Ethereum's birthday week
Published 2026-08-17 · data: 2026-07-27 → 2026-08-02 (UTC)
Ethereum turned eleven years old a few weeks ago, on Thursday July 30th. This year has seen a lot of discussion about the "return to L1". But what was happening onchain in the week of Ethereum's birthday?
The week's transactions and block auction in a timeline. We will dig into the Sunday transaction surge and the auction intensity on Thursday in more detail below.
Business as usual
In some ways it was a regular week. 12.9M transactions landed in 50,224 blocks. 24% of transactions were ETH transfers, 23% were stablecoins, and 2.5M addresses sent a transaction.
48% of blocks were built by Titan, with Quasar coming in second at 19%. Titan's share of block rewards was best in class at 54% — its average block paid 0.0157 ETH vs the field's 0.0140. Meanwhile ~20% of proposer blocks went to Lido, with a steady proposer fee.
219K blobs were posted, mostly by rollups. Robinhood was the biggest spender, with Aztec coming in second.
These are pretty normal background statistics, at least in recent months. So what was unusual about this week?
A quiet week
It was a relatively quiet week from a gas perspective, with six out of seven days below the 12-week median base fee.
Only one day saw a significant spike, on Thursday 30th July, the birthday itself - what could have been driving this intensity?
A tale of two mints
The obvious candidate is the 11x11 mint, a collaborative art creation from EthLabs and the networked.art team to celebrate Ethereum's big day. This allowed people to create a custom 11x11 pixel canvas, stored and rendered onchain as an SVG. The mint took place over five days, with 35,245 artworks created by 10,880 unique addresses — and collected a further 11,639 times through editions, for 46,884 mints in all.
The most-collected artworks. 11×11 pixels, eleven colours, for Ethereum's eleventh birthday. The top piece, ONDI's #45, was collected 750 times.
But this activity was spread over a couple of days, so despite the 11x11 mint making up 0.4% of the week's transactions, it wasn't responsible for the gas spike (though in its peak hour the birthday mint made up 10.8% of all Ethereum transactions)
Instead it was Compas, a paid mint of 10,000 NFTs using OpenSea's SeaDrop contract, which kicked off at 15:15 UTC and briefly took 62% of all gas in one 5-minute window, driving the base fee to 12.2 gwei.
The other mint: a fraction of the transactions, but gas-priced aggressively enough to drive the base fee from 1.6 to 12.2 gwei.
A sample of Compas pieces, rendered from the contract's onchain SVGs.
The Compas mint cost 42.4 ETH in transaction fees against 11x11's 2.9 ETH, for less than a third as many tokens overall.
Notably, both NFT collections store their metadata fully onchain!
Fake World Assets and fake addresses
Transaction categorisation highlights a couple of other noteworthy things from this week - FWA and Dust transfers.
The first is also NFT related - it was the second week of Fake World Assets, the latest experiment from Tokenworks. FWA combines an onchain gacha mechanism with a native token and a whitelist of existing NFT collections. While the second week was quieter than the first, this playful new primitive grabbed attention and gas fees, with spinners and depositors spending ~6% of the week's gas.
The second is less cheerful. It was a quiet weekend on Ethereum, with gas prices at a low ebb, which gave a bad actor the opportunity to make large numbers of tiny transfers ("Dust transfers").
Far from innocent, this is so-called "address poisoning", where small transfers are made to addresses similar to those which have interacted recently with active addresses, in the hope that the victims accidentally copy-paste the similar address from a block explorer or equivalent when they go to make another transfer. This is always there in the background at some level, but made up half of transactions on Sunday.
The weekend blitz was driven by a significant increase in direct ETH dust transfers, while batched transfers of ERC20 tokens made up the greater share of malicious transfers during the week (in a smaller number of transactions).
This will often take some time to pay off, but this week five wallets did pay poisoned addresses across six payments totalling 1.21 ETH. That is a floor: it only counts same-day ETH payments, and poisonings typically pay off weeks later. Stay safe out there folks (and block explorers: help people out!)
The six payments: 0.080 and 0.010 ETH (Wed, same victim twice), 0.005 (Fri), then Sunday's 0.597, 0.512 and 0.006 ETH.
Depegs and arbitrages
From millions of tiny transactions, we will finally pay attention to just three blocks which generated significant MEV, employing three different mechanisms to do so.
Who ended up with the money depends on the mechanism: the exclusive opportunity let the builder keep 77%, the contested ones pushed ~74–89% through to the proposer.
The msUSD depeg liquidation
The first was a liquidation on Thursday morning, as a particularly large looping position between frxUSD and Metronome's msUSD on Morpho was unwound during the msUSD depeg. You can read more about the overall depeg in this analysis from Webacy. The transaction of interest made 11.09 ETH (of which the builder Ultrasound kept 8.5), as the 0.97 price (caused by some early exiters) enabled liquidation of an overleveraged position well before the postmortem was published, and the rest of the pool emptied as the price dropped to $0.70.
The other two were arbitrage blocks. Ultrasound struck again, later on Thursday, landing a GALA / ZIK & SUSHI / ZIK arbitrage across stale pools in block 25,646,867 (sharing 89% of the 3.84 ETH with the proposer).
The GALA arbitrage: harvesting stale pools
And on Saturday Titan landed 79 CEX-DEX arbs in block 25,661,826, after a significant 1.3% ETH price drop. All 79 came from a single operation across three coordinated contracts (a WETH book working the Uniswap V2/V3 pools, a native ETH leg for V4, and a Curve leg, each funded by the book mid-transaction). Overall they sold ~$3M of ETH into onchain pools (presumably buying offchain for an overall profit).
Never a dull week in the dark forest.
Conclusion
So as Ethereum turned eleven, there was plenty to pay attention to at a high level and in the details, even with this relatively cursory look. With new primitives landing, and more and more happening on the L1 itself, there will be all the more reasons to keep checking the chain.
This is an exploratory analysis, feedback and corrections are very welcome: please reach out at adam@azfuller.com.
Methodology
How the numbers were made: sources, attribution methods, and caveats
Every claim above comes from one of the frozen datasets in data/. The pipeline in pipelines/ethereum-birthday-week/ produces them. The methods, by section:
Scope and core data. The dataset covers every Ethereum mainnet block with a timestamp in 2026-07-27 to 2026-08-02 UTC: 50,224 blocks and 12.9M transactions, fetched with full transactions and receipts from an archive RPC. Winning-bid data merges two independent sources: the CC-0 bulk archive at dataalways/mevboost-data (ten relays) and our own polling of six relays' proposer_payload_delivered APIs. The two sources agree on all 31,595 blocks they share, with zero disagreements (compared at 1-gwei tolerance, since the bulk archive stores rounded wei). Whether a transaction was ever public comes from Flashbots Mempool Dumpster. Contract names come from Etherscan, checked against Blockscout where a label mattered.
Builders and proposers. Builders are identified by the block's extraData tag, confirmed by the relay archives' delivered-payload records (builder label and pubkey). Both sources agree on every block cited. Builder profit uses the change in the fee recipient's balance. This method mismeasures builders that pay bids from a separate address or settle off-chain; sampled stress-tests show the major builders are clean and the long tail is approximate. Proposers are attributed by grouping validators that share a withdrawal address (credentials from the beacon chain for all 37,019 distinct proposers of the week — both 0x01 and 0x02 compounding credentials embed the operator's execution address) and spreading the relay archives' partial labels across each group. Nothing is guessed, coverage reaches 44% of blocks, and all shares are minimums. Payment comparisons between operators use medians, because one lucky high-value block can move a small operator's average.
| builder | blocks | share | rewards (ETH) | reward share |
|---|---|---|---|---|
| Titan | 24,227 | 48.2% | 381.0 | 54.1% |
| Quasar | 9,566 | 19.0% | 87.7 | 12.5% |
| BuilderNet | 5,930 | 11.8% | 102.7 | 14.6% |
| Eureka | 4,506 | 9.0% | 50.0 | 7.1% |
| proposer entity | share of blocks (floor) | bids received (ETH) | median bid/block |
|---|---|---|---|
| Lido (33 operators) | 19.1% | 140.4 | 0.0073 |
| Binance | 8.0% | 56.7 | 0.0072 |
| ether.fi | 3.3% | 24.5 | 0.0071 |
| Blockdaemon | 1.8% | 12.8 | 0.0079 |
| Everstake | 1.7% | 12.1 | 0.0081 |
| Coinbase | 1.6% | 14.6 | 0.0075 |
| unattributed | 55.8% | - | - |
The week in context. The 12-week base-fee baseline samples eth_feeHistory in three windows of 1,024 blocks per day. That covers about 43% of blocks, so spikes between windows can be missed, and records are stated as "highest sampled". Historical transaction counts, active addresses and fee totals come from growthepie. Percentiles for each day of the week are computed against the prior 77 days.
| day | median base fee (gwei) | percentile vs prior 77 days |
|---|---|---|
| Mon 07-27 | 0.0877 | 16.9 |
| Tue 07-28 | 0.0901 | 16.9 |
| Wed 07-29 | 0.0684 | 5.2 |
| Thu 07-30 | 0.0972 | 22.1 |
| Fri 07-31 | 0.1351 | 53.2 |
| Sat 08-01 | 0.0493 | 1.3 |
| Sun 08-02 | 0.0414 | 1.3 |
| prior 77-day median | 0.1325 | - |
Rollups and blobs. Blob counts and fees are read from the node: eth_feeHistory gives baseFeePerBlobGas, cross-checked against the blobGasPrice on receipts. (Recomputing the fee from excessBlobGas needs fork constants that have changed several times, and stale constants silently produce garbage.) Rollup spend is attributed by the inbox address each blob-carrying transaction is sent to. OP Stack chain IDs are decoded from the 0xff00… inbox address; the digits are decimal, not hex, and misreading them turns Base into a different real chain. Aztec posts through Multicall3, so its transactions are resolved by decoding the inner call. Three inboxes that no explorer labels were identified by matching their daily blob counts against growthepie's per-chain volumes; the worst daily gap on a match is 1.7%. On chains both sources track, our counts agree with growthepie within 0.2% to 3.3%.
| rollup | blobs | blob ETH | exec ETH | total ETH |
|---|---|---|---|---|
| Robinhood | 53,982 | 0.122 | 3.338 | 3.460 |
| Aztec | 8,203 | 0.014 | 1.118 | 1.132 |
| Base | 52,924 | 0.114 | 0.514 | 0.628 |
| Arbitrum One | 14,331 | 0.033 | 0.293 | 0.326 |
| World Chain | 12,849 | 0.026 | 0.281 | 0.307 |
| OP Mainnet | 19,312 | 0.026 | 0.096 | 0.122 |
| all blob posters (week) | 219,158 | 0.45 | 7.52 | 7.96 |
The two mints. Mint transactions are the set that emitted a Transfer from the zero address on each collection. Fees are computed per transaction as gasUsed × baseFee + tip from stored receipts, and verified against receipts sampled on-chain. Both galleries render from the contracts' own onchain tokenURI SVGs. For 11x11, the colour palette was recovered by lining up one 11-colour artwork's pixel indexes against its SVG, with zero conflicts, and artwork numbers map to token ids through mint-order Transfer logs, verified against each token's name field (33 of 33 matched). The fee event was attributed by decoding the nftContract argument of sampled SeaDrop.mintPublic calls in the peak window: 37 of 40 pointed at Compas.
| tokens | mint txs | gas | burned (ETH) | tips (ETH) | total fees (ETH) | |
|---|---|---|---|---|---|---|
| 11x11 | 35,245 (+11,639 editions) | 34,922 | 7.98B | 1.71 | 1.23 | 2.93 |
| Compas | 9,980 of 10,000 | 4,544 | 14.10B | 35.06 | 7.33 | 42.39 |
FWA and the dust campaign. Transactions are classified by a curated contract map, with spam patterns checked first. FWA's protocol figure (6.1% of gas) spans its five contracts. Its activity curve counts AcquisitionRequested and ListingStaged events only, so it tracks pack demand rather than gas. The dust campaign is fingerprinted by transfers of exactly 1 gwei and by lookalike address matching: the poison address shares the first and last four hex characters with a counterparty the victim recently paid. Batch-dispatched transfers are counted exactly, from every receipt of the three batch contracts. Payoffs are counted only when real funds later flow from the dusted wallet to the poison address, in genuine tokens (counterfeit lookalike tokens fake $4.5M of payoffs), on the same day, in ETH. The 1.21 ETH from five victims is therefore a minimum.
| poisoning arm | txs | transfers | gas cost (ETH) |
|---|---|---|---|
| direct blitz (1-gwei dust) | 1,790,082 | 1,790,082 | 3.478 |
| batch: Poisoner | 29,376 | 524,212 | 3.530 |
| batch: MockB | 7,998 | 1,134,024 | - |
| batch: batcher-2 | 22,275 | 1,336,064 | - |
| week total | 1,849,731 | 4,784,382 | - |
| payoff | ETH | day |
|---|---|---|
| victim 1, first payment | 0.0800 | Wed 07-29 |
| victim 1, second payment | 0.0100 | Wed 07-29 |
| victim 2 | 0.0053 | Fri 07-31 |
| victim 3 | 0.5966 | Sun 08-02 |
| victim 4 | 0.5124 | Sun 08-02 |
| victim 5 | 0.0062 | Sun 08-02 |
| total | 1.2105 |
The depeg and the three MEV blocks. The pool drain is sampled reserve state on the Curve frxUSD/msUSD pool, and the disclosure time is the post-mortem's publication time. Block anatomies use trace_block: payments to the fee recipient, ETH flows per address, and token flows netted per participant. The liquidation is decoded from Morpho Blue's Liquidate event, and the borrower's history (opened 3 July, about 25 loops, eight liquidations on the 30th) is rebuilt from their full event log on that market. The arbitrage classifications rest on measurable fingerprints. The GALA searcher ends the block with +4.43 WETH and exactly zero in every other token, a closed onchain loop through two pools that trade about 17 times a day. The Saturday arbs are one operation across three contracts. One holds the book, a second trades native ETH on Uniswap V4, and a third handles the Curve pools. The two satellites are funded with the book's WETH mid-transaction and deliver their proceeds straight back to it. The operation ends the block having net-sold 1,723 WETH. The proceeds, mostly stables plus a little WBTC, accumulate on the book with no closing leg onchain, and that inventory only makes sense if the offsetting buy happens on an exchange. All three blocks' MEV transactions are absent from the public mempool, while most other transactions in the same blocks are present. One honesty note: only winning bids are stored, so builder-level competition is inferred from how much value passes through to proposers, not observed in losing bids.
| block | MEV paid in (ETH) | bid to proposer | builder kept | searcher net |
|---|---|---|---|---|
| 25,643,399: depeg liquidation | 11.09 | 2.566 | 8.527 | 0.0012 ETH |
| 25,646,867: GALA stale-pool arb | 3.84 | 3.426 | 0.413 | +1.17 WETH of 4.23 extracted |
| 25,661,826: CEX-DEX cascade | 3.42 | 2.531 | 0.888 | open inventory, hedged off-chain |
Reproduction. pipelines/ethereum-birthday-week/README.md lists the run order. Every chart reads only the frozen JSON in data/, so the page rebuilds without network access.