Westwyrd
Protocol
A deterministic execution and verification layer engineered for institutional-grade settlement. The protocol enforces cryptographic validation of state transitions through zero-knowledge proofs, eliminating reliance on trust-based intermediaries while maintaining formal correctness guarantees across all execution paths.
The Westwyrd Protocol operates across a three-layer architecture in which each stratum enforces a distinct class of invariant. Layer 1 serves as the immutable settlement foundation, anchoring all reserve claims through cryptographic proofs of inclusion verified against canonical chain state. Layer 2 manages the propagation of verified state between settlement domains through Merkle proof attestation, watchtower monitoring, and challenge-response mechanisms derived from optimistic rollup theory.
Layer 3 provides the deterministic execution fabric — a institutional rollup with its own sequencer, block production, and fraud proof system in which every state transition produces a verifiable proof of correct execution. This layered separation enforces a strict dependency hierarchy in which no higher-layer operation can override or circumvent the settlement guarantees established at the base layer.
The protocol does not custody assets, make discretionary decisions, or introduce probabilistic execution paths. All outcomes are derived from verifiable inputs, circuit-bound constraints, and on-chain enforcement — eliminating the class of systemic risks associated with intermediary discretion.
Layer 1 establishes the immutable reserve anchor for the entire protocol stack. All base assets are verified at this level through cryptographic proofs of inclusion computed against the canonical chain state. Settlement guarantees are enforced through smart contract logic formally verified using Rocq, with each contract function bound to specific circuit constraints that prevent divergence between the proof system and the execution logic.
The L1 layer serves as the root of trust for the entire protocol stack. Its security model is inherited directly from the underlying blockchain consensus mechanism — Ethereum's proof-of-stake consensus provides probabilistic finality within approximately 12 minutes (two epochs), after which state is considered irreversible. No higher-layer operation can override or circumvent the settlement guarantees established at this level.
Layer 2 manages the propagation of verified state between settlement domains. Bridge operations are secured through Merkle proofs of inclusion verified by on-chain contracts, watchtower monitoring that provides liveness guarantees, and challenge-response mechanisms that allow any honest observer to contest fraudulent state roots.
State transitions at this layer are validated through zero-knowledge proofs before propagation. The proving circuit encodes the state transition function as a system of R1CS equations. The verifier — deployed as an immutable smart contract on L1 — checks the proof against the circuit's verification key, accepting only proofs that demonstrate correct execution. This provides the same security guarantees as re-executing every transaction at a fraction of the cost.
The relay infrastructure employs a 3-of-5 watchtower model in which five independent observers monitor cross-domain message passing. A message is considered valid if at least three watchtowers attest to its correctness — Byzantine fault-tolerant up to two adversarial nodes.
Layer 3 enforces execution integrity through deterministic computation bound to circuit constraints. The execution environment operates as a institutional rollup with its own sequencer, block production cadence, and fraud proof system. Every state transition produces a verifiable proof of correct execution committed to L1 for final settlement. The forced-inclusion mechanism guarantees censorship resistance: if the sequencer refuses to include a valid transaction, the user can submit it directly to L1.
Block production follows a deterministic ordering protocol in which transactions are sequenced by arrival time within each epoch. The sequencer has no discretionary reordering capability — eliminating MEV extraction by the sequencer operator. This is critical for institutional settlement where predictable execution ordering is a prerequisite for regulatory compliance and fiduciary obligation.
The proving pipeline operates asynchronously: transactions are executed immediately, and proofs are generated in a parallel pipeline running approximately 2–4 minutes behind the execution frontier. This provides low-latency execution while maintaining full ZK proof security guarantees.
Every unit of value is mathematically provable and independently verifiable. State transitions are deterministic and circuit-bound — given the same inputs and pre-state, the protocol always produces the same output. There are no probabilistic execution paths or discretionary decision points. This property is enforced at the circuit level where the R1CS constraint system rejects any non-satisfying witness.
The protocol does not custody, rehypothecate, or exercise control over user assets at any point in the settlement lifecycle. Asset ownership is determined exclusively by cryptographic key possession and state transitions require explicit cryptographic authorization from the asset owner. This eliminates counterparty risk in the custodial sense — a property that distinguishes the protocol from centralized settlement systems.
Settlement outcomes are resistant to manipulation through cryptographic enforcement. The security model does not rely on economic incentives, game-theoretic assumptions, or behavioral expectations. Correctness is guaranteed by the soundness of Groth16, the collision resistance of Poseidon, and the binding property of Pedersen commitments.
WEFT is a source medium, not an effect. Where generative models optimize for plausibility — media that looks real — WEFT optimizes for provenance: media that is verifiable. A WEFT frame carries a proof of the physics and the recipe that produced it. Not “this file was not edited,” but “this image is the verified render of this provable field under this law.”
This defines a category: Provable Media. It does not compete with generative AI — it is the complement. Every convincing synthetic frame raises the market value of one that can prove what it is. The more the industry generates, the more a verifiable frame is worth.
The distinction is structural, not incremental. A diffusion model cannot produce a proof of its own process even in principle — its generation is stochastic sampling, with no deterministic recipe to fold into a root. To offer what WEFT offers, a generative system would have to abandon its architecture and become a physics engine. Provability is not a missing feature; it is a property their foundation excludes.
WEFT was built under a protocol — WYRD: adversarial collaboration for heterogeneous AI agents. No shared framework, no common vendor, no orchestration runtime; a git repository, append-only text files, and discipline. It turns agents from different labs — different vendors, sizes, context windows — into a system where no claim survives unexamined and every artifact carries its provenance.
Wyrd — Old English for “that which has been woven”: the binding, unwritable record of deeds from which present truth is read. The ninth-century word for an append-only ledger of witnessed actions determining trust. The backronym and the etymology say the same sentence — the mark of a name that was already true.
Work becomes a claim only when witnessed — scoped, evidenced, written before checking.
The seat yields its claim to adversarial verification. The humility is structural.
The verdict lands append-only. Git is the audit log. Nothing unwritten.
The record determines trust — what stands written is what you build on.
Named roles with charters — Operator, Conductor, Implementer, Verifier, Wisdom.
Append-only proof-of-work plus a board. Git history is the audit log.
Every gate is proven able to fail, in a sandbox, before it is trusted.
Assert don’t launder; verify don’t trust; one writer per file; the operator’s hand on every irreversible act.
A gateway directory, an optional config-driven server, and a git repo.