On 18 April 2026 an attacker pulled about 116,500 rsETH, Kelp DAO’s restaked ether token, out of its LayerZero bridge without the ether that was supposed to back it. Part of that unbacked rsETH went straight into Aave V3 as collateral. Against 89,567 rsETH, worth around 221 million dollars at the time, the attacker borrowed 82,650 WETH and 821 staking means depositing a proof of stake coin with a protocol that stakes it for you and issues a transferable token (such as Lido stETH or rETH) representing the deposit plus rewards, which you can use or sell while the coins stay staked.">wstETH, according to figures from Aave’s incident report published by The Block. Real ether left the blockchain that runs exactly as written when it is called, holding and moving funds according to its code, without an intermediary able to change the outcome.">smart contract holding reserves of two or more tokens, supplied by users, that a decentralized exchange uses to execute trades automatically at a price set by a formula.">pool; tokens of doubtful value stayed behind, and the largest venue for crypto lending and borrowing had a hole in its books.
Depositors reacted within hours. DefiLlama’s series for Aave V3 fell from 25.95 billion dollars on 18 April to 17.54 billion on 20 April, and Galaxy Research found that WETH utilisation on Aave went from about 88.6% to effectively 100% within six hours, then stayed above 99% for 12.7 days. For almost two weeks, most WETH suppliers could not take their ether out. Nobody broke Aave’s code. The damage came through the two machines every lending protocol runs: one that sets interest rates, and one that decides what collateral is worth and when to sell it.
Those two machines explain most of crypto lending and borrowing. They work differently in a shared pool (Aave, Compound, Spark) and in isolated markets (Morpho, Euler v2, Kamino). Below: a real interest rate curve with the parameters in force on 28 October 2026, the end of Aave’s stable rate, eMode, a supply and a borrow on the Sepolia testnet with test tokens, and the cases in which lenders actually lost money.
Two reminders before going further. DeFi means financial services run by smart contracts on a public blockchain, with no company holding customer funds; the basics are in our guide what DeFi is and how it works. When you supply to Aave you receive an aToken whose balance grows block after block; the previous lesson, how interest-bearing tokens work, explains that mechanism, so it is not repeated here.
How crypto lending and borrowing works on a protocol
A DeFi loan has three parties. A lender deposits an asset, often a stablecoin such as USDC, into a smart contract. A borrower deposits collateral, for example ether, and takes the lender’s asset out. The contract sits between them, holds the collateral and applies fixed rules. There is no credit check, no identity and no repayment date.
What replaces the credit check is overcollateralisation. The borrower must lock more value than they take. On Aave V3 Ethereum, as read on-chain on 28 October 2026, WETH has a loan-to-value (LTV) of 81%: 10,000 dollars of WETH lets you borrow at most 8,100 dollars. If the collateral loses value and the loan is no longer covered with a margin, anyone can repay part of the debt and take part of the collateral at a discount. That is a liquidation, and it is how the lender gets paid back without a court or a collector.
Why would anyone borrow less than they own? To keep exposure to an asset they do not want to sell, to get dollars for a few weeks, to take a leveraged position, or to farm a yield elsewhere. Lenders, on their side, earn the interest borrowers pay, minus a cut kept by the protocol. Nothing else funds the yield. When a protocol advertises a lending rate, it is a share of what borrowers are paying at that moment.
Lending protocols are also among the biggest pieces of DeFi. A note on reading their size: DefiLlama’s TVL for a lender counts what is supplied minus what is borrowed, and shows borrowed funds as a separate series. Our guide to measuring DeFi TVL explains why. The figures in the table further down follow that convention.
Pool markets: Aave, Compound and Spark share one pot per asset
Aave V3: one pool, one rate, many collaterals
In a pool design, every supplier of an asset on a given chain puts it into the same pot. All USDC supplied to Aave V3 on Ethereum sits in one reserve, and every borrower of USDC draws from it, whatever collateral they posted. There is one borrow rate and one supply rate per asset, recalculated at each interaction. A borrower can mix several collaterals in one account, and the protocol values the whole basket.
Each asset carries its own risk settings, voted by the Aave DAO. These are the main ones for five large reserves of Aave V3 Ethereum, read from the Pool contract on 28 October 2026 at about 09:46 UTC (block 26,290,418). Caps are in token units.
| Asset (Aave V3 Ethereum, 28 Oct 2026) | LTV | Liquidation threshold | Liquidation bonus | Reserve factor | Supply cap | Borrow cap |
|---|---|---|---|---|---|---|
| USDC | 75% | 78% | 4.5% | 10% | 3.0 billion | 2.7 billion |
| USDT | 75% | 78% | 4.5% | 10% | 3.48 billion | 3.09 billion |
| WETH | 81% | 84% | 5.0% | 15% | 2.7 million | 2.4 million |
| wstETH | 79% | 82% | 6.0% | 35% | 950,000 | 28,000 |
| WBTC | 81% | 85% | 5.0% | 50% | 38,200 | 1,500 |
The LTV is how much you may borrow when you open the loan. The liquidation threshold is the level at which the position can be liquidated; the gap between the two is your buffer. The liquidation bonus is the discount a liquidator gets on the collateral it buys. The reserve factor is the share of interest the protocol keeps. Caps limit how much of a risky asset the pool can accept: WBTC can be supplied up to 38,200 coins but only 1,500 can be borrowed, so it works mainly as collateral.
The strength of the pool is liquidity: one deep pot per asset, so a supplier can usually withdraw and a borrower can usually find funds. Its weakness is the reverse side. If one collateral asset turns out to be worthless, the loss lands on everyone who supplied the asset that was borrowed against it. That is exactly what the rsETH weekend tested.
Compound III: one borrowable asset, collateral that earns nothing
Compound, the protocol whose COMP token launched the 2020 farming wave told in our history of DeFi, rebuilt its design in version III, also called Comet. Each Compound III market has a single base asset that can be borrowed, USDC on its first Ethereum deployment. Other assets are accepted only as collateral, and the Compound III documentation states that supplied collateral does not earn interest. The upside is a simpler risk picture for USDC suppliers. As of 28 October 2026, DefiLlama shows Compound V3 at 1.37 billion dollars of TVL and 0.64 billion borrowed, against a peak of 2.56 billion in August 2025.
SparkLend: stablecoin liquidity from Sky
SparkLend is the money market of Spark, a project tied to Sky, the former MakerDAO. Its documentation describes it as a market whose stablecoin liquidity comes directly from Sky, rather than only from outside depositors. It works like a pool: one reserve per asset, several collaterals. Our earlier lesson on CDP stablecoins such as DAI and USDS shows where that liquidity is created. DefiLlama gives SparkLend 5.05 billion dollars of TVL and 2.69 billion borrowed on 28 October 2026.
Aave V4: shared hubs and spokes since 30 March 2026
Aave V4 went live on Ethereum mainnet on 30 March 2026, according to Aave’s launch post. It changes the pool idea. Supplied funds go into a Liquidity Hub, which holds assets centrally; Spokes connect to a Hub and each Spoke sets its own collateral types, risk parameters and liquidation rules. Liquidity stays shared, but the rules can differ between, say, a market for one ecosystem and an institutional one. Three Hubs launched (Core, Prime and Plus) with conservative supply and borrow caps that the DAO plans to raise.
It is still small next to its predecessor. On 28 October 2026 DefiLlama counts 0.75 billion dollars of TVL on Aave V4 against 16.51 billion on Aave V3, which keeps running. Our Aave project page follows the protocol’s versions and governance.
Isolated and modular markets: Morpho, Euler v2 and Kamino
Morpho: one collateral, one loan asset, rules frozen at creation
Morpho’s base layer takes the opposite approach. According to Morpho’s documentation, a market pairs one collateral asset with one loan asset and is defined by five settings: the two assets, the liquidation loan-to-value (LLTV), the oracle that prices the collateral and the interest rate model. Once deployed, a market cannot be changed. Anyone can create one without a vote, but the LLTV must come from a governance list (0%, 38.5%, 62.5%, 77%, 86%, 91.5%, 94.5%, 96.5% and 98%) and the only approved rate model is called AdaptiveCurveIRM.
Isolation is the point. USDC lent in a cbBTC/USDC market is exposed to cbBTC, to that market’s oracle and to Morpho’s code, and to no other collateral. If a strange token collapses in another market, its lenders take the hit and yours do not. The price of this design is fragmented liquidity: each market is its own small pool, and a lender who wants spread must split deposits across many markets.
Vaults and curators sit on top
Few people pick markets one by one. Many lenders deposit into a Morpho vault, a token vault that spreads deposits across several markets chosen by a curator, a risk firm or team that sets which markets the vault may use and with what caps. Morpho’s API on 28 October 2026 lists vaults such as Steakhouse’s steakUSDC and Gauntlet’s gtUSDC Prime. The vault looks like one product to the depositor, but the risk is the curator’s choice of markets. Our Morpho project page covers the team and the token.
Permissionless creation has a visible side effect. The same API returns vaults with no recognised curator showing improbable yields, and markets stuck at 100% utilisation whose rate has climbed to the model’s ceiling because the collateral is no longer being repaid. The main Morpho app curates what it shows, but the contracts accept anything. A very high rate on an unknown market is a warning sign, not an offer.
Euler v2: vaults you wire together
Euler v2 builds lending from separate vaults. Each vault holds one asset and has its own caps, oracle, rate model and list of accepted collaterals; a connector contract, the Ethereum Vault Connector, lets one account combine positions across vaults. A vault can be governed (a DAO, a curator or a single person can change its settings) or ungoverned, with settings fixed forever. Euler v2 is a rewrite: version 1 lost about 196.9 million dollars to an exploit on 13 March 2023, and the attacker returned the recoverable funds by 3 April 2023, Decrypt reported. DefiLlama shows Euler V2 at 0.28 billion dollars of TVL on 28 October 2026, down about 87% from 2.2 billion in October 2025.
Kamino: independent markets on Solana
Kamino Lend, the largest lender on Solana in DefiLlama’s list, sits between the two models. Kamino’s documentation describes a market as an independent instance of its lending program with one reserve per token: a market listing USDC, SOL and three other tokens has five reserves that work like a small pool. Liquidations and interest stay inside each market, so a position in one has no exposure to another. Curators can deploy their own market and run vaults across reserves. So Kamino is not Morpho’s one-pair design; it is a set of separate pools. DefiLlama: 1.27 billion dollars TVL, 0.86 billion borrowed on 28 October 2026.
Shared liquidity or contained risk: the trade-off in one table
| Protocol | Design | Who sets risk rules | TVL, billion USD | Borrowed, billion USD |
|---|---|---|---|---|
| Aave V3 | Pool per asset, many collaterals | Aave DAO | 16.51 | 11.57 |
| Morpho | One collateral, one loan asset per market; vaults on top | Fixed at creation; curators for vaults | 11.26 | 5.77 |
| SparkLend | Pool per asset | Spark / Sky governance | 5.05 | 2.69 |
| Compound V3 | One borrowable asset per market | Compound governance | 1.37 | 0.64 |
| Kamino Lend | Independent markets, pool inside each | Market owner or curator | 1.27 | 0.86 |
| Aave V4 | Shared hubs, spokes with own rules | Aave DAO | 0.75 | 0.47 |
| Euler V2 | Linked single-asset vaults | Vault governor, or nobody | 0.28 | 0.23 |
Figures from DefiLlama, retrieved 28 October 2026. Read the table this way. In a pool, liquidity is deep and one bad collateral is everybody’s problem. In isolated markets, a bad collateral stays in its market, but liquidity is thinner and someone, the lender or the curator, must choose markets carefully. Would isolation have contained the rsETH damage? It is an argument, not a tested fact: the same attacker also borrowed on Compound and Euler, and Morpho vault lenders took losses in the Stream case described below.
The interest rate curve on a real market: Aave USDC, 28 October 2026
Rates in DeFi lending are set by one number: utilisation, the share of supplied funds that is borrowed. Low utilisation means idle money, so rates are low to attract borrowers. High utilisation means suppliers may not be able to withdraw, so rates rise to pull in deposits and push borrowers to repay.
Aave uses a curve with two slopes and a kink at an optimal utilisation. For USDC on Aave V3 Ethereum, the rate strategy contract returned these values on 28 October 2026: optimal utilisation 94%, base rate 0%, slope 1 of 4.5%, slope 2 of 10%, with a 10% reserve factor. Below 94%, the borrow rate rises gently from 0 to 4.5%. Above 94%, it adds up to 10 more points over the last six points of utilisation.
The supplier’s rate follows from the borrow rate: borrow rate times utilisation times (1 minus the reserve factor). Suppliers earn less than borrowers pay for two reasons: part of the pool is not lent, and the protocol keeps its share. Here is what the curve gives with those parameters.
| Utilisation (Aave V3 USDC, parameters of 28 Oct 2026) | Borrow rate (APR) | Supply rate (APR) |
|---|---|---|
| 50% | 2.39% | 1.08% |
| 80% | 3.83% | 2.76% |
| 90% | 4.31% | 3.49% |
| 94% (optimal) | 4.50% | 3.81% |
| 94.16% (live at 09:46 UTC) | 4.77% | 4.04% |
| 97% | 9.50% | 8.29% |
| 100% | 14.50% | 13.05% |
All rows are computed from on-chain parameters; the live row matches the rates the contract reported at the same block (about 2,363 million USDC supplied, 2,225 million borrowed). The DAO can change these settings at any time, and the rate moves with every deposit, loan or repayment. At 100% utilisation the same curve would pay USDC suppliers 13.05%. During the rsETH episode it was the WETH market that stayed pinned near 100%; stablecoin utilisation rose as well, USDT from 77.4% to 93.27% and USDC from 76.97% to 82.72% in the first six hours, according to Galaxy Research.
The same day, USDT sat at 94.5% utilisation with a 7.61% borrow rate and 6.47% supply rate, and WETH at 78.9% utilisation with 1.93% and 1.30%. WETH uses an optimal point of 92% with slopes of 2.2% and 6%.
Morpho’s curve moves on its own
Morpho’s AdaptiveCurveIRM, per its documentation, targets 90% utilisation. At the target the borrow rate equals a value called the rate at target; at 100% it is four times that value. The rate at target then drifts: up while utilisation stays above 90%, down while it stays below, between a floor of 0.1% and a ceiling of 200% a year. At the fastest speed, five days of 100% utilisation roughly double it. So a Morpho market corrects over days, without a governance vote.
A live example: the cbBTC/USDC market on Ethereum, with an 86% LLTV, showed 90.07% utilisation, about 344 million dollars supplied, a 5.01% borrow APY and a 4.50% supply APY on 28 October 2026 at about 09:48 UTC, as reported by Morpho’s API. The fee was zero, so suppliers received the whole borrow rate times utilisation.
Variable rate only: how Aave retired the stable rate
Aave once let borrowers choose between a variable rate and a so-called stable rate that stayed fixed unless the market moved far enough to trigger a rebalance. That option no longer exists. Stable-rate borrowing was never enabled on Aave V3 Ethereum, and on 4 November 2023 a critical bug linked to the stable rate was reported through Aave’s bug bounty. The fix halted new stable borrowing. In a proposal dated 31 January 2024, BGD Labs, a developer for the DAO, asked to move all remaining stable positions on V2 and V3 to variable. Version 3.2 then removed all the leftover stable-rate logic from the code: BGD Labs reported on Aave’s governance forum that every active Aave V3 instance had been upgraded to v3.2 on 8 October 2024.
For a borrower today, every Aave loan is variable. A loan opened at 4% can cost 9% next week if utilisation crosses the kink. Fixed-rate borrowing exists elsewhere in DeFi through other designs, which this lesson does not cover.
eMode, isolation mode and caps: borrowing more, or less, by design
The Aave V3 documentation describes Efficiency Mode, or eMode, as a way to get a higher loan-to-value when what you supply and what you borrow belong to the same category of correlated assets. On Aave V3 Ethereum, WETH and wstETH both sit in eMode category 1 on 28 October 2026. Borrowing WETH against wstETH is a bet on two prices that move together, so the protocol accepts a thinner margin than for WETH against USDC. Version 3.2 extended the idea with so-called Liquid eModes: an asset can now belong to several eMode categories instead of one. In V4, eMode becomes dedicated Spokes.
Higher leverage means a smaller buffer. If wstETH trades below ether for any reason, an eMode position close to its limit can be liquidated even though both assets look like ether. The category settings are on the asset page of the Aave app; check them there rather than assuming a number.
Isolation mode goes the other way. A newly listed or risky asset can be accepted as collateral only to borrow a short list of assets, usually stablecoins, and only up to a debt ceiling for the whole protocol. Supply and borrow caps, shown in the parameter table above, add another limit: they cap how much exposure the pool can take to one token, whatever its price. The rsETH case shows why caps matter: the less of a bridged token a pool accepts, the smaller the hole it can leave.
Health factor and liquidation, with numbers
Aave tracks each account with a health factor. Its help page gives the formula: total collateral value times the weighted average liquidation threshold, divided by total borrowed value. Above 1 the position is safe; below 1 it can be liquidated. A later lesson goes deeper into managing it; here is one worked example.
Suppose you supply 10,000 dollars of WETH on Aave V3 Ethereum (liquidation threshold 84%) and borrow 5,000 USDC. Your health factor is 10,000 x 0.84 / 5,000 = 1.68. It falls to 1 when the WETH is worth 5,000 / 0.84, about 5,952 dollars, that is after a drop of roughly 40%. Interest on the debt also pushes the factor down slowly.
At that point a liquidator can repay up to 50% of the debt (up to 100% if the health factor is 0.95 or below, or if the collateral or the debt is worth less than 2,000 dollars, per Aave’s help page) and receive collateral worth the repaid amount plus the 5% WETH bonus. Repaying 2,500 USDC buys it 2,625 dollars of your WETH. You keep the rest, but you have lost the bonus and your exposure.
All of this depends on the price feed. The collateral is valued by an oracle, and a wrong or manipulated price can trigger liquidations or let bad loans through; our lesson on how DeFi oracles are attacked and defended covers that risk.
Try it on Sepolia: an Aave testnet supply and borrow
Aave’s developer documentation states that the protocol is available for testing on the Sepolia testnet, and Aave’s help pages list V3 test markets on Ethereum Sepolia, Arbitrum Sepolia, Base Sepolia, Optimism Sepolia and others. Test tokens have no value, so this is the safe way to see every step. The sequence below follows Aave’s documentation; the app’s labels may change, so follow what it shows.
- Install a browser wallet you control and create a separate account for testing, with no real funds on it.
- Get a little Sepolia ETH for gas from a public faucet. These faucets are run by third parties, not by Aave, and some ask for a small mainnet balance or a sign-in.
- Open app.aave.com (type the address yourself) and switch on testnet mode with the toggle at the top right, then pick the Sepolia V3 market.
- Open the Faucet tab of that market and request test tokens, for example test USDC and test WETH.
- Supply test WETH. The wallet first asks you to approve the token, then to confirm the deposit. You receive the matching aToken and the asset appears as collateral.
- Borrow test USDC. Watch the health factor the app displays before you confirm, and try two amounts to see how it moves.
- Repay part of the loan, then all of it, and withdraw the collateral. Note the gas used by each step in your wallet’s history.
What the exercise teaches: approvals are separate transactions, the health factor reacts to every borrow, and repaying requires the borrowed token plus interest. Moving to real money changes three things. Gas is paid in real ether: on Ethereum mainnet each step can cost from cents to several dollars depending on the moment, so check a live gas tracker, or use Aave on a layer 2 where fees are lower. Liquidation is real. And phishing is real: fake Aave sites are pushed through search ads and social media, so type the address yourself, and never give your seed phrase to anyone who says they are Aave support. Start with an amount you can lose entirely.
When lenders lose money: the real cases
Lending stablecoins feels like a savings account. It is not one, and the cases below show five different routes to a loss.
Collateral risk first. In the Kelp rsETH case, a funding proposal posted on Aave’s governance forum by TokenLogic on 24 April 2026 says a vulnerability allowed an unauthorised release of rsETH from the Ethereum LayerZero adapter and, citing LlamaRisk, counts 152,577 rsETH taken from the lockbox; press reports used 116,500. Aave’s incident report gave two scenarios, not a final figure: bad debt of about 123.7 million dollars if Kelp spread the loss over every rsETH holder, and about 230.1 million if the loss fell only on rsETH on layer 2s, The Block reported. Which one applies depends on how Kelp settles rsETH accounting, a decision outside Aave’s control. The Arbitrum Security Council froze 30,766 ETH held by the attacker, Kelp froze 40,373 rsETH, and a coalition named DeFi United gathered pledges; the Aave proposal asked the DAO for 25,000 ETH. In the sources available to us on 28 October 2026, we found no published final accounting of the bad debt.
Curator and counterparty risk next. On 3 November 2025 Stream Finance said an external fund manager had lost about 93 million dollars of its assets and suspended deposits and withdrawals. Its xUSD token fell to about 0.50 dollars soon after, according to PeckShield, and CoinGecko showed 0.26 dollars the next day, Decrypt reported. Curators including TelosC, Elixir, MEV Capital and Varlamore had exposure through markets on Euler, Silo, Morpho and Gearbox; the research group Yields and More put the direct debt exposure at nearly 285 million dollars. Lenders of plain USDC in those vaults had never bought xUSD; their curator had accepted it as collateral. Our lesson on composability and money legos follows how such chains build up.
Liquidity risk is the quietest one. At 100% utilisation every unit supplied has been lent, so a supplier cannot withdraw until borrowers repay or new deposits arrive. That is what WETH suppliers on Aave faced for almost two weeks in April 2026. The rate curve is designed to fix this within hours or days, but there is no guarantee of when.
Smart contract risk: Euler v1 lost about 196.9 million dollars in March 2023 through a flaw in its own code, and users depended on the attacker returning the funds. Liquidation-system failure: on Black Thursday, 12 March 2020, MakerDAO auctions cleared at zero bids, a story told in our history of DeFi. Depeg risk: a stablecoin you lend or hold as collateral can lose its peg, a risk that depends on its reserves or its design.
| Case | Date | What failed | Who bore the loss |
|---|---|---|---|
| Kelp rsETH on Aave | 18 April 2026 | Bridge adapter released unbacked collateral | WETH lenders exposed under Aave’s scenarios; coalition recovery, final split not published |
| Stream Finance xUSD | 3 November 2025 | External fund manager loss; xUSD used as collateral | Holders of xUSD and lenders in vaults that accepted it |
| Euler v1 | 13 March 2023 | Protocol code flaw | Depositors, until the attacker returned funds |
| BlockFi (centralised) | 14 February 2022 | Unregistered interest accounts; misleading risk statement (SEC findings) | BlockFi paid 100 million dollars to the SEC and states; customers were exposed when it filed for bankruptcy in November 2022 |
| MakerDAO Black Thursday | 12 March 2020 | Liquidation auctions at zero bids | Vault owners |
BlockFi, in the table, is outside DeFi but worth the contrast: depositing with a centralised lender is an unsecured loan to a company. On 14 February 2022 BlockFi agreed to pay 100 million dollars, 50 million to the SEC and 50 million to 32 states, after the SEC found its interest accounts were unregistered securities and that its website misdescribed the risk of its loan portfolio. Celsius and FTX went further, as our history of DeFi recounts. On Aave or Morpho the loan is overcollateralised and visible on-chain; the risks move to code, oracles, collateral and curators.
Rules and taxes for US, UK and EU readers
Rules on crypto lending and borrowing differ by jurisdiction, and DeFi sits in the gaps of most of them.
United States. The SEC’s BlockFi case is the main precedent for centralised interest products; no company takes deposits in a protocol such as Aave, and we found no SEC or CFTC rule written for DeFi lending. The IRS treats digital assets as property, but its virtual currency FAQ, last reviewed on 30 June 2026, does not address crypto-collateralised loans or DeFi lending income. Whether a deposit that returns an aToken is a disposal, and when interest becomes taxable, are open questions; our guide to US crypto taxes covers the basics, and a tax professional can apply them to your case.
United Kingdom. HMRC’s Cryptoassets Manual has a section on DeFi lending and staking. Its page CRYPTO61620 says a DeFi loan is a disposal for capital gains when the lender transfers beneficial ownership of the tokens, which depends on the terms: a recipient free to deal with the tokens points to a transfer. We did not review FCA rules for this lesson.
European Union. MiCA does not cover lending and borrowing of crypto-assets. Its Article 142 asked the Commission for a report; the EBA and ESMA answered in January 2025, estimating DeFi lending and borrowing in the EU at about 1.8 billion euros and noting that 13 protocols handle 86% of the market, per their factsheet. The Commission’s 2026 consultation on the MiCA review asks whether lending and borrowing should be regulated (question 67); as of 28 October 2026 we found no legislative proposal on it. The French and Italian tax administrations have not published a specific position on DeFi lending interest in the sources we checked, so the treatment is unsettled.
A checklist before you lend or borrow
- Find out which design you are using: a shared pool, an isolated market, or a vault whose curator picks markets for you.
- Read the collateral list of the pool or vault you lend into. Your stablecoin is only as safe as the worst collateral accepted against it.
- Check utilisation. Above the kink, rates are high because withdrawals may be hard.
- Treat an unusually high rate as a sign of risk and look for the reason before depositing.
- As a borrower, keep a health factor well above 1 and know the price at which you would be liquidated.
- Remember that every Aave rate is variable and the DAO can change parameters.
- Note the oracle used for your collateral and whether the asset is bridged.
- Try the steps on a testnet first, then start small, and keep records of every transaction for taxes.
The next lessons of the DeFi path go deeper into managing a health factor and into the strategies built on top of these markets.
