Veramem Kernel — deterministic trust layer for cognitive and zero-knowledge systems

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