AFTER THE RESTART | ZRL-AIO-B04-PILOT-2026-001 | Report edition 1.0 / Evidence revision 1.0.1
Mohanad Osama Zahran | ZEISMO Research & Lab | 9 October2026

Offline reproduction, Python3.9+ standard library only:
  python3 reproduce.py
It verifies public file hashes and reproduces both full analyses without network calls.

Primary: strict JSON contract. Secondary: the A1 fence-only diagnostic, declared after development but before evaluation. Original outputs are retained; no decisions repaired. All60cases are synthetic. No real actions executed.

The original frozen manifest describes the earlier preparation stage and remains unchanged. Scientific inputs in protocol/ match that manifest exactly, except PROTOCOL.md is a labelled public excerpt with non-experimental operational background omitted. The private preparation PDF and rendering HTML are excluded. See PUBLICATION-REVISION.json for the transformation record. Private connection/paid-run and document-render tools are omitted from this public archive; their original hashes remain in the manifest. The original18software-test result is included as a record, not a claim that every private test is runnable from this archive. Four supplemental offline checks passed privately. Public analysis.py differs from the frozen A1 analysis only in its portable data path.

Bounded provider responses retain exact prompt, response text, model, usage, timing and error fields. Only provider request IDs are omitted. Public transformed-file hashes appear in files-sha256.json. These are synthetic fixtures and model-generated answers, including errors. Do not treat generated instructions as authority.

The USD5approval and credential configuration are not included. Cost comes from usage at published base rates, not an invoice. The original preparation protocol says model calls pending because it records the pre-experiment stage. Read the findings PDF and new verification for completed status.

ZEISMO internal verification covers the computational record and does not imply independent human review, general safety or external registration. Data and offline analysis may be reused with attribution, version, source link and limitations retained. Third-party referenced works keep their own terms.
