Skip to content

feat(reference): Multiverseal Twin substrate reference impl [E13/WS-C] - #281

Merged
mdheller merged 2 commits into
mainfrom
feat/multiverseal-twin-refimpl
Aug 4, 2026
Merged

feat(reference): Multiverseal Twin substrate reference impl [E13/WS-C]#281
mdheller merged 2 commits into
mainfrom
feat/multiverseal-twin-refimpl

Conversation

@mdheller

@mdheller mdheller commented Aug 4, 2026

Copy link
Copy Markdown
Contributor

Multiverseal Twin — substrate reference implementation [E13 / WS-C]

Executable substrate for schemas/{MultiversealTwin,TwinAttestation,InterferometricDiff}.json (ADR-0014). Turns the twin contracts into a running, tested engine on FHRR (Fourier Holographic Reduced Representations): V = C^D unit-modulus phasors, so phase = provenance is literal.

bind(o,r)=o⊙r   bundle(ts)=Σt_k   unbind(H,r)=H⊙conj(r)

Proven, with measured numbers (python3 test_mvtwin.py, D=8192, N=8)

Property Measured
Reference-gating (authored ref reconstructs; else noise) recover 0.348 vs wrong-ref 0.003 / opaque −0.009 (~100× SNR)
ε-unlinkability (distinct refs near-orthogonal) cross-context leakage L=0.004
Fringe = leading indicator (drift shows in phase before score) score move 0.00016 vs mean|fringe| 0.020 (~125×)
Holographic tamper-evidence (local write → global fringe) 100% of spectrum perturbed

Scope & honesty

Linear substrate only. Reference-at-ingest, ε-budget, VRF mint/verify, and the impersonation wall are enforced at the contract/policy layer (ADR-0014); Sybil-resistance / nonlinear trust policy live in a separate layer that must not leak back into this medium (twin spec §H / WS-F). Lands under reference/ (adjacent to the contracts it exercises), so it doesn't touch schema CI. Next increments: VRF-derived references, QEC [[n,k,d]] sharing, and the InterferometricDiff replay (Fresnel forward-propagation).

🤖 Generated with Claude Code

Executable FHRR substrate for the twin contracts (ADR-0014). V=C^D unit phasors, so
phase=provenance is literal. bind/bundle/unbind + interferometric diff + unlinkability
leakage. Self-test proves, with measured numbers, the four load-bearing properties:
- reference-gating: recover 0.348 vs wrong-ref 0.003 / opaque -0.009 (~100x SNR)
- ε-unlinkability: cross-context leakage L=0.004 (near-orthogonal)
- fringe is the leading indicator: score move 0.00016 vs mean|fringe| 0.020 (~125x)
- holographic tamper-evidence: one local write perturbs 100% of the spectrum
Linear substrate ONLY; VRF/QEC/impersonation-wall at the contract layer, Sybil/nonlinear
policy in a separate layer (WS-F). Built in an isolated worktree.
…to make validate

test_mvtwin.py proves the four load-bearing properties (reference-gating,
epsilon-unlinkability, fringe-as-leading-indicator, holographic tamper-evidence)
with measured numbers and asserts, but nothing invoked it: it lives outside
tools/, so pytest tools/ -q (tools-tests.yml) never collects it, and it was on
no Makefile target either. A self-test nobody runs in CI proves nothing about
the PR that shipped it.
@mdheller
mdheller force-pushed the feat/multiverseal-twin-refimpl branch from 8fbd1a1 to 0dccd12 Compare August 4, 2026 19:43
@mdheller
mdheller merged commit 1b15184 into main Aug 4, 2026
8 checks passed
@mdheller
mdheller deleted the feat/multiverseal-twin-refimpl branch August 4, 2026 19:46
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant