TL;DR
Uniswap is still solid. This run re-read every admin key on Ethereum and then checked what governance actually executed since the last report.
- Admin keys: unchanged. The v4 PoolManager and the v2 fee setter answer to the governance Timelock A built-in delay between an admin approving a change and the change taking effect, so users can withdraw first if they disagree.. The v3 factory and the v4 fee controller are fee-adapter contracts owned by that same Timelock, which also holds their feeSetter role. So does the v4 adapter's fee policy contract. The delay is still 2 days.
- Governance since 2 October: two proposals executed. #101 rotated sentinel addresses on the three Uniswap Earn Vault and curator A shared deposit pool run by code or by a curator. The curator chooses which markets get your money and sets the limits, for a fee.. #102 switched on protocol fees on Arc. Neither touched an Ethereum pool or a fee adapter.
- TVL: $3.57B, down 11.4% from 2 October. ETH fell 17.2% over the same days, so most of the drop looks like price. I could not split it exactly.
- Fees: $81.4M over 30 days, against $197.2M in the last report. The yield is still real. There is just less of it.
- Hooks: one new hook-layer loss is on the hacks list. GoldPesa ($115k, Base) lost funds through its own v4 hook. DeFiLlama records that v4 itself behaved as designed.
Your risks are still the ones you choose: impermanent loss, a thin pool, or somebody else's v4 hook.
Checklist
Audits & contracts. Deployed v2 pairs, v3 pools and the v4 PoolManager have no proxy, upgrade path, pause or selfdestruct. That was confirmed from source on 2026-08-12, and bytecode at a fixed address cannot change afterwards.
I re-listed both Audit A security firm's review of a protocol's code. It lowers the odds of a bug but does not rule one out, so check what was in scope. folders on GitHub this run:
- v3-core
audits/:abdkandtob, for ABDK Consulting and Trail of Bits. - v4-core
docs/security/audits/: final reports from OpenZeppelin and Trail of Bits, plus reports from ABDK, Certora and Spearbit whose filenames mark them DRAFT.
The blog's totals of nine audits and a $2.35M competition were not re-checked. Neither was the Cantina bounty for v4 core: its page loads but renders only with JavaScript.
Admin control. I read every key on Ethereum by eth_call at the latest block (about 26,363,000).
- Timelock: 0x1a9C...35BC.
delay()returns 172,800 seconds (2 days). Itsadmin()is GovernorBravo 0x408E...24C3, andpendingAdmin()is empty, so no handover is queued. - v4 PoolManager:
owner()is the Timelock.protocolFeeController()is still V4FeeAdapter 0x89a5...51db. - v2 factory:
feeToSetter()is the Timelock.feeTo()is 0xf385...5f85, the same TokenJar address that both adapters report asTOKEN_JAR(). - v3 factory:
owner()is still V3OpenFeeAdapter 0xf237...bf5a. - Both adapters:
owner()andfeeSetter()both return the Timelock. - v4 fee policy: the V4FeeAdapter reads its fees from a
policy()contract, 0x1cd8...a314. Itsowner()andfeeSetter()are also the Timelock. The last report did not name this contract.
GovernorBravo's proposalCount() is 102, and the last proposal executed before this window was #100 in July. I read #101 and #102 on-chain. Both were queued on 3 October, became executable on 5 October and now read Executed. I decoded their actions with getActions.
- #101: six
setIsSentinelcalls on three vaults whosesymbol()reads uniUSDC, uniUSDT and uniETH. All three are owned by the Timelock and share one curator address, which the proposal text names as Gauntlet. This is Uniswap Earn, a separate lending product on Morpho, not the AMM. - #102: "Protocol Fee Expansion: Arc". Three actions, starting with
setWormholePeer, that send a cross-chain message turning on v2, v3 and v4 fees on Circle's Arc chain.
Neither proposal targets an Ethereum pool, either fee adapter, the fee policy or the PoolManager.
The adapters also received a large volume of calls from addresses other than the Timelock. From the evening of 1 October, the V3 adapter took 959 calls: 894 batchTriggerFeeUpdateByPool and 65 collect. The V4 adapter took at least 1,000 batchTriggerFeeUpdate calls between 1 and 9 October alone, which is where Etherscan's page limit cut off. I matched every selector against the verified ABI.
These calls push the fee governance already set onto individual pools, or move protocol fees the pools already owe into the TokenJar. None of them can change the fee setting itself, because only the feeSetter can, and the feeSetter is the Timelock.
The ceilings in pool code are unchanged:
- v3: a protocol share between one tenth and one quarter of the swap fee. The V3 adapter's
defaultFee()returns 0x44, which decodes to one quarter in each direction, the top of that range. - v4: at most 1,000 pips per direction.
- v2: a fixed one sixth of sqrt(k) growth.
Oracle. The core reads no external price feed. Uniswap supplies prices rather than consuming them, so TWAP manipulation on a thin pool hurts the protocols that trust it. If you LP (liquidity provider) Depositing two tokens into a trading pool to earn its swap fees. If their prices drift apart, you end up holding more of the one that fell. into a v4 pool whose hook reads an Oracle The price feed a protocol trusts to value collateral. A wrong or manipulated price can cause unfair liquidations or let bad loans through., you take on that hook's oracle risk.
Liquidity & exit. TVL Total value locked: the dollar value of everything deposited in a protocol or pool. It shows size, not safety. across live versions is $3.57B: $1.52B in v3, $1.18B in v4, $0.87B in v2 and $3.5M in v1.
By DeFiLlama's daily series, the parent stood at $4.03B on 2 October, so the drop is 11.4%. Against the $4.06B intraday figure in the last report, it is 12.1%. ETH went from $2,705 to $2,240 over the same window, a fall of 17.2%.
A TVL drop smaller than ETH's fall fits a mostly price-driven decline, since a large share of Uniswap Liquidity How much you can sell or withdraw right now without moving the price or waiting. Thin liquidity means a slow or costly exit. is paired against ETH. I could not prove it. DeFiLlama publishes no token breakdown for v3 or v4, and the v2 breakdown looked incomplete. So the split between price and outflow is not verified in this review.
Ethereum holds $2.32B, Base $466M, Robinhood Chain $254M and Arbitrum $170M. Exit is one permissionless transaction with no queue. In a hooked v4 pool, a hook that reverts on liquidity removal can still lock LPs in.
Yield: real vs emission. No UNI Emissions and points Emissions are new tokens a protocol issues to pay rewards; they count in the reward yield. Points may turn into tokens later and are never counted in a rate. pay LPs. The yield is trading fees: $14.7M over 7 days, $81.4M over 30 days and $924.6M over the past year (DeFiLlama). The 30-day figure in the last report was $197.2M, so the latest window is about 59% lower. Protocol revenue fell from $15.2M to $8.3M over 30 days.
The decline was gradual, not a one-day gap in the data. Daily fees ran $3M to $5.5M in late September and mostly $1.3M to $3.5M through October, apart from a spike on 29 and 30 October.
Lower volume means a lower fee APR Yearly rate without compounding. Funding here is shown as an APR: the hourly rate × 8,760. on the same pool. Impermanent loss still decides your net. The fee mechanics are unchanged:
- v3: the protocol fee comes out of the LP fee.
- v4: it is charged on top of the trade.
Holder concentration. The top-holder table is still unmeasured. Etherscan's endpoint and Nansen's holder endpoint both returned Pro-subscription errors, the fourth blocked run in a row.
Two balances could be read directly on-chain:
- Governance Timelock (the treasury): 262.2M UNI.
- Burn address 0x...dEaD: 113.9M UNI.
The UNI contract's own totalSupply() still returns 1B, because burning here means sending to the dead address. CoinGecko's total of 886.1M is that 1B minus the dead balance, to the token. It was 887.5M on 2 October, so about 1.37M UNI was burned in 36 days. The previous 51 days burned about 4.0M, and the slower pace matches the lower fees.
UNI holders control the fee dial through governance. The pool code caps that dial, and UNI holders have no claim on your LP position.
Track record. No exploit of v2, v3 or v4 core since 2018. DeFiLlama's hacks list has 1,340 entries, the latest dated 2026-11-05. Forty are dated on or after 2 October, and none is an attack on Uniswap core.
One of the forty is a hook-layer loss. On 2 October GoldPesa lost $114,900 on Base. Its own GPXHooks contract rebalanced liquidity through v4's shared PositionManager without isolating its balances, and an attacker used that to settle an unpaid position. DeFiLlama's entry states that v4 itself behaved as designed. That makes five hook-layer incidents on the list, alongside Cork V1, Bunni V2, VTSwapHook and BoostHook from earlier reports.
Three more of the forty are other projects that only sound related. Ethervista ($18.6k, an overflow in its own pair contract) and Yapeswap ($21.9k, a malicious governance proposal) are separate DEXs. UNCX Network V2 ($86k) was a flaw in that service's vesting contract. None of them is Uniswap.
The list still has only two rows named Uniswap: v1 reentrancy (2020, $220k) and a 2022 phishing campaign ($8M) that drained users' wallets, not contracts.
An X search covering 2 October to 7 November reported no exploit, frontend or DNS compromise attributed to Uniswap. Grok summarised that search, and I did not open the individual posts it cited. It did not surface the GoldPesa incident, so treat it as a weak check.
Worst case
For a v2, v3 or hookless v4 position, theft is not the realistic worst case. The worst case is impermanent loss outrunning fees in a volatile pair, made more likely now that fees run at less than half last month's pace. Governance could also raise your pool's protocol fee to its coded ceiling. That takes a vote plus two days in the Timelock, so nobody can do it quietly.
For a hooked v4 pool, the worst case is losing the whole position. Anyone can deploy any hook, and a malicious or buggy hook can drain the pool or block withdrawals. GoldPesa this month is the fifth such case on the hacks list. The bounty does not cover hooks, and nothing in this stack is insured.
Uniswap Earn is a different risk again. It is a set of Morpho vaults curated by Gauntlet that the Timelock owns. Its risks are lending risks, and this verdict does not cover it.
Bottom line
Solid. The core is immutable, and every governance lever on Ethereum still runs through a two-day Timelock and stops at a fee ceiling written into pool code. The month's governance activity touched a separate vault product and a new chain, not your pools.
What did change is the size of the fee yield, which fell by more than half. This verdict covers canonical v2 and v3 pools and hookless v4 pools in deep pairs. Judge a hooked v4 pool or Uniswap Earn as a separate protocol, and size any LP position for impermanent loss rather than for the APR.
Data appendix
- TVL: $3,569,253,587 for the DeFiLlama parent
uniswap. Children:uniswap-v3$1,515.9M,uniswap-v4$1,180.1M,uniswap-v2$870.5M anduniswap-v1$3.5M. - TVL trend (parent, DeFiLlama daily series): 2026-10-02 $4.027B (-11.4%); 2026-10-08 $3.967B (30d, -10.0%); 2026-08-09 $3.004B (90d, +18.8%); 2025-11-07 $4.732B (1y, -24.6%). All-time high $10.24B on 2021-05-12.
- Price vs outflow: ETH $2,705 on 2026-10-02 (CoinGecko) and about $2,240 on 2026-11-07 (Coinbase $2,236 at 12:45 UTC), about -17.2%. A token-level split of the TVL change is not verified in this review, because DeFiLlama returns no token breakdown for v3 or v4.
- Chain split (top): Ethereum $2.32B, Base $466M, Robinhood Chain $254M, Arbitrum $170M, BSC $76M, Polygon $59M, X Layer $54M, Monad $46M.
- Fees: $1.79M 24h, $14.66M 7d, $81.40M 30d, $924.59M 1y, $6.144B all time.
- Protocol revenue: $180k 24h, $1.62M 7d, $8.33M 30d, $63.52M 1y.
- Audits:
- v3-core
audits/:abdkandtob(GitHub API, this run). - v4-core
docs/security/audits/:OpenZeppelin_audit_core.pdf,TrailOfBits_audit_core.pdf,DRAFT_ABDK_audit_core.pdf,DRAFT_Certora_audit_core.pdfandDRAFT_Spearbit_audit_core.pdf(GitHub API, this run). - Blog totals and Cantina bounty status: not verified in this review.
- v3-core
- Admin and governance (eth_call on Ethereum, latest block, about 26,363,000):
- Timelock 0x1a9C8182C09F50C8318d769245beA52c32BE35BC:
delay()= 172,800 s.admin()= 0x408ED6354d4973f66138C91495F2f2FCbd8724C3 (GovernorBravo).pendingAdmin()= zero. - v4 PoolManager 0x000000000004444c5dc75cB358380D2e3dE08A90:
owner()= Timelock.protocolFeeController()= 0x89a5d5bf00a27d55c02951e49078a5c5771051db (V4FeeAdapter), whoseowner()andfeeSetter()are the Timelock. - V4 fee policy:
policy()on the V4FeeAdapter = 0x1cd822b70a0591420f65e94b9b3a0d0b0fb3a314. Itsowner()andfeeSetter()are the Timelock, and itsPOOL_MANAGER()is the PoolManager above. - v2 factory:
feeToSetter()= Timelock.feeTo()= 0xf38521f130fccf29db1961597bc5d2b60f995f85, which both adapters also return asTOKEN_JAR(). - v3 factory:
owner()= 0xf2371551fe3937db7c750f4dfabe5c2fffdcbf5a (V3OpenFeeAdapter), whoseowner()andfeeSetter()are the Timelock. ItsdefaultFee()returns 0x44 (one quarter per direction). That value was not read on 2 October, so I cannot compare it. Per-tier and per-pool overrides were not read. No executed proposal since then targeted this adapter. - Executed proposals since 2026-10-02: #101 and #102.
proposals()gives both an eta of 2026-10-05 (queued 2026-10-03), andstate()returns Executed for both.proposalCount()= 102. #100 had an eta of 2026-07-27.- #101: six
setIsSentinelcalls on uniUSDC 0x5b45...8be7, uniUSDT 0xb827...d703 and uniETH 0x98d2...2f9d. Each vault'sowner()is the Timelock, and all three share curator 0x9e33...0585. - #102: Arc fee expansion, three actions starting with
setWormholePeer, plus a cross-chain message.
- #101: six
- Adapter calls (Etherscan, selectors matched to the verified ABIs): V3 adapter, 959 calls from 2026-10-01 to 2026-11-07 (894
batchTriggerFeeUpdateByPool, 65collect) from six senders, none of them the Timelock. V4 adapter, 1,000batchTriggerFeeUpdatecalls from one sender between 2026-10-01 and 2026-10-09, where the page limit was hit. The V4 total since then is not verified in this review. - Adapter event logs since 2 October: not verified in this review. The public RPC refused archive
eth_getLogs.
- Timelock 0x1a9C8182C09F50C8318d769245beA52c32BE35BC:
- Oracle: the core consumes none.
- Token: UNI $7.05 (from $8.98 on 2026-10-02), market cap $4.29B (rank 24), FDV $6.25B, -8.6% over 30 days (CoinGecko).
- Supply: CoinGecko total 886,148,419, which equals on-chain
totalSupply()(1B) minus 113,851,581 UNI held at 0x...dEaD. That is about 1.37M burned since 2 October. - CoinGecko circulating reads 607.9M, against 625.2M last run. The methodology change behind that is not verified in this review.
- Supply: CoinGecko total 886,148,419, which equals on-chain
- Holder concentration:
- The top-holder table is not verified in this review. Etherscan
tokenTopHolderswas not re-tried after three Pro-paywall failures, and Nansentoken_current_top_holdersreturned a Pro-subscription 403. - On-chain balances: Timelock 262,247,996 UNI; burn address 113,851,581 UNI.
- The top-holder table is not verified in this review. Etherscan
- Incident record: DeFiLlama hacks list, 1,340 entries, latest 2026-11-05, 40 since 2026-10-02.
- Hook-layer: "GoldPesa", 2026-10-02, $114,900, Base. Its GPXHooks contract mishandled v4 PositionManager accounting, and DeFiLlama states that "Uniswap v4 itself behaved as designed". Earlier hook-layer rows: Cork V1 ($12M), Bunni V2 ($8.4M), VTSwapHook (amount null), BoostHook ($187,000).
- Similar names that are not Uniswap: Ethervista, UNCX Network V2 (vesting contract) and Yapeswap.
- Rows named Uniswap: "Uniswap V1", 2020-04-18, $220,000, reentrancy; "Uniswap (Phishing Attack)", 2022-07-11, $8,000,000.
- Recent-news scan: an X search (Grok) for 2026-10-02 to 2026-11-07 found no Uniswap exploit or frontend compromise. Posts covered fee burns and the TokenJar. I did not open the individual posts, and the scan missed the GoldPesa hook incident.
Maintained monthly. Methodology: DeFi Research Instruction v2.