USL BIBLE ATLAS — RECEIVER RESULTS
Prepared 8 September 2026
A Way of Reading, Carried in Colour: The Bible Atlas Experiment

RESULT
One ChatGPT Sol receiving trial recovered the exact reviewed atlas, interpreted
its connected reading of the retained 66-book KJV collection, and challenged
its mostly single-frame treatment of competing interpretations. Local replay
confirms the returned artifacts and code results.

METHODS
Michael confirms that this run used an existing conversation with earlier context. The available record does not establish whether this exact sender interpretation had been seen beforehand. We do not describe the run as blinded: it tests recovery and use of this artifact in that contextual setting.

The model label is the organizer's report; an exact backend version was not
supplied. No Claude receiving trial is included. Sol used supplied reference
code and a separately written Python cross-check. Complete Carrier schema and
allocation validation still relied on the supplied validators. This was not
a wholly independent implementation of all layers. Local rerunning of code
does not constitute a second AI interpretation. Our qualitative comparison
with the sender rubric had access to both sides.

FROZEN ARTIFACTS
PNG: 295,322 bytes, 344 x 344 pixels, 288.40 KiB
SHA-256: 33e04bea3e3f10e7169bfcc70aaa214c9a4cfd5422b297de92a8395d7a8ccf02

USL1: 978,124 bytes
SHA-256: c47b934954a45db90cd88f5b3f1d5cc8a30dc0d4d96d30af9811e0877a9a9a17

Exact expanded expression: 1,056,076 UTF-8 bytes
SHA-256: e4d2e31c3c2a1169896a2e2012880727afd8be334fb47a4e278cda5ac7ab65a3

Original returned ZIP: 2,403,136 bytes
SHA-256: 930e322e485c4133136e4d4af4be82cee69d6d14b733a845e96184f2aa3a48cd

RECOVERY
All 12 inputs match the frozen edition 90 sender kit. All 38 entries in the
returned checksum manifest match its 39 files, excluding the manifest itself.
Four replay commands passed without modifying returned code; 33 tests passed.
All five recovery outputs and the full decoded message reproduce byte for
byte. Diagnostic changes are limited to Node version and Windows newlines.

Data grid: 320 x 320, one pixel per module; 60 unused black modules.
Physical pixels checked: 118,336.
RS: 1,204 codewords x 32 = 38,528 zero syndromes; no repair needed.
Application gzip: 255,201 bytes, expanding to 978,124-byte USL1.
Tokens: 133,225 = 66,423 lexical + 66,613 literal/space + 189 concept.
Concepts: 182 distinct checked references, 189 actual bindings, zero unresolved.
Structure: 5 stance principles, 24 reading methods, 14 thematic facets,
5 comparisons, 66 book frames.
State: 446 checked, 3 working, 458 allocation records, 11 dropped-history records.
Checked head: d436ebafcb6626cf892bcebc417ce801eb7ea00bb665485431dbb8753324fb44

WARNING PRESERVED
LEGACY_ID_HISTORY records earlier reuse of working IDs w3 through w9. Current
selected checked-head/ID bindings are unaffected. The history was not rewritten.

SEMANTIC FINDING
The recipient recovered a proposed reading which connects the collection
without making every book speak in one voice. It preserves genre, attributed
claims, difficult comparisons, qualifications and permission to disagree.
One highlighted comparison preserves common standing in Christ (checked/351)
beside obedience within enslavement (checked/354), without automatic erasure,
equivalence or contradiction.

The recipient's substantive challenge is retained: allowing disagreement is
not yet a worked representation of competing full frames. The repeated lens
of care alongside coercion is an editorial choice. A future experiment could
encode an alternative frame with its supporting reasons. It was not run here.

SOURCE FOLLOW-UP — AFTER THE RECEIVER TRIAL
The original receiver lacked the complete Bible TXT and could check only the
internal consistency of its carried evidence. The publication adds a later
local comparison against the retained full source, without changing the kit.
All 602 quotation occurrences / 522 distinct verses match source identities,
byte boundaries and whitespace-normalized wording. All 31,102 source-index
slices, 177 evidence mappings, 14 source-book lists and 66 book frames pass.
Eight deliberate source-audit corruptions are rejected.
Full TXT: 4,455,950 bytes
SHA-256: 0204adaed1f25700aa854218cae63c7172228c41088f335e99167a071eed83c0

SCOPE
The PNG carries a composed overview, full frozen Carrier state and selected
evidence. English lexical forms and reference tools are separate inputs.
It cannot reconstruct every Bible verse. It is not a smallest-Bible record,
optical/damage robustness test, estimate of population reliability, independent
authentication of the reviewer, or proof of theological or historical truth.
Matching bytes and a faithful reading do not require agreement with the author.

REPRODUCE THE RETURN
Extract usl-bible-atlas-chatgpt-sol-return.zip. From its usl_bible_atlas_recovery folder:

    cd receiver
    node --require ../offline.cjs recover.cjs
    cd ..
    node --require ./offline.cjs inspect_reference.cjs
    python -X utf8 audit_atlas.py
    python -X utf8 test_audit.py

Use a new extraction; these commands regenerate local results. Node and Python
are required; no third-party packages, browsing or agents are needed. Compare
the packet and expression with the hashes above. The research archive contains
the frozen sender key, full-source follow-up, local reports and reproducible
audit additions. It is not material for a blinded receiving exercise.
