Status and limits

This is a working protocol scaffold, not a finished product, and the distinction is worth stating plainly rather than leaving a reader to infer it.

What's real

• The pool's accounting, access control, pausability, and proof-binding logic are written and tested against MockVerifier.
• All three circuits exist, and real on-chain verifiers for all three are wired, tested, and proven working against real proofs. See The circuits and Proof system. This includes a real LatensPool.borrow() call gated by a genuine solvency proof and a real LatensPool.liquidate() call gated by a genuine liquidation proof, each also confirmed to reject a tampered public input.
• Interest is real and utilization-driven, and supplier yield genuinely compounds through a per-asset index rather than a placeholder number. See Interest, yield & minting.
• Every deployed contract is source-verified on the block explorer for its network, see Deployment for the current set of addresses.

What isn't finished yet

• No independent security audit has been done. A security self-review exists in the repository (contracts/SECURITY_REVIEW.md). It is a genuine review, written by the same author as the code, and is explicitly not a substitute for an independent one.
• Client-side proof generation now exists: frontend/lib/proving/ runs real UltraHonk proving in a Web Worker (@noir-lang/noir_js for witness generation, @aztec/bb.js's UltraHonkBackend for proving), wired into every action that previously submitted a mock "0x" proof. Its output has been cross-checked against the real, deployed Solidity verifiers for all three circuits, confirming byte-for-byte compatibility, not just internal self-consistency. See The circuits's "Client-side proving is real, end to end" section.
• The live testnet deployment has been switched to the real verifiers. script/deployRealVerifiersTestnet.js deployed the real Honk verifiers and called setVerifiers() on the already-live LatensPool and LatensCDP, without redeploying either pool. A full supply → borrow → repay → withdraw cycle was then driven against the live pool with a fresh wallet and real UltraHonk proofs (frontend/scripts/genLiveE2EProofs.mjs
  • script/liveE2ETestnet.js), confirming client-side proving and the real verifiers work together for an actual user flow, not just in isolation. See The circuits's "Confirmed live" section.
• The mock verifier must never reach mainnet. There is no on-chain guard preventing a misconfigured deployment from using it; that gate belongs to the deploy process and to milestone-acceptance review, not to the contract itself.
• Circuit soundness has not been independently audited. The empirical language checks described in The circuits are real, but narrow: they rule out specific, identified failure modes, not every possible one. A circuit audit is a different discipline from a Solidity review, and neither has been performed independently.
• Cross-margin, multi-asset positions are out of scope for this milestone. Positions are single-collateral, single-debt-asset, and isolated per user. Supplying a second asset into a position that already holds one is rejected on-chain, and the interface disables the action rather than letting it fail in a wallet.
• The testnet price feed is a mock that needs a heartbeat. Both contracts reject any solvency-gated call whose price is more than an hour old. A production oracle updates itself; MockPriceOracle only advances when something calls it, so a deployment left alone for an hour starts rejecting borrow, withdraw-against-debt, mint and liquidate with StaleOraclePrice, while supply, repay and burn keep working. The interface re-stamps the feed itself before each affected action, and script/refreshPrices.js does the same from the command line. The re-stamp is permissionless and cannot change what a price says, only how recently it was checked.
• Liquidation amounts become public. The seized collateral and repaid debt amounts for a liquidated position are visible on-chain at the moment of liquidation. See Privacy model for the full boundary of what stays private and what doesn't.

Roadmap

Milestone	Target	Description
M1	Q1 2027	Core privacy capability: confidential deposit, borrow, and health-factor proofs live on testnet.
M2	Q2 2027	Independent security audit of the proof system and liquidation logic, before mainnet exposure.
M3	Q3 2027	Mainnet usage: real deposits and borrows on Horizen, demonstrating product-market fit.