Why Zava.

Every credit tier issued by Zava is preceded by an UltraHonk proof that runs on-chain — real Sumcheck + Shplemini + BN254 pairing check. Not a stubbed byte-length, not a "we'd add this later," not a claim in a whitepaper.

Lenders receive a tier + loan cap. That's the whole transmission. Your amounts, client names, and deposit history stay in your browser and on chain in a form only your scan key decrypts.

The whole system is auditable in one click: `honk verified true` events stream live at /proof-log. Tampered proofs return `honk verified false` on the same infrastructure. Completeness and soundness — provable, not asserted.

Stellar's Protocol 26 (Yardstick, live May 6 2026) shipped native BN254 host functions via CAP-74 — the exact primitives UltraHonk needs to verify a real proof on chain in a bounded fee envelope.

The verifier crate is 26 KiB of WASM. Every proof lands in a 4-6 second Soroban transaction. Payment settlement stays in the same 3-5 second window Stellar already gives you. The user experience is a bank's dream — with none of the exposure.

The frontend prover is byte-identical to the CLI prover (both `@aztec/bb.js@0.87.0`), so a browser-generated proof is guaranteed to verify against the same on-chain VK. Not architecturally similar. Identical.

There is no Zava-controlled scoring service. The `CreditRecord` writes to Soroban keyed by your wallet. Any lender that reads Soroban can query it. Any lender that trusts UltraHonk on Protocol 26 trusts the tier.

The record has a 90-day expiry — a design choice. Credit is time-sensitive; a snapshot from six months ago doesn't reflect current discipline. Users re-prove when they need to.

Withdraw everything from the vault and the tier reflects it immediately. `active_weeks` counts what's locked now; `withdrawn_weeks` reflects what was proven historically but has been drawn. The lender sees both.

Traditional credit scoring is trained on salaried employment: fixed paycheck, employer name, tax withholding, credit-card history. None of that describes a designer paid in USDC by four clients across three continents.

Zava reads the pattern the freelancer actually has — consistent weekly savings from irregular income — and proves it without demanding a paper trail the freelancer can't produce.

A 24-week track record at $200/week unlocks up to ~$28,800 XLM in loan capacity. That's not a promise; it's the deterministic output of `active_weeks × range × multiplier`. The formula is in the contract, on chain, forever.