Hello everyone,
I’m sharing an open-source project that may be relevant to people working on zero-knowledge systems, verifiable computation, and distributed trust infrastructure.
Veramem Kernel is a deterministic core for:
-
append-only, verifiable timelines
-
invariant-enforced state transitions
-
canonical encoding (strict TLV, domain separation)
-
fork detection / deterministic merge / reconciliation
-
cryptographic commitments (hashchain-based)
-
signed commitments (HMAC + Ed25519)
-
device attestation (challenge–response, replay-resistant)
It is not a blockchain and does not require global consensus.
It is not a CRDT and does not prioritize eventual convergence over safety.
The model is:
Safety > liveness
Determinism > convenience
Explicit abstention > unsafe merge
All state transitions are:
-
deterministic,
-
replayable,
-
auditable,
-
structurally validated via invariants.
Golden conformance fixtures are included for:
-
timeline delta / fork / merge / reconcile
-
device attestation (HMAC + Ed25519)
-
canonical wire format
The goal is to provide a minimal truth layer on top of which:
-
ZK proofs,
-
selective disclosure,
-
verifiable reasoning,
-
and privacy-preserving cognitive systems
can be anchored.
The kernel does not enforce a specific proof system — it provides a deterministic and cryptographically anchored substrate that can be integrated with proof systems.
Repository (specs + code + conformance vectors):
Documentation is available in /protocol, /docs, and the conformance suite.
Open Questions
-
How would you integrate zk-SNARK / zk-STARK proofs over timeline commitments?
-
Would Merkle-based commitments improve proof-friendliness vs the current hashchain model?
-
What PQ migration strategy would you recommend for long-horizon memory systems?
-
How would you model fork proofs inside a ZK constraint system?
Thank you for attention.
Julien Lefauconnier