The House Edge

Appearance

Outline: The House Edge ​

The full table of contents. The agent works against this. Chapter numbers are stable; section headings inside each chapter are determined at Phase 2.

Prologue — A Dollar Walks Into a DEX ​

The book opens on a single $10,000 trade. We meet our protagonist (a fictional but plausible business reader making her first significant crypto swap), watch the trade complete, and surface the question the rest of the book answers: where did the $73 of slippage actually go, and to whom? Sets up the follow-the-money spine.

Part I — The Game Board ​

Chapter 1 — What Is a Trade, On-Chain ​

Three phases that happen to every trade, and what the chain does to each one.

Every trade has the same three phases: intent, settlement, finality. TradFi parallels for each. What "on-chain" means versus "on a centralized exchange," and why the distinction matters less than people think. First introduction to the public mempool. No adversarial content yet — the reader needs the board before they see the moves.

Final section headings (Phase 3, 2026-05-13):

  • Cold open — NYSE T+1 transition (28 May 2024) as the TradFi anchor; "ten years to shorten settlement by one day" vs "seconds to final on-chain"
  • What this chapter answers
  • The setup — the three phases as a universal concept; Alice plant
  • The worked example — Alice runs the same $10K USDC→SOL swap twice (Coinbase Advanced Trade, then Solana via Jupiter)
  • The mechanics, in detail
    • Intent — visibility differences across venues (Coinbase invisible; Solana Gulf Stream limited; Ethereum public mempool; Hyperliquid none)
    • Settlement — explicit fees (Coinbase) vs gas + priority + slippage (Solana); PFOF aside citing SEC DERA 2025 paper (crypto PFOF 45× equities)
    • Finality — the curve from "venue says done" to "no one can roll it back"; Solana 0.4/0.6/12.8s; Alpenglow shipping ~150ms; Ethereum 12.8min; Hyperliquid ~200ms; L2 7-day fraud-proof window
  • How this plays out on each chain — Solana / Hyperliquid / Ethereum and L2s
  • Who wins, who loses, why — structural verdict (light touch, the chapter is setup)
  • What changes when… — transition to Chapter 3 (mempool extraction)
  • Footnotes and sources (17 numbered citations)

Chapter 2 — Where Liquidity Lives ​

Two architectures, four actors, and the quiet displacement of the passive liquidity provider.

The two architectures: automated market makers (AMMs) and central limit order books (CLOBs). Plus aggregators and RFQ as overlays. Why Solana converged on prop-AMMs routed through Jupiter, why Hyperliquid built a CLOB that worked, why Ethereum supports both. The economics of being a passive LP versus a market maker. First sidebar: Meet the Market Maker.

Final section headings (Phase 3, 2026-05-13):

  • Cold open — Phoenix's collapse: $3.7M Q2 2024 → $68,604 Q1 2026 (≈54× decline), and Ellipsis's pivot to perpetuals; ends with the chapter's argument statement
  • What this chapter answers
  • The setup — liquidity, AMM, CLOB, aggregators, RFQ, plus Bob's $50K LP plant
  • The worked example — Bob deposits into Uniswap V3 ETH/USDC 5-bps, Feb 2026
  • The mechanics, in detail
    • The AMM curve and the loss the LP doesn't see (xy=k, concentrated liquidity, Topaze 2021 finding, LVR framework, Bob's monthly P&L)
    • The CLOB, and why every on-chain attempt at spot has failed except Hyperliquid (with Meet the Market Maker sidebar — Wintermute, Flowdesk; the HyperBFT difference)
    • Aggregators, solvers, and the retail interface that replaced the pool (Jupiter 93.6% share; CoW Swap, UniswapX, 1inch Fusion; SCP and Wintermute filling >90% of UniswapX)
    • The prop-AMM displacement (HumidiFi reportedly Temporal, SolFi Ellipsis, Tessera; the >50%-of-Solana-spot finding)
  • How this plays out on each chain — Solana (prop-AMMs + Jupiter), Hyperliquid (native CLOB + HLP), Ethereum + L2s (both)
  • Who wins, who loses, why — the verdict (lighter-touch than Ch 4 since this chapter is structural)
  • What changes when… — transition to Chapter 3 (mempool dynamics, the visibility that lets prop-AMMs work)
  • Footnotes and sources (19 numbered citations)

Chapter 3 — The Mempool and What Replaces It ​

A surface, three responses, and the actor whose job is to read it.

What a public mempool is, why it's a billboard for predators, and how each chain has tried to address that. Jito's auction. Hyperliquid's no-mempool design. Private order flow on Ethereum (Flashbots Protect, MEV-Share). This chapter is the hinge — once the reader understands transaction visibility, the adversarial content lands.

Final section headings (Phase 3, 2026-05-14):

  • Cold open — Jito Labs' 8 March 2024 mempool shutdown (run rate ~10K SOL/day in tips, ~$500M+ annualised; ~10 months of broad searcher access plus ~6 prior months of validator-only operation; gone within hours; "negative externalities untenable")
  • What this chapter answers
  • The setup — public mempool defined; Alice and Carlos as the chapter's two traders
  • The worked example — Alice's $10,000 USDC→SOL swap, public RPC vs Jupiter Beam
  • The mechanics, in detail
    • Solana: the mempool that wasn't, the mempool that was, and what replaced both (Jito mempool history; March 2024 shutdown; June 2024 Solana Foundation removing ~32 validators; Beam at router; BAM at block construction; Harmonic at validator selection; client-stack stake shares; April 8 2026 patch; Constellation as forward-looking)
    • Ethereum: the public surface and its private replacement (Wahrstätter 54.6% of block value private; Carlos through Flashbots Protect; Flashbots Protect, MEV-Blocker, MEV-Share; MEV-Boost concentration with Titan 50.6%, BuilderNet, Quasar = ~91%)
    • L2s: the sequencer as a one-firm private mempool (Base / Arbitrum Timeboost / Optimism Unichain; Espresso & Astria as alternatives; no major L2 decentralised yet)
  • How this plays out on each chain — Solana / Hyperliquid / Ethereum + L2s (Hyperliquid as architectural elimination)
  • Who wins, who loses, why — adversarial verdict; "the volume has contracted; the structure has not"
  • What changes when… — transition to Chapter 7 (exclusive flow; relationship between bundle-submitter and block-decider)
  • Footnotes and sources (24 numbered citations)

Part II — The Actors ​

Chapter 4 — The Searcher ​

Why someone is always watching your trade — and what they get paid for it.

The first adversarial actor. What a searcher does: arbitrage, liquidations, sandwiches. The three flavors of MEV with worked examples for each. Why arbitrage is morally neutral, liquidations are useful, and sandwiches are pure extraction. Sidebar: Meet the Searcher.

Final section headings (Phase 3, 2026-05-13):

  • Cold open — the Vpe scene (Dec 2024 – Jan 2025; $13.4M extracted by a single program)
  • What this chapter answers
  • The setup — defining MEV in plain language; introducing the searcher; the HFT analogue
  • The worked example — Alice's $10,000 USDC→SOL Jupiter swap, Feb 2025
  • The mechanics, in detail
    • Arbitrage: closing a gap (2Fast cameo; CEX–DEX as the darker variant)
    • Liquidations: forced repayment (the March 10, 2026 Aave wstETH event)
    • Sandwiches: the pure-extraction case (Alice resumes; D1 sandwich sequence; D2 dollar split; JIT footnote; Jupiter MEV-Protect and the April 8, 2026 Solana patch)
    • Why this is a real business (Paradigm's theoretical ceiling; Flashbots' Base economics; consolidation; the Vpe P&L table)
  • How this plays out on each chain — Solana, Hyperliquid, Ethereum and L2s
  • Who wins, who loses, why — the verdict
  • What changes when… — transition to Chapter 5 (where Vpe's $4.6M in tips went)
  • Footnotes and sources (15 numbered citations)

Chapter 5 — The Validator (and the Builder) ​

The role that was supposed to be neutral plumbing — and the firms that took the rest of the job.

Validators were supposed to be neutral plumbing. They aren't. How Ethereum's proposer/builder separation reshaped the role. How Solana's leader schedule and Jito's auction create the same dynamic via a different architecture. How Hyperliquid sidesteps it. Sidebar: Meet the Validator and Meet the Builder.

Final section headings (Phase 3, 2026-05-14):

  • Cold open — The Merge (15 Sep 2022) → MEV-Boost launch (7 Nov 2022) → >90% adoption within 9 months
  • What this chapter answers
  • The setup — what a validator is; the bifurcation across three chains; Coinbase Cloud + Titan as the chapter's two characters
  • The worked example — a single Ethereum block: searcher → Titan → Ultrasound Relay → Coinbase Cloud → staker (Mermaid sequence diagram inline)
  • The mechanics, in detail
    • Ethereum: the proposer who doesn't build the block (PBS history; Titan 51.5% / BuilderNet 24% / Quasar 15.3% concentration; ePBS slipping past Glamsterdam; Meet the Builder sidebar)
    • Solana: the client stack is the moat (Agave / Jito-Solana / JitoBAM / Frankendancer / Firedancer at 32/28/17/12/2% stake; perf vs balanced vs revenue scheduler modes; Harmonic Performance/Balanced configurations at +101%/+39%/+36% priority fees; Figment Frankendancer migration as empirical anchor; Timely Vote Credits and Chorus One's 2.03% skip rate; Alpenglow (98.27% approval; testnet 11 May 2026; VAT economics drop profitable-validator threshold tenfold); Meet the Validator sidebar; client+scheduler matrix table)
    • Hyperliquid: the permissioned set that sidesteps PBS entirely (top-21 active set; 10K HYPE self-delegation; HLP separate from validator stack)
  • How this plays out on each chain — Solana / Hyperliquid / Ethereum + L2s (with the slashing record as operational-excellence anchor)
  • Who wins, who loses, why — block-construction firms, validator-client developers, institutional validators win; small validators without builder access or sophisticated client choice lose; Coinbase Cloud's optional OFAC filtering as the chapter's closing structural observation
  • What changes when… — transition to Chapter 7 (the four-actor flow collapses to one actor)
  • Footnotes and sources (22 numbered citations)

Chapter 6 — The Infrastructure Layer ​

The firms you've never heard of that route most of the order flow — and what they sell, to whom, and why.

The companies you've never heard of that route most of the order flow. Jito, bloXroute, Astralane, Eden, Temporal/Nozomi/Harmonic on Solana; Flashbots, Titan, beaverbuild on Ethereum. What they sell, who buys it, and why their existence is the most under-appreciated fact about on-chain trading. Sidebar: Meet the Infrastructure Provider.

Final section headings (Phase 3, 2026-05-14):

  • Cold open — Five firms touched one trade (Phantom, Helius, Jupiter, Jito Labs, Jito-Solana; possibly BAM as a sixth)
  • What this chapter answers
  • The setup — the four functional layers (RPC, builder, relay, OFA); paired Helius/Flashbots anchor; D1 stack flowchart inline
  • The worked example — Helius and Flashbots full firm profiles
  • The mechanics, in detail
    • The RPC providers (Helius full + Triton + Astralane; the 50/50 wallet-rebate split as the PFOF analogue)
    • The block builders and bundle markets (Titan 51.55% + BuilderNet 24.09% + Quasar 15.29% with unknown founders on Ethereum; Jito Labs full treatment including the JIP-24 transition; Temporal + Paladin; the Ben Coverston Temporal-Harmonic CEO-overlap paragraph)
    • The relays and routing infrastructure (bloXroute full treatment; Ultrasound, Titan Relay, Aestus as non-profit, Flashbots Relay at 2.15%)
    • The Order Flow Auction operators and refund services (Flashbots Protect, MEV-Share, SUAVE archival, MEV-Blocker transferred to Consensys SMG January 2026, Kolibrio, CoW DAO)
  • Meet the Infrastructure Provider sidebar — instantiated against Helius
  • How this plays out on each chain — Solana (operator-concentration across layers), Hyperliquid (no significant third-party infrastructure layer), Ethereum (layer-concentration with different firms at each function)
  • Who wins, who loses, why — D2 firm-by-layer table; aggregate take rate estimate; Eden Network as defunct firm; Beaverbuild as migrated-into-BuilderNet
  • What changes when… — transition to Chapter 7 (exclusive flow arrangements; Titan + Banana Gun as seed)
  • Footnotes and sources (22 numbered citations)

Chapter 7 — Exclusive Order Flow ​

The contracts behind the dollar — and the reason a top-quintile validator is no longer just a better-run one.

The chapter that pulls together the previous three. What happens when a single infrastructure provider has exclusive access to a slice of order flow, and sells block-construction privileges to a single validator. The TradFi parallel: payment for order flow at Robinhood and Citadel. Why this is the most important undiscussed dynamic in validator economics. The chapter contains the book's strongest claim: the difference between a top-quintile and bottom-quintile validator on most chains is no longer operational excellence — it is access to exclusive flow.

Final section headings (Phase 3, 2026-05-14):

  • Cold open — Banana Gun / Titan dollar trace: 4,466.89 ETH paid by users → 2,915.65 ETH to validators / 2,271.26 ETH retained exclusively by Titan; Titan <1% → ~40% → ~52% builder share trajectory under one contract
  • What this chapter answers
  • The setup — public order flow vs exclusive order flow on a permissionless chain; the three structural shapes (direct exclusives, pipeline rebates, validator-side fragmented surfaces); the chapter's load-bearing access-vs-operational claim previewed
  • The worked example — Banana Gun + Titan developed in full (the four-actor pipeline: Banana Gun's user → Banana Gun → Titan → the proposer; who pays for the 17.75% margin)
  • The mechanics, in detail
    • Direct exclusive arrangements: the Ethereum builder–trading-firm model (Banana Gun / Titan, Maestro / Beaverbuild; Wu et al.'s 75 EOF arrangements / ~71% of trading-related builder revenue; Pahari & Canidio's 77.2%–84% fee-share for exclusive transactions; Mermaid flowchart for two named exclusive pairs inline)
    • Pipeline relationships: the Helius / wallet-rebate model (50/50 split; Phantom, Backpack, Solflare, Trust, Bitwise, Coinbase, Jupiter as named wallet customers; the structural opacity around downstream rebate flow)
    • Validator-side fragmented surfaces: the persistent grey market (DeezNode + Vpe ~$13.43M sandwich profit reprise; Tim Garcia's "enforcement is ongoing" framing; demand surface fragmented into RPC-layer, BAM-layer, and less-public validator arrangements)
    • The validator-quintile decomposition: the empirical argument (Syndica's +33% to +101% access signal; Chorus One's ~+3% operational signal; Placeholder VC's Gini ~0.93; Figment Q4 2025 SRR 6.44% vs 4.70% network average; inline matrix table comparing top/median/bottom-quintile validators)
    • The TradFi parallel: payment for order flow in 2026 (SEC DERA paper concentration data; EU PFOF ban effective 30 June 2026 with only Germany having exercised the transitional exemption; SEC's 12 June 2025 withdrawal of the Order Competition Rule and Regulation Best Execution)
  • How this plays out on each chain — Solana (mix of pipeline and validator-side surfaces; no Ethereum-style builder-direct equivalent); Hyperliquid (architecture eliminates the surface; HIP-3 builder codes operate at a different layer; Bitwise's approval-not-exclusivity ETF counterparties); Ethereum (the chain where exclusive flow reached its most mature form; BuilderNet as the explicit non-exclusive structural counter-attempt)
  • Who wins, who loses, why — Titan and the builder-trading-firm pairs; Helius and the named wallet customers; Jito Labs (BAM stack as analogous exclusive surface with JIP-24 routing tip revenue to the DAO); SCP and Barter at the solver layer; Lido curated operator set; bottom-quintile validators as the primary loser; Banana Gun's users as the named-loser case; the public-goods narrative as the structural loser
  • What changes when… — Part II ends; Part III (chain-specific architectures) opens
  • Footnotes and sources (15 numbered citations)

Part III — The Architectures ​

Chapter 8 — Solana: AMMs, Auctions, and the Speed Game ​

The chain where the validator IS the slot leader IS the block constructor — and what that produces when you wire eight cities and three software stacks into one market.

Deep on Solana's market structure. Why aggregators (Jupiter) won spot, why no CLOB has won trading, what the Jito auction actually auctions, what the Frankendancer/Firedancer/Rakurai client diversity does and doesn't change. The under-explained role of geographic proximity to block engines. Where retail loses, where MMs win, where validators with exclusive flow win the most.

Final section headings (Phase 3, 2026-05-14):

  • Cold open — Solana's 19 Jan 2025 $56.8M single-day REV anchor; 2025 annual $1.4B → Q1 2026 $89.5M (−68% YoY); Solana #2 by REV behind Hyperliquid for the first time
  • What this chapter answers
  • The setup — Solana's validator = slot leader = block constructor architecture; the three stratified findings (retail loses less in 2026 than 2024; MMs won spot via prop-AMM displacement; validators with infrastructure relationships have the largest dollar magnitude per actor)
  • The worked example — Alice's $10,000 USDC → SOL swap through the full Solana 2026 stack (Phantom → Helius → Jupiter → Beam → Jito Frankfurt → Frankendancer Harmonic Performance in Equinix FR5); per-layer dollar trace lands at ~$8–$15 cumulative take vs ~$73 of 2024 formal-sandwich slippage; D1 Mermaid flowchart inline
  • The mechanics, in detail
    • Aggregators ate the interface — and Jupiter took the aggregator layer (Jupiter ~93.6% aggregator share; ~74% of total DEX volume aggregator-routed; Ultra ~95% of aggregator volume + ~50% of total; Beam's +0.006% vs −0.14% Ultra V3 slippage; Beam public Transaction Submission API April 2026)
    • The founding vision, and the CLOB experiment that ended (Anatoly Yakovenko's 2017 NASDAQ-on-Solana whitepaper pitch; the Resnick + Yakovenko May 2025 Path to Decentralized NASDAQ essay admission that "five years in, we have not succeeded in that goal"; Phoenix Q2 2024 $3.7M → Q1 2026 $68,604; Manifest <2% of $284.5B Q1 2026 spot volume; March 2026 prop-AMM share ~55% / $36B of $65B; competitively contested; HumidiFi → Temporal / SolFi → Ellipsis Labs / Tessera V → Wintermute / BisonFi → Forward Industries; the new institutional cameo)
    • What the Jito auction actually auctions (three products in one: inclusion priority + atomic bundle execution + routing access; ~94% Jito-Solana-variant stake; late-2024 ~$2.5M/day tips + $14.7M single-day high; post-JIP-24 entire 6% routes to DAO; BAM launch 21 July 2025 with four launch operators; ~25% stake by Feb 2026; new sites Dublin + Dallas; open-source target mid-Q2 2026)
    • The client + geography matrix — Chapter 8's signature material (March 2026 client distribution Agave Jito 32% / JitoBAM 28% / Agave Harmonic 17% / Frankendancer 12% / Rakurai 6% / Firedancer 2%; none sandwich-default in 2026; Ben Coverston CEO of Temporal + Harmonic concentration fact; Rakurai's "no timing games" + Figment migration anchor; stake by city — Frankfurt 19% / Amsterdam 16% / London 12% / Vilnius 6% / Tokyo 4% / Ashburn 4%; Jito's eight mainnet block engines; Equinix FR5 + NY5 as anchor data centres; D2 Mermaid block-engine ↔ stake-city map inline)
    • SFDP, Alpenglow, and Constellation: the response from inside (May 2026 <25% ASN + <15% data-centre concentration rules; ~150 validators projected to lose Foundation stake; DeezNode forward-reference only; Alpenglow testnet 11 May 2026 + late Q3/early Q4 2026 mainnet; vote-tx removal + VAT + ~10× threshold drop; Constellation unveiled 5 May 2026 at Solana Accelerate as the multi-leader response to the access gap; ~16 concurrent proposers / 256 attesters / 50ms economic tick; preprocessor to Alpenglow; Q3 2026 mainnet target; full architectural treatment of both is Chapter 12)
  • How this plays out on each chain — Solana (vertical integration + access-vs-operational gap as software-and-geography function); Hyperliquid (no-mempool / in-consensus-matching, #1 Q1 2026 REV — full treatment Chapter 9); Ethereum (PBS separation + most mature exclusive-flow surface — full treatment Chapter 10)
  • Who wins, who loses, why — three-actor stratification: validators with infrastructure relationships > market makers (prop-AMM operators internalising spread) > retail (loses less in 2026 than 2024 but still pays cumulative stack take); the structural closing on five-firms-touching-one-trade as Solana's defining shape
  • What changes when… — transition to Chapter 9 (Hyperliquid as the structural counterpoint that eliminates the surface Solana could only compress)
  • Footnotes and sources (25 numbered citations)

Chapter 9 — Hyperliquid: The CLOB That Worked ​

The chain that eliminated the extraction surface — and where the value re-concentrated when it did.

Why Hyperliquid succeeded at what every Solana CLOB attempt failed at. The no-mempool design, the on-chain matching engine, the HLP. What HIP-3 is and why volume is migrating there. The trade-offs Hyperliquid made (less programmability, smaller ecosystem) to deliver one thing well. Where the extraction still happens — it's just concentrated differently.

Final section headings (Phase 3, 2026-05-14):

  • Cold open — BHYP launch 15 May 2026 + validator set 24→27 + Q1 2026 REV reversal (Hyperliquid $144.8M / Solana $89.5M); Jeff Yan + ~11-person team + zero outside VC; $619B Q1 2026 perp volume on HyperBFT
  • What this chapter answers
  • The setup — HyperBFT-consensus matching engine; HyperCore + HyperEVM dual-state; no mempool / no PBS / no relays; extraction surface eliminated but value re-concentrated into four named loci (HLP / validator set / HIP-3 deployers / HYPE holders via buyback machine); architectural trade-off as deliberate (less programmability, smaller ecosystem, single-product excellence)
  • The worked example — Alice's $10,000 long-BTC perpetual through the Hyperliquid 2026 stack (Circle CCTP → HyperCore match against HLP / Wintermute / Flowdesk / FalconX / Nonco → fee flow to HLP and Assistance Fund → exit); per-layer dollar trace lands at ~2–4 bps all-in vs Ch 8 Solana spot stack-take; D1 fee-flow Mermaid inline
  • The mechanics, in detail
    • The CLOB inside consensus — why Hyperliquid works where Phoenix didn't (visibility + latency twice-closed; ~200K orders/sec; dual-state HyperCore + HyperEVM)
    • The HLP — the chain's house ($383M TVL; ~$121.8M lifetime PnL through Oct 2025; JELLY incident 26 Mar 2025 as stress test; Meet the House sidebar developing HLP-as-actor distinct from MM/passive LP/exchange risk desk)
    • The validator set — what it does without a block construction layer (24 expanding to 27; slot #21 ~525K HYPE; 10K HYPE self-delegation; 2.37% APR inverse-square-root formula; named operators Imperator/Bharvest/P2P.org/Figment/Hyperliquid Strategies × Unit Labs/Foundation Nodes; slashing not yet automatic at consensus layer — jailing is; updates Ch 5 "permissionless-top-21" framing)
    • HIP-3 — the chain's programmability primitive, in motion (mainnet 13 Oct 2025; 500K HYPE deployment stake; 50/50 deployer/protocol fee split + 2× user fee; aggregate 30-day volume ~$2.82B / 20,348 traders; HIP-3 >35% of all Hyperliquid trading volume; TradeXYZ at >90% HIP-3 OI / $12.7B cumulative; permissioned-by-stake-not-by-permission framing)
    • HYPE tokenomics and the buyback machine (1B max supply / ~333M circulating; 99% of validator-perp fees → Assistance Fund; cumulative buybacks crossed $1B in 15 months; Dec 2025 ~37.5M HYPE / ~$912M burn; Jan 2026 ~13% circulating supply burn proposal; 3%/97% vs 99%/1% framing reconciled; structural breakeven by design; BHYP institutional access anchor)
  • How this plays out on each chain — Hyperliquid (eliminates the surface; trade-off as deliberate); Solana (compresses but doesn't eliminate; comparison via Ch 8); Ethereum (the third structural option with most mature PBS-and-exclusive-flow surface; full treatment Ch 10; Arbitrum bridge as hint at L2 thesis)
  • Who wins, who loses, why — four-actor stratification: HLP depositors (paid by surface they fund; tail-exposed); validator set (incumbent stakers); HIP-3 deployers (gated by 500K HYPE stake moat, comparable to Ethereum builder economies-of-scale); retail (pays less per trade than any comparable on-chain venue; pays it more visibly); "concentrated differently" verdict landing
  • What changes when… — transition to Chapter 10 (Ethereum + L2s; the third structural option; Arbitrum bridge being deprecated by Hyperliquid hints at L2 value-capture pattern)
  • Footnotes and sources (23 numbered citations)

Chapter 10 — Ethereum and Its L2s: The Most Studied, Least Settled ​

The chain where the extraction surface is most explicitly architected — and where the value capture dispersed along architectural lines but concentrated along operator ones.

Why Ethereum is the chain where MEV was named, measured, and partially industrialized. Proposer/builder separation, MEV-Boost, the relay layer. The L2 fragmentation problem (Arbitrum, Base, Optimism, Linea) and how it creates a new extraction surface: cross-chain MEV. Why Hyperliquid's Arbitrum bridge tells you something about where Ethereum's L2 thesis is headed.

Final section headings (Phase 3, 2026-05-14):

  • Cold open — Ethereum Q1 2026 chain REV ~$82M (#3 behind Hyperliquid $144.8M / Solana $89.5M); Titan ~52% / BuilderNet ~25% / Quasar ~15% builder share; Base ~62% of L2 fee revenue (~$67M/year, ~65% from 250 addresses); EigenLayer ~$18B / 4.6M ETH / ~1,900 operators; blob fee L2→L1 collapse from ~$113M (2024) to ~$10M (2025)
  • What this chapter answers
  • The setup — Ethereum's deliberate dispersion across PBS + L2 sequencers + restaking; dispersion real along architectural lines but not along operator lines (Coinbase / Consensys / Flashbots / Lido appear at multiple layers); three previewed findings (L1 fee economy contracted by design; value moved off L1 to L2 sequencers; cross-chain MEV is a new measured surface)
  • The worked example — Alice's $10K USDC → ETH swap on L1 (~$15–25 take through MetaMask → Flashbots Protect → CoW → SCP solver → Titan → Ultrasound → Lido validator) vs Base (~$2–5 take through Coinbase Wallet → Aerodrome → Base sequencer → blob settlement); D1 two-leg Mermaid flowchart inline
  • The mechanics, in detail
    • L1 in 2026 — the chain that moved its fee economy (validator count ~1.06–1.1M; staked ETH ~35.86M; L1 daily fees ~$6.3M down from ~$23M peak; ETH ~0.23%/year inflationary; blob payments $113M→$10M; Fusaka 3 Dec 2025 + two BPO forks lifting blob target 3→14)
    • PBS in 2026 — Titan, BuilderNet, Quasar, and the new entrant (live relayscan.io shares; Eureka Labs as new $6.7M-seed-funded ~2% builder; ePBS slipping to Q3/Q4 2026 with 2027 risk if FOCIL bundled; Justin Drake's 7-fork Strawmap; MEV-Boost as de facto enshrined PBS for ~4 years and counting)
    • The L2 sequencer economy — Base wins, Arbitrum monetizes, the rest catch up (L2 TVL ~$48B / ~73 rollups / Arb+Base+OP ~90% volume; every major L2 single-sequencer; Base ~62% of L2 fee revenue at Coinbase / ~$185K/day / 64.9% of priority fees from 250 addresses / only profitable L2; Arbitrum Timeboost as L2-EOF anchor / Castro et al. 2 firms win >90%; Espresso mainnet 12 Feb 2026 / decentralized sequencers not yet at scale)
    • Cross-chain MEV — the new extraction surface (Maire et al. 242,535 arbitrages / $868.64M volume / $10.05M revenue / 5.5× growth / 58.35% L1↔L2 / 35.67% L2↔L2 / top-5 addresses >50% of trades / one address ~40% post-Dencun; Gogol et al. >500K cross-rollup opportunities / 0.03–0.25% of volume / 10–20 block persistence; atomicity as structural difference)
    • Restaking — EigenLayer as the third extraction surface (~$18B TVL / 4.6M ETH / ~1,900 operators / ~94% market share; slashing live 17 April 2025; ELIP-006 Redistributable Slashing 2026; AVS roster EigenDA / AltLayer MACH / Hyperlane / Witness Chain; Vertical AVS trend; the cross-chain extraction surface restaking unlocks as Ethereum's trifecta third layer)
  • How this plays out on each chain — Ethereum (PBS + L2 + restaking trifecta; dispersion real architecturally not by operator; Coinbase/Consensys/Flashbots/Lido cross-layer); Solana (vertical integration compressed surface; access gap as software-and-geography function; Ch 8); Hyperliquid (eliminated surface; four-locus re-concentration; Ch 9); three-chain pattern: architecture choices produce different value-capture shapes but comparable operator concentration across all three chains
  • Who wins, who loses, why — four-layer stratification: L1 builder oligopoly (Titan / BuilderNet / Quasar 91% / Wu et al. 75 EOF / Eureka as bounded entry); L2 sequencer monopolies (Base / Arbitrum / OP / Linea with single-sequencer concentration); cross-chain searchers (top-5 addresses >50% per Maire); restaking operators (~1,900 across $18B); retail traders lose more visibly on L1 (~$15–25 on $10K) than on L2s (~$2–5) but L2 fee flow concentrates into single sequencer operators; closing structural observation — architecture produces legible map of operator concentration without producing operator competition
  • What changes when… — transition to Chapter 11 (Part IV opens; the four chronic-loser categories across all three chains)
  • Footnotes and sources (20 numbered citations)

Part IV — The Verdict ​

Chapter 11 — Who's at a Disadvantage, and Why They Don't Move ​

Four chronic-loser categories across three chain architectures — and the consent gap that keeps the table populated.

The honest chapter. The four chronic losers in on-chain trading: retail traders unaware of toxic flow, passive LPs in informed-flow pools, slow market makers, validators without infrastructure relationships. Why each group continues to participate anyway — and why "just leave the table" is harder than it sounds.

Final section headings (Phase 3, 2026-05-14):

  • Cold open — Pump.fun's 30.1% (June 2025) → 73.3% (April 2026) profitable-wallet share is survivorship not learning; 65% MAU exodus from 5.2M → 1.8M
  • What this chapter answers
  • The setup — participation ≠ consent; the four loser categories along two axes (loss magnitude + informational gap); three previewed findings (loser categories are not symmetric; each category's exit path is structurally constrained; the consent gap is the chapter's clinical core)
  • The worked examples — Carlos (Pump.fun chronic loser); Bob's V3 LP revisited; Erika the SFDP-removed Solana validator; slow MM developed at category level
  • The mechanics, in detail
    • Retail traders unaware of toxic flow (BIS Bulletin 69 73–81% retail loss / 47.89% average; Pump.fun survivorship pattern; Banana Gun 1.3M users / $635 avg trade / Titan ~50.8% capture; FCA 2025 UK ownership 12%→8% with disclosure reducing participation not per-trade losses; Carlos's UX abstraction)
    • Passive LPs in informed-flow pools (Milionis et al. LVR framework ~5–7% capital/year; Topaze Blue 49.5% net-negative V3 LPs / $60.8M worse than HODL; Uniswap Labs V3 5-bps tier underperformed V2 by 68%; HLP as structurally-different passive-LP-class with informed consent; JELLY stress test; toxic flow vs uninformed flow CrocSwap finding)
    • Slow market makers (Phoenix $3.7M → $69K 54× collapse as category-level displacement; Wu et al. 75 EOF / 71% of trading-related Ethereum builder revenue; +33–101% Harmonic Performance gap vs ~+3% operational; Wintermute mid-May 2026 90% Hyperliquid liquidity compression as winner-can-become-loser event; slow MMs as the consent-aware category)
    • Validators without infrastructure relationships (Solana validator count 2,560 → ~770 / ~150 SFDP removals May 2026; +33–101% access vs ~+3% operational decomposition; Gini ~0.93; Hyperliquid 525K HYPE slot-21 / 10K self-delegation 52× capital ratio; Ethereum solo home staker as winner-by-yield / loser-by-share)
  • The consent gap — the chapter's clinical core (D1 inline table: four categories × loss magnitude / Layer-1 visibility / Layer-2 visibility / consent type; uninformed retail vs partial-LP vs informed slow-MM and validator; regulatory frameworks address asset-exposure layer not trade-execution layer)
  • Why "just leave the table" is harder than it sounds (per-category exit path constraints; DEX-to-CEX ratio tripled 2021–2025 in opposite direction of "retail moves to safer venues"; the pivot-to-prop-AMM availability bounded to top 5–10 firms; the small validator's option-value framing)
  • What changes when… — transition to Chapter 12 (forward-looking; does the structure change?)
  • Footnotes and sources (23 numbered citations)

Chapter 12 — Where This Is Going ​

Three forward-looking trends that reshape where the extraction happens — and what they do not change.

The forward-looking chapter. Three trends that will reshape the game over the next 3–5 years: intent-based architectures (Anoma, CoW, UniswapX), appchains and sequencers (Hyperliquid's HIP-3, app-specific rollups, SVM appchains), and cross-chain execution layers. What each does to the existing extraction pattern. What it doesn't change.

Final section headings (Phase 3, 2026-05-14):

  • Cold open — HIP-3 OI $280M (Jan 2026) → $2.50B (May 2026) 9× growth with TradeXYZ >90%; EF Checkpoint #9 Glamsterdam "trickier than anticipated"; Astria mainnet sunset; Alpenglow live validator testing 11 May 2026
  • What this chapter answers
  • The setup — architecture variable produces value-capture shape but operator concentration does not vary across architectures; three forward-looking trends + two protocol-level events (ePBS + Alpenglow); three previewed findings (intents deepen consent gap; appchains relocate moat to capital-gated stake; cross-chain layers compress surface but concentrate operators)
  • The worked examples — three short hypothetical traces (Alice's UniswapX intent ~2–5 bps with execution path hidden; Carlos's hypothetical HIP-3 deployment with 500K HYPE moat; the cross-chain searcher's reallocation as surface compresses + ERC-7683 rises)
  • The mechanics, in detail
    • Intent-based architectures — what they change and don't change (UniswapX SCP + Wintermute >90%; CoW Swap Barter consolidating ~28% → targeting >50% via Copium acquisition; Anoma Phase 1 mainnet 29 Sep 2025; 1inch Fusion+ cross-chain; the consent gap deepens)
    • Appchains and sequencers — what they change and don't change (HIP-3 OI 9× growth with TradeXYZ >90%; HIP-4 launched 2 May 2026 at 1M HYPE Phase 2 stake / 2× moat; Optimism Superchain native interop + Espresso shared sequencer; Astria sunset; Eclipse retrenched; SVM modularisation via Anza; DoubleZero Edge 16 Apr 2026)
    • Cross-chain execution layers — what they change and don't change (LayerZero ~75% bridge volume + Stargate acquisition; CCIP 1,972% growth + Kraken adoption 14 May 2026; Across 88% ERC-7683 + top-5 relayers ~70%; Maire et al. ~$10M/year cross-chain MEV; ERC-7683 ratified + Open Intents Framework 30+ teams; surface compresses but operators concentrate)
    • ePBS and Alpenglow — two protocol-level events on the immediate roadmap (EF Checkpoint #9 Glamsterdam slipping with Q3/Q4 2026 realistic + 2027 risk if FOCIL bundled; Justin Drake Strawmap 7 forks through 2029 + Hegotá w/ FOCIL+EIP-8141; Alpenglow live validator testing 11 May 2026 + Agave 4.1 late 2026; Votor + VAT + ~10× threshold drop; ePBS doesn't dent builder concentration; Alpenglow restores bottom not top)
  • The closing structural argument — D1 inline table: three trends × architectural change / what doesn't change / consent-gap effect; architecture choices change actor shape not concentration shape; ePBS + Alpenglow reinforce finding; tentative answer to Ch 11 consent-gap question is "no in retail's case; partial in informed-consent cases"
  • What changes when… — the book ends here; structural map handed to reader; verdict belongs to Epilogue
  • Footnotes and sources (22 numbered citations)

Epilogue — A Note on Who's Reading This ​

A short closing note (written by Nick, not the agent). Identifies Figment, explains why the firm cared enough to commission this book, and frames the call to action: if you've read this far, you understand why blockspace and execution — not validator uptime — are the products that matter, and you should talk to us.

Appendices ​

  • Glossary — every term, alphabetized, with first-use chapter reference
  • Bibliography — every source cited, by chapter
  • A note on numbers — methodology for the dollar figures used, caveats on what's measured and what isn't
  • Further reading — 10–15 primary sources for the reader who wants to go deeper

Production order ​

The agent does not write in TOC order. Recommended order:

  1. Chapter 4 (The Searcher) first — it's the most self-contained and lets the agent's voice calibrate on adversarial content early.
  2. Then Chapter 2 (Liquidity) — establishes the AMM/CLOB foundation everything else builds on.
  3. Then Chapter 1, 3 — setup chapters, written once we know what voice works.
  4. Then Chapters 5–7 — the actor chapters, in order.
  5. Then Chapters 8–10 — the chain chapters, which can borrow language and structure from each other.
  6. Then Chapters 11–12 — the verdict and forward-looking chapters, which depend on everything else.
  7. Prologue and epilogue last.

Each chapter goes through Phase 1 → Nick review → Phase 2 → Nick review → Phase 3 → Nick review before moving on. Phase 1 of the next chapter can start while Phase 3 of the current chapter is in review, but no chapter goes into Phase 3 before the previous chapter has finished Phase 3.