Akita’s Post-Quantum PCS Makes the Verifier the New ZK Bottleneck

A new zero-knowledge primitive is targeting one of the hardest post-quantum trade-offs: keeping proofs small while making verification practical.

Akita is a lattice-based polynomial commitment scheme introduced by LayerZero Labs researchers with contributors from a16z crypto, Carnegie Mellon University, USC, and Georgetown University. Polynomial commitments sit beneath many proving systems: they let a prover commit to a computation trace and later prove evaluations of that committed data.

The paper’s design uses setup offloading. Public setup matrices are committed ahead of time, while the verifier’s work for processing those matrices is deferred and proved against the commitments. The stated goal is to avoid the square-root verification behavior associated with an earlier lattice-based design while retaining logarithmic proof size under the paper’s asymptotic model.

The reported measurements are notable. In the authors’ benchmarks, Akita proofs measure 61–70 KB and verify 10× to 94× faster than the compared lattice-based systems. When integrated into Jolt, the paper reports a 1.3×–2.2× prover speedup and a 2.2×–7.4× verifier speedup over Jolt using Dory, with every evaluated proof below 100 KB.

That shifts the engineering question for zkVM builders. Post-quantum security is often associated with hash-based constructions whose proofs can become large. Akita instead targets sparse workloads, where the committed polynomial may be enormous but contain relatively few nonzero entries. The paper says its prover’s memory overhead grows sublinearly beyond storage for the polynomial itself, a property that matters for large execution traces and distributed proving.

The implementation is public in Rust and is organized as a modular crate family. The repository separates verifier-facing components from prover APIs, and it includes versioned schedule artifacts that are checked for consistency before use. That separation is useful for systems such as ICP integrations: a canister or off-chain verifier can depend on a narrower verification surface instead of inheriting the entire proving stack.

The practical caveat is important. “Production-ready” is LayerZero’s characterization, while the performance figures are the authors’ own benchmark results, not an independent audit or a guarantee for every workload. The announced Jolt deployment is also described as a future first deployment, not evidence that a live ICP-compatible production path already exists.

For ICP builders, Akita is therefore best read as an infrastructure signal rather than an immediate dependency recommendation. If post-quantum proofs become relevant to identity, cross-chain attestations, or long-lived state commitments, the decisive metric may be verifier cost and memory—not merely whether a scheme avoids elliptic curves.