ChainFeeds Research’s PRO #162 brings together a wide set of updates across Ethereum market structure, Bitcoin Lightning security, MEV research, and academic work on DeFi transaction classification.

How PropAMMs operate on Ethereum
Ethereum researchers Mike Neuder and Maryam Bahrani describe PropAMMs as a different model from traditional AMMs that passively quote prices through fixed formulas. In a PropAMM, professional market makers continuously update prices and liquidity. According to the report, PropAMMs now account for about 5% to 10% of swap volume on Ethereum mainnet. On Solana, leading PropAMMs are already close to 40% of total volume across major trading venues.
The main challenge on Ethereum is the 12-second slot. A maker’s quote can become stale before the block is built, leaving it open to arbitrage. The report frames the problem around three constraints: quote update latency, transaction ordering, and the cost of writing updates on-chain.
Titan’s approach relies on three guarantees from builders. The first is Maker Priority: if a transaction touches a PropAMM, the maker’s latest quote update must be executed first. The second is Maker Freshness: a taker trade must trail the maker’s latest update by at least about 50 ms, creating a brief last-look window for the maker. The third is Conditional Inclusion: makers do not need to post every quote change on-chain, and the builder inserts the latest update into the block only when a trade actually uses that PropAMM.
That setup is meant to avoid the need for millions of quote updates per day, as seen on Solana, while still keeping prices fresh enough during Ethereum’s longer slot time. The tradeoff is trust. Makers must rely on builders to follow these ordering rules because the guarantees are not enforced by the protocol today.
The report says the effects of wider PropAMM adoption may go beyond execution efficiency. Since market makers need direct connectivity to builders and must trust them, barriers to entry for block builders could rise. Execution quality may also vary across builders. At the same time, if users and liquidity move from traditional DEXs to PropAMMs, less CEX-DEX and DEX-DEX arbitrage could reduce MEV income for validators, while builders could move from highly competitive MEV auctions toward steadier fee models tied to trading volume.
Bitcoin protocol updates: PQLN, Eclair bugs, and LN-GAP
PQLN targets Lightning’s off-chain communications
Ahmet Kurt developed PQLN as an attempt to protect Lightning Network communications before Bitcoin consensus itself receives a post-quantum upgrade. The report notes that keys used in funding, commitment, and penalty transactions still depend on a Bitcoin consensus-layer upgrade for full post-quantum security. Gossip, node-to-node communication, invoices, offers, and onion routing are different. Those are messages exchanged between nodes, so they can migrate earlier through software upgrades.
PQLN has already been implemented in rust-lightning, covering BOLT 4, 7, 8, 11, and 12, and it uses post-quantum algorithms including ML-DSA and ML-KEM.
The biggest issue is not computation but data size. In testing, the slowest ML-DSA signature took only about 0.33 ms, but post-quantum public keys, signatures, and ciphertexts are much larger. In practical tests, a PQLN node had to download about 10 times as much gossip data as a standard node and store about 9 times as much data. Using the smaller FN-DSA can reduce that expansion to about 4 times.
PQLN also supports gradual migration. If every node on a route supports PQ mode, the route uses post-quantum protection. If a standard node appears, the system falls back to the traditional mode by default, though a require-PQ setting can reject that downgrade.
Two Eclair DoS bugs were fixed in v0.14.1
Bitcoin Optech Newsletter #427 says Eclair v0.14.0 and earlier versions had two denial-of-service vulnerabilities in the channel opening flow. An attacker could exploit a race condition involving temporary_channel_id or bypass the limit on pending channels, steadily consuming node memory.
In testing, one of the attacks leaked about 1 MB of memory per second. The other could exhaust 4 GB of memory in about 48 minutes without any on-chain cost. The issues were fixed in Eclair v0.14.1.
LN-GAP explores complex two-party contracts inside Lightning channels
Adam Gibson proposed LN-GAP as a way to run complex, multi-round two-party contracts inside a Lightning channel while keeping dispute resolution to a fixed number of on-chain transactions. The report uses chess as an example. Under normal conditions, every move stays off-chain. If one party stops responding, a committee is introduced only to prove whether a given message was published before a deadline, and the on-chain contract then resolves the dispute according to the rules.
Ethereum research: priority fees and the value of block position
Engineer Dan Guo used a tool called TEMBRI to study transaction placement, priority fees, and transaction types inside Ethereum blocks. The question was simple: how much extra do transactions near the front of a block actually pay?
The dataset covers 16 sample days from July 2025 to October 2026 and about 115,000 blocks. The results were described as highly stable. Swaps in the first 10% of a block by gas consistently paid much higher average priority fees than swaps in the middle of the block, typically 8 to 17 times the level seen in the 30% to 60% range. The sharpest fee drop happened within the first 20% of the block. The report says this points to clear position value at the front of Ethereum blocks.
Priority fees, however, do not fully explain ordering. In August 2026, about half of swaps in the first 5% of block gas had a priority fee of 0. By October, roughly one-eighth of those front-of-block swaps were still zero-tip trades. Titan-built blocks contained many such transactions, suggesting that some trades reached the front through private order flow, bundles, direct payments, or internal builder ordering.
Ordering patterns also differed across builders. Quasar, BuilderNet, and Titan showed a stable negative relationship between priority fee and block position, while locally built blocks came closest to simple fee ordering. The report’s conclusion is that Ethereum does not currently operate under a single rule where higher tips always buy earlier placement.
Speed versus screening in market making
Community contributor nuconstruct studied two main defenses against stale quotes. One is faster quote updates, which reduce the time a price is exposed. The other is screening, where a market maker identifies and rejects order flow that is more likely to be informed or toxic. The study asks how much informed flow a maker updating every 12 seconds would need to filter out in order to match the break-even spread of a venue that updates every second and does no screening.
The calibration used Binance ETH-USDT prices from June through August 2026 and about 1.72 million swaps across four USDC-WETH Uniswap V2 and V3 pools on Ethereum.
The result is that speed matters far more than a simple Gaussian model would suggest. Real one-second price changes are highly non-Gaussian: about 73% of one-second intervals show no price move at all, but when the price does move, the jump can be large. Cutting the quote interval from 12 seconds to 1 second reduces expected price impact by about 84.5%.
Under the study’s estimates, a maker updating every 12 seconds would need to filter out at least about 84.5% to 85.2% of informed flow to match a one-second market with no screening. Put differently, it would need to block about 6 out of every 7 toxic orders. A traditional Gaussian model would estimate only about 71.1%, materially understating the value of speed.
Screening still helps, but the paper says it is hard to treat it as a full substitute for speed. If two markets update at the same pace, filtering out one-quarter of informed flow can still let a maker quote a tighter spread. But when update speeds differ by a factor of 12, screening alone requires extremely high detection accuracy. The study also finds that in clear trend conditions, trend-following trades receive execution prices that are better by about 11.1 bps on average than trend-fading trades, indicating that stale quotes are actively picked off by participants in faster markets. For PropAMMs and a future Ethereum with shorter slots, reducing quote staleness is itself a core part of market quality.
Selected MEV-related items
The MEV Letter #157 was summarized with several highlighted pieces:
- One Node, Two Roles: Simultaneous Contests for Validation and Attention in Rollups studies how Optimistic Rollups can reward validators and attention nodes for independently executing state transitions, instead of allowing one operator to collect rewards through multiple identities.
- Ethereum PropAMMs: today and tomorrow explains how Ethereum PropAMMs rely on trusted builders for Maker Priority, Freshness, and Conditional Inclusion, and examines how wider adoption could raise builder entry barriers and change MEV revenue distribution.
- Actionable Signals in Prediction Markets: Evaluating Public and Private Information through Systematic Trading looks at whether Polymarket history contains signals usable by automated agents. The summary says risk arbitrage was loss-making overall, though some signals, including large positions, still showed limited realizable returns.
- The unleashed power of stacking EEZ & FCR seen through an example shows how combining EEZ’s Rollup-L1 atomic execution with FCR’s conditional fast confirmation could let L2 users complete payment flows involving L1 state in about two slots.
- Informed Diffusion: making home builders publish fast proposes a block propagation mechanism that helps home builders with limited upload bandwidth publish blocks faster by using known block structure, narrowing the latency gap with professional builders.
- Rapidblocks, take two: a pipelined consensus for independent blocks proposes a pipelined consensus design for 2-second Ethereum slots, allowing production of a later block to overlap with validation of the previous one rather than forcing every step into sequence.
- Actually, AI makes crypto better. Ethlabs Week 16. updates readers on AI-assisted protocol research, FCR adoption, and progress after Quick Slots entered Hegotá Considered for Inclusion.
- The cryptographic world computer is a website that visualizes Vitalik’s writing on the “Cryptographic World Computer,” showing how Ethereum combines on-chain consensus, cryptographic verification, and decentralized off-chain computation into one computing layer.
Paper: classifying DeFi methods with graphlets
In the paper section, ChainFeeds highlights Graphlet-based Decentralized Finance Transaction Method Classification from Duke University. The authors ask whether the internal network structure of a transaction can reveal what kind of DeFi action it performed.
The study converts interactions among accounts, smart contracts, and tokens into directed graphlet features and uses them to classify the 30 most common DeFi methods. According to the summary, graphlets improve on traditional event-log counts and raise classification accuracy. With SHAP, the approach can also identify structural patterns specific to different methods, offering another on-chain analysis tool for recognizing complex DeFi behavior.


