Semantic Switch Proof Pack: Verified Relation Preservation

Synthetic fixtures, portable verification harnesses, raw results, and a reproducible 20/20 + 23/23 evidence trail.

By Blayne Aina · October 2026 · AINA Governance Specification

Executive reading guide · 60 seconds

Problem: A protected-computation claim should be reproducible, not accepted from a polished article alone.
What it proves: The pack contains synthetic A/B fixtures, verification harnesses, and raw results for 20/20 invariance blocks and 23/23 exact relations.
Commercial relevance: Run the included verifier with your Semantic Switch binary and compare its result with the published evidence.
Next step: Review the measurements and claim ceiling, then bring one bounded entity-resolution problem.

A public, reproducible evidence bundle for relation-isomorphic protected computation. The fixtures are synthetic; no customer records, private crosswalk, credentials, native identifiers, local keys, or provider calls are included.

Read the reproduction guide Download verify.py

Verified results

Verification surface	Published result
Relabeling-invariance blocks	20/20
Eligible native edit relations preserved exactly	23/23
Records verified per relation-isomorphism run	53/53
Field values verified per run	529/529
Batch-restoration adversarial cases	14/14
Transformation adversarial cases	10/10
Provider calls during preparation	0

The pipeline being tested

Native synthetic records
→ Semantic Switch protected representation
→ qBlock / Qdrant Edge candidate partitions
→ Fracture adjudication
→ Forge canonical selections
→ local exact restoration through the private crosswalk
→ Shield forward-only baseline defense

The two input files describe the same relational object under different native relabelings. A correct generator must preserve the relational structure, not the native token identities. The relation audit separately requires every eligible native edit type to remain visible in the fictional representation.

Reproduce the evidence

Obtain or build a Semantic Switch executable with the relation-isomorphism-v1 feature.
Download verify.py and this pack's fixtures and harnesses.
Run python verify.py --binary /path/to/semantic-switch.
Compare the fresh JSON summary with the retained invariance result and exact-relation result.

The driver creates a temporary random local key, prepares both inputs, verifies every record and field, computes fresh reports, and deletes the temporary workspace. It exits successfully only when all 20 blocks and all 23 eligible relations pass.

Artifact inventory

Claim ceiling

These results establish deterministic transport fidelity, exact relation preservation, complete local verification, and adversarial restoration behavior on the declared synthetic fixtures. They do not establish anonymity, inference resistance, universal privacy, matching accuracy, production merge correctness, or independent validation of a customer population. Native rows, keys, and the private crosswalk remain under client control.