovery.

Q3: What, if anything, follows for O/0? Separate rejection of a Bell-local
model class from universal ontological identity, consciousness, no-signaling,
and the number of fundamental sources. Do not call the result proof of O/0.

## Data

Hensen et al. (Delft), Nature 526, 682-686 (2015), DOI 10.1038/nature15759.
Original repository DOI 10.4121/uuid:6e19e9b2-4a2d-40b5-8dd3-a660bf3c0a31.
The repository API specifies a 152,913-byte data.zip with MD5
342f29f8288c46575818acd2acebd535. Validate the original archive before parsing.
Do not execute the downloaded Python 2 analysis example. Do not redistribute
the source archive under an assumed license; fetch it from its official host.

## Fixed analysis

- Parse all 17 CSV fields; retain the first as timestamp and the rest as integers.
- First/second herald windows: channel-0 start 5,426,350 ps, channel-1 start
  5,425,700 ps; length 52,450 ps; second-window offset 250,000 ps. Inclusive
  start, exclusive end; opposite detector channels required.
- Quality: no excitation click on either side; invalid markers either absent
  (zero) or older than 250 sync pulses. These reproduce the original protocol.
- Readout is +1 if a photon click occurs after 10,620 ns and no later than
  14,320 ns; otherwise -1. Retain all selected trial outcomes.
- Settings a,b are the two recorded random bits. S=E00+E01+E10-E11.
- Win if x*y=(-1)^(a*b). Count all selected trials; no outcome-dependent cuts.
- Primary null tail adopts the original calibration tau=2*5.4e-6 and
  q=3/4+3*(tau+tau^2). Report binomial survival P[W>=k | n,q]. Also report
  the ideal-uniform-settings tail, clearly labeled. Recompute with a separate
  high-precision standard-library binomial sum to check the SciPy result.
- Enumerate all local deterministic models. Validate the quantum ideal
  using explicit Pauli operators and a Bell state; distinguish this theoretical
  calibration from measured data. Report reduced-state purities as an exact
  example of global information absent from local marginals.
- Secondary negative control: 2,000 seeded permutations of Bob's outcome
  within each of Bob's setting strata, preserving local setting-specific
  marginals while breaking paired intersite correlation. Seed 11115000.
- Report the 4x4 setting/outcome count table, settings counts, correlations,
  descriptive S standard error, both tail probabilities, source hashes and
  all checks. Do not convert the tail into probability that O/0 is true.

## Interpretation rule

If the original counts and rounded S/p cannot be reproduced, flag the mismatch
and stop before making a physical-evidence claim. If they match, report
REPRODUCED_BELL_VIOLATION under the original experimental assumptions, with
explicit credit to the original investigators. The analysis does not re-check
spacetime timing, detector hardware, the RNG calibration or laboratory records
outside the deposited event log.

The theorem is an exact conditional statement. The physical inference is
statistical. The extension to the full O/0 ontology remains unestablished.
