Aave Labs traces the project to the November 2017 ETHLend token sale and a peer-to-peer lending model. Matching individual lenders and borrowers proved difficult to scale. The team rebuilt around pooled liquidity, launching Aave V1 in January 2020. This change is the historical hinge: borrowers could interact with a market’s available liquidity instead of waiting for a particular counterparty.

The Labs retrospective is a first-party governance contribution, not a neutral evaluation. A March 2020 community AMA provides another contemporary record, with direct questions about lending and the ETHLend transition.

Supplying collateral and borrowing against it are separate decisions. Aave tracks the relationship between the value of a user’s collateral, the applicable liquidation thresholds and outstanding debt through a health factor. Below the required boundary, a position becomes eligible for liquidation. A liquidator repays debt and receives collateral with an incentive for performing that work.

This is why collateralization should not be mistaken for a guarantee. Prices can move, interest can grow and a position can become vulnerable while its owner is offline. The documentation explicitly avoids naming one universally safe health factor. Editorially, readers should also distinguish a protocol’s solvency from an individual borrower’s outcome: a liquidation can protect the market while imposing a loss on that borrower.

Aave V4 launched on Ethereum on March 30, 2026. Its Hub and Spoke model separates shared liquidity from markets with their own collateral and risk configurations. A Hub holds liquidity; connected Spokes offer different borrowing environments. The design aims to use capital across markets without making every market identical.

On July 15, 2026, Labs announced V4’s first expansion to another chain, Avalanche. Version and deployment therefore matter when reading any Aave guide. An explanation of an older Ethereum market is not a specification for every V4 Spoke. The useful research habit is to identify the chain, contracts, version and market before comparing rates, collateral rules or risk.

GHO is Aave’s dollar-targeting, overcollateralized stablecoin. The documentation describes users creating it against collateral and governance controlling parameters; interest from minters supports the DAO treasury. GHO and AAVE have different roles: one aims to function as a stable-value asset, while the other participates in governance and the project’s token economy.

A dollar target is a design objective, not a promise that every market trade clears at exactly one dollar. Readers should ask about collateral quality, redemption or repayment paths, available liquidity and the people authorized to change settings. GHO adoption may strengthen the ecosystem, but the link from usage to the value of AAVE still depends on revenues, expenses and governance decisions.

The 2025 Aavenomics discussions proposed a clearer connection between protocol finances and AAVE, including funded buybacks and changes to the safety system. The implementation thread records voting progress and an April 2025 update announcing that purchases had begun. That supports a dated execution claim; it does not make the initial proposed weekly budget a permanent promise.

Debate extended beyond whether holders wanted purchases. Participants considered treasury resilience, risk coverage and how discretion should be exercised. Areta’s published voting rationale supported the direction while calling for monitoring. Our interpretation is that buybacks belong in a full capital-allocation account: gross purchases, ongoing expenses, treasury runway and expected losses all influence sustainability.

Umbrella’s 2025 rollout introduced an asset-focused approach to covering deficits and prompted discussion about shifting incentives away from the earlier Safety Module. Coverage has conditions and limits. A high displayed reward rate should never be separated from the risk the staker is being paid to accept.

Before V4’s launch, Labs published a security update covering reviews, verification and testing. These are inspectable safeguards, not proof that every future market configuration is safe. For a lending system, contract correctness, oracle behavior, collateral concentration, liquidation liquidity and governance are interdependent. An excellent contract can still participate in a badly configured market; an audit needs to be read with its date and scope.

V4's liquidation engine replaces a fixed close factor with a target health factor configured for each Spoke. The liquidator can repay enough debt to restore that target instead of automatically taking a fixed fraction of the position. Bonuses also respond to how unhealthy the position has become. This gives governance additional control over borrower outcomes, but it does not eliminate forced sales. It changes the rules determining their size and compensation.

A borrower comparing versions needs the parameters of the actual Spoke, particularly when a small price change approaches the threshold.

TokenLogic's February 2026 study illustrates why choosing those parameters is a policy decision. It replays V3 events under hypothetical V4 configurations, contrasting smaller interventions with stronger liquidator incentives and larger safety buffers. The authors explicitly present boundary cases rather than expected operating regimes. More revenue recaptured from liquidation activity is not automatically better for the borrower whose collateral was sold. Conversely, an attractive borrower experience can leave too little compensation for reliable execution during congestion.

The simulation helps explain that tradeoff; it cannot establish how a future shock will unfold.

The March 2026 CAPO incident demonstrated an operational failure between off-chain calculation and on-chain enforcement. Chaos Labs acknowledged that it operated the oracle whose update mismatched a stale ratio with a newer timestamp. A rate limiter prevented the two values from advancing consistently. The team reported manual realignment and proposed broader stale-state tests. This was not a market decline making ordinary collateral insufficient; the reported price cap itself was wrong.

Independent checks must therefore examine parameter updates and their interactions, not merely whether a familiar oracle provider appears in the architecture diagram.

TokenLogic's reimbursement proposal identified thirty-four affected accounts and separated the amount owed to users from the net treasury cost after recovered liquidation proceeds. It linked a machine-readable calculation and proposed a governance-funded payment. Those distinctions matter for accountability: a position's liquidated collateral, the liquidator's bonus and the DAO's eventual expense are different quantities. The proposal is a record of the compensation process, not an independent receipt for every payment.

Its existence also shows that correcting protocol state does not by itself reverse the economic harm of transactions already executed.

The service providers' April 20 report traced the rsETH crisis to Kelp's LayerZero route, where a single-verifier configuration accepted a message without the corresponding source burn. The attacker then supplied released collateral to Aave and borrowed other assets. According to the report, Aave's contracts were not compromised, yet their markets faced exposure because an accepted asset's backing had changed. Guardians froze affected reserves and stewards adjusted parameters.

This is the difference between software following its rules and a lending market remaining economically sound: correct execution cannot manufacture backing missing elsewhere.

The June 1 development report says attacker positions had been liquidated through AIP 478 and ETH liquidity restored across affected deployments, while rsETH remained frozen as a precaution. It also describes recovery coordination and lockbox work. That later record is more informative than leaving April's modeled shortfall as a permanent current balance. It remains a publisher's dated operational report, not a full independently reconciled recovery statement.

Neither successful collective fundraising nor reopening one reserve proves that every affected asset, legal claim and user's opportunity cost was resolved identically.

The April 22 finance proposal disclosed that buybacks had stopped on April 19, following the rsETH incident. TokenLogic's rationale was to retain flexibility while the DAO's possible response and external recoveries remained uncertain. This makes the earlier execution history more precise: purchases did begin, but the program was not an immutable promise to keep buying through every crisis. A treasury holding its own token and a treasury holding assets available for remediation have different capacities.

The dated pause does not by itself establish the status of every later funding proposal or transaction.

The disagreement continued into September. ApuMallku repeatedly asked for an actionable timetable for token economics, intellectual property and the proposed brand structure. Stani replied on September 14 that work was progressing and an ARFC would follow. Zodd's later hope for a launch before year-end was a participant's expectation. These posts support describing unfinished public accountability questions, not treating a reported future buyback design as deployed code.

The distinction preserves both the holder's frustration and the boundary between a public commitment, a governance decision and an executed mechanism.

The March Aave Will Win framework proposed directing product revenue to the DAO while funding Labs and absorbing responsibilities from departing contributors. It separated brand governance and V4 ratification into their own processes. The budget combined stablecoin funding, vested AAVE and milestone-dependent product grants. This is an organizational bargain rather than a cost-free transfer of income. Tokenholders would need to evaluate delivery, contributor concentration and the expenses required to generate that revenue.

A larger gross inflow is not the same thing as a larger surplus available for reserves or token purchases.

The subsequent April development report identifies concrete fee-routing changes after approval: Horizon fees and Aave's share of CoW partner fees went to DAO-controlled collectors. It explicitly says the routing did not create a new fee source or change the underlying economics. This is narrower and stronger evidence than assuming every Aave-branded product automatically belongs to tokenholders. The practical verification is the recipient and settlement path of each flow.

An interface integration, its company's operating costs and the lending contracts' own fees remain separate accounting subjects even when governance seeks to align them.

The V4 licensing proposal couples a time-limited business license with a contributor agreement. Labs describes the agreement as a grant of usage rights rather than an outright transfer of each contributor's ownership. MconnectDAO supported clarity while questioning whether contributor friction and a five-year restriction were necessary. Labs responded that the DAO could shorten the period.

The thread documents a genuine disagreement within an open-development culture: publicly inspectable code, unrestricted reuse and enforceable community ownership are related aims, but they are not identical legal arrangements.

September's custodied-collateral proposal illustrates another boundary. The proposed isolated market would represent assets held at Anchorage with non-transferable CoCT receipts maintained through Chainlink infrastructure. The receipt itself would carry no direct claim on the custodied asset; legal recourse would follow the account-control agreement. This is a proposal for combining contractual custody with on-chain lending, not proof of universally permissionless ownership. A borrower, supplier and liquidator would encounter different admission rules and dependencies.

The forum's discussion stage is not evidence that this particular market had completed on-chain activation.

Current sGHO documentation describes an Ethereum ERC-4626 vault whose shares accrue value in GHO. It distinguishes that vault from the legacy Merit-based savings contract and exposes both a configured rate and a paused state. Its stated design retains deposited GHO rather than lending it onward. These features explain the product without promising a permanent rate or uninterrupted access. A dollar-denominated balance still depends on GHO's value and the deployed contracts.

Old tutorials referring to the previous savings contract should not be treated as interchangeable with the current share accounting.

Umbrella serves a different purpose. Its documentation describes staking aTokens into asset-specific coverage, with those positions available to absorb deficits and rewards paid for accepting that exposure. Coverage also has network boundaries, offsets and withdrawal timing. A person supplying an asset, staking the resulting receipt into Umbrella and buying AAVE has taken three different positions. A familiar application can present all of them without making their risks interchangeable.

Descriptions of reduced legacy slashing should therefore not be generalized into a promise that current Umbrella positions cannot lose principal.

mrjoaoms introduced a retention proposal as an affected Arbitrum WETH supplier who said they had never held rsETH or used the bridge. The revised plan sought temporary incentives and withdrawal limits after normal access returned. MconnectDAO and bobgodwinx helped shape it, including changes intended to avoid privileged withdrawal classes. This is a participant-authored proposal, not an enacted withdrawal policy or verified balance statement.

It shows why repairing collateral backing and retaining suppliers are different tasks: a market can survive an incident while users reconsider the cost of making their liquidity available.