Uniswap V4 Hooks: Programmable Liquidity Architecture

Uniswap Labs shipped aggregator hooks on October 22, 2026, on Ethereum mainnet, Base and Robinhood Chain, connecting seven external liquidity sources that Uniswap says cover more than 90 percent of decentralized exchange activity on those chains. The mechanism is a contract wrapper that makes a venue Uniswap does not operate appear to routing software as a v4 pool. It demonstrates what this architecture actually permits. A pool is no longer a place where assets sit. It is a promise that a swap will settle, and the hook is what makes the promise.

A Pool Is a Promise, Not a Place

In v3 a factory deployed a contract per pool and the liquidity sat in that contract. In v4 a single PoolManager holds the accounting for every pool, and a hook contract attaches at pool creation. The PoolManager calls that hook at ten defined points: before and after initialize, add liquidity, remove liquidity, swap and donate.

Four of those callbacks can return a balance delta. The one before a swap matters most, because it lets the hook satisfy the order itself and hand the PoolManager a balance change rather than letting the pool's curve price the trade. Deltas have to net to zero before the transaction's unlock call returns, or everything reverts.

An aggregator hook is that permission at its limit. The hook takes the order, sources the fill elsewhere and settles the delta, so the pool need not hold the liquidity that filled the trade. Routers refresh quotes each block, combine native pools with wrapped venues, settle in a single transaction, and pay protocol fees that feed the UNI burn.

Where the Trust Boundary Moved

v3 had a property worth naming before it is lost. Every pool ran identical code, so reading the contract once meant reading every pool, the only variables being the token pair and the fee tier. An LP faced market risk and nothing else.

v4 replaces that with a composition. A pool is the PoolManager, shared and audited and identical for everyone, plus a hook that is pool-specific, arbitrary and unaudited by default. The hook is immutable in the sense that matters least: it cannot be swapped out once the pool exists, while the hook contract itself can be a proxy, so its logic can change after the audit that cleared it.

So providing liquidity on Uniswap v4 is not a risk statement. It names the half of the system that is safe. The operative question is which hook, and the answer carries code, upgrade authority and, with the delta permissions, the ability to alter amounts.

The Permission Set Is in the Address

One control here is unusually good and almost nobody uses it. The PoolManager does not ask a hook what it can do. It reads the hook's permissions out of the bits of the hook's own contract address, as a bitwise test against one flag per callback, so deploying a hook with a given permission set requires mining a salt until the address carries the right bits.

Capability is therefore public, immutable and readable without touching the code. Before trading against a pool or providing to it, decode the hook address and you know whether it intercepts swaps and whether it holds any of the four delta-returning flags, which is what determines whether the hook can change your amounts. That check costs nothing and needs nobody's cooperation. It cuts the other way too: a bit set for a function the hook does not implement makes the PoolManager call into nothing and revert, a denial of service produced by a deployment error rather than by an attacker.

What the Next Order of Magnitude Takes

Adoption has moved quickly. Hooks linked to deployed pools went from roughly 22,600 in January 2026 to more than 90,000 by September, across 4,689 tracked v4 pools, with v4 at 48 percent of weekly Uniswap volume against v3 at 52 percent, and cumulative v4 volume near $355 billion through June.

What a further order of magnitude requires is not more liquidity. v3's diligence cost was fixed, since one audit covered every pool that would ever exist. v4's scales with the hook population, because each hook is a separate program with its own permissions and upgrade authority. Cheap pool creation is the mechanism by which the review burden grows.

0x analyzed 84,163 hooks across six chains as of September 11 and classified 19.4 percent safe, 54.2 percent malicious and 26.4 percent likely malicious. Treat the proportions as indicative, since 0x runs a competing aggregator and the classification is its own. The structural point survives: the population is large, growing and individually unreviewed.

Which is what makes October 22 the interesting event. Uniswap did not solve that problem, it routed around it by curating seven venues and wrapping them. Scale came from the sponsor picking the counterparties, which is the opposite of permissionless, and it worked.

Failure Modes, and What Is Getting Better

The attack this design enables is divergence between quote and settlement. A hook can present an attractive price during routing and behave differently at execution, and in the worst cases 0x observed, traders received as much as 50 percent less than quoted. That is not a pricing error but a hook using a permission it legitimately holds. Realized losses are moderate and real: roughly $11 million at Cork Protocol in May 2025 and $8.4 million at Bunni in September 2025.

Three defenses hold. Decode the hook address first and reject any delta-returning permission with no stated reason to be there. Require simulated execution against the actual hook rather than a quote, because the quote is the thing being spoofed. And check whether the hook is a proxy, since auditing an upgradeable hook certifies a version rather than a contract.

The direction of travel is better than those statistics suggest. Capability is verifiable from the address, guidance on secure hook construction is published, the population is being measured by parties with an incentive to measure it, and curated wrappers give an institution a reviewed subset to work inside. The honest framing for a technical owner is that v4 is not an AMM with extensions. It is a settlement engine that delegates pricing to third-party code, and it asks for vendor diligence per pool rather than protocol diligence once.

For informational purposes only. Not an offer to buy or sell any security. Available only to accredited investors who meet regulatory requirements.

‍