Summary: Silo Finance is best understood as a permissionless isolated lending-market framework that tries to localize credit risk at the market level and add a protocol-native fallback when normal DEX liquidations fail. Instead of one pooled money market with shared asset risk, each Silo market is a one-sided two-vault pair with its own collateral rules, oracle setup, liquidation thresholds, and interest-rate configuration. The reusable mechanism insight is that Silo does not just isolate listings; it also introduces a second liquidation path — collateral-debt swap (CDS) defaulting — that explicitly writes off debt and reimburses lenders with collateral share tokens when external liquidity is unavailable.
What it does:
Lets anyone deploy isolated lending markets made from two ERC-4626 vaults, one for collateral and one for the loan asset
Uses one-sided markets where Token A can be pledged to borrow Token B, but not vice versa inside the same market
Offers both borrowable deposits and protected collateral deposits, so users can choose whether supplied assets are available for borrowing
Runs two liquidation paths: standard collateral-sale liquidations and an internal collateral-debt swap (CDS) path that writes off debt and distributes collateral share tokens to lenders
Exposes a hooks-based architecture so liquidation logic and other protocol actions can be extended through modular external components
Key claims:
The user docs explicitly frame Silo v3 as a money-market design that aims to guarantee lender solvency without relying on instant market liquidity, which is the clearest statement of the protocol’s design ambition
Each market is isolated and one-sided, so lenders in a given market are underwriting that specific borrower/collateral/oracle setup rather than a protocol-wide asset pool
The architecture docs say markets are permissionless to deploy through Silo Factory and immutable by default unless the deployer opts into upgradable settings, which makes market creation and deployer configuration meaningful control surfaces
Silo’s dual-liquidation model matters more than the headline “isolated lending” label: standard liquidations sell collateral on a DEX, while CDS liquidations forgive debt internally and compensate lenders with collateral share tokens instead of forcing an onchain sale
The lender-protection docs say CDS can coexist with standard liquidations and is designed as a fallback for stressed or illiquid markets, with lower liquidator rewards because the liquidator no longer performs execution-risky swaps
The developer docs show that each asset can use separate solvency and max-LTV oracle settings, which means practical authority can sit in oracle choice and price-source design as much as in headline collateral factors
The interest-rate model is described as a PI-controller-style dynamic IRM with utilization targets and safety bands, which is analytically useful because Silo treats rate control as an endogenous policy surface rather than a static slope table
The risk docs temper the stronger marketing language: risk is localized, not eliminated, and lenders are still underwriting the borrower, collateral, oracle, and liquidity conditions of the chosen market
Whitepaper: No single canonical whitepaper or litepaper stood out in this pass. The strongest primary materials were Silo’s official docs for intro, isolated pairs, liquidation design, architecture, and risk disclosures, plus the public silo-contracts-v3 repository; see ../whitepapers/silo-finance-primary-sources-2026-05-08.md.