The House Edge

Appearance

Chapter 10 — Review Notes (Phase 3 self-check) ​

Status: REVIEW NOTES (Phase 3 — written by agent, accompanying DRAFT.md) Date: 2026-05-14 Working chapter: 10 — Ethereum and Its L2s: The Most Studied, Least Settled Word count: ≈7,400 words including footnotes (≈6,800 words of prose plus 20 footnotes plus one inline Mermaid two-leg flowchart and two inline dollar-trace tables). Slightly above the Bible's 6,000-word target — Chapter 10 is the closer for Part III and carries the heaviest cross-referencing load (Chapters 5, 6, 7, 8, 9 all referenced) plus three new empirical surfaces (cross-chain MEV via Maire et al., L2 sequencer economy, restaking via EigenLayer). The chapter's structural finding — the PBS + L2 fragmentation + restaking trifecta — needed enough room to land cleanly against the comparison with Chs 8 and 9.

The eight required questions ​

  1. Could a smart business reader with zero crypto background follow this chapter on first read?Yes — the chapter is a closer / synthesis chapter that names every concept it uses against an already-developed Ch 5/6/7 baseline. The chapter introduces five named primitives (L2 sequencer, Timeboost, EigenLayer, AVS, cross-chain MEV) plus the trifecta verdict frame. Most named actors return from earlier chapters (Titan, BuilderNet, Quasar, Flashbots, Coinbase Cloud, Lido, Consensys); the only structurally new institutional actor is Eureka Labs as the new builder entrant. The chapter's load is concentrated in §4c (the L2 sequencer economy with the Base 62% concentration plus the Timeboost two-firm finding) and §4d (cross-chain MEV with the Maire et al. concentration and the Gogol et al. cross-rollup measurement) — both dense but each anchored to a single empirical finding with a single sourced number.

  2. Is every actor named, and is it clear how each one makes money?Yes. Titan (~52% builder share; revenue from exclusive-flow contracts ~17.75% margin); BuilderNet (~25% builder share; multi-operator non-exclusive design); Quasar (~15% builder share); Eureka Labs (the new entrant; $6.7M seed in Q2 2025 from Spark + Collider; ~2% share; "programmable blocks"); Flashbots (co-operates BuilderNet, runs MEV-Boost / Protect / Relay / SUAVE / Flashnet R&D); Coinbase Cloud (~5–12% Ethereum staker; Base sequencer; OFAC-filtering-optional validator; EigenLayer operator); Consensys (Linea sequencer; MetaMask; Infura; MEV-Blocker post-Jan 2026); Lido (~24% of all staked ETH; named EigenLayer restaking participant); the L2 sequencer operators (Offchain Labs / Coinbase / OP Labs / Uniswap Labs); Espresso Systems (12 Feb 2026 mainnet); Astria; named AVSs (EigenDA, AltLayer MACH, Hyperlane, Witness Chain); the cross-chain searcher top-5 addresses (per Maire et al.); the Timeboost two-firm winners (per Castro et al.). Each one's revenue mechanism is explicit.

  3. Is there a worked example with specific dollar amounts threaded through the chapter?Yes — sharpest two-leg comparison in the book. Alice's $10,000 USDC → ETH swap is walked through the L1 path (MetaMask → Flashbots Protect → CoW Swap → SCP solver → Titan → Ultrasound Relay → Lido validator) with a ~$15–25 cumulative take, and through the Base path (Coinbase Wallet → Aerodrome → Base sequencer → blob settlement) with a ~$2–5 cumulative take. The two-leg dollar trace is the chapter's central illustration of the architectural finding: L2 architecture produces lower per-trade costs for retail than L1, but concentrates the fee surface into a single sequencer operator whose economic position is meaningfully different from any L1 validator. The chapter explicitly acknowledges two illustrative elements (the Lido staker-share is composed; the SCP/Barter solver retention is pulled from Ch 7) and frames the comparison as the chapter's clinical comparison rather than a moral judgement. The Maire et al. cross-chain searcher pattern is the worked example's third reference point, deployed in §4d.

  4. Does the chapter end with a clear "who wins, who loses" verdict?Yes — sharpest four-layer stratification in Part III. §6 orders the four winning categories (L1 builder oligopoly; L2 sequencer monopolies; cross-chain searchers; restaking operators) and notes retail traders' L1 losses (~$15–25 on $10K) vs L2 lower-cost but single-operator concentration (~$2–5). The closing structural observation lands the chapter's editorial contribution: "the architecture does not produce operator competition; the architecture produces a legible map of where the operator concentration sits." The clinical comparison — "whether knowing exactly which firm extracts which dollar at which layer is meaningfully different from not knowing" — is the chapter's open question and Part III's closing argument.

  5. Are all numbers sourced in footnotes?Yes. 20 footnotes. Every dollar figure, percentage, validator count, builder share, sequencer revenue, TVL, latency, and AVS count has a footnote with URL and access date. Cross-references to footnotes already cited in earlier chapters (Coinpedia / PANews Q1 2026 from Chs 8, 9; relayscan.io from Chs 5, 6; Wu et al. and Pahari-Canidio from Ch 7; EF Checkpoint #9 from Ch 5; Hyperliquid bridge deprecation from Ch 9; Pectra recap from Ch 5) are marked "Already cited in Chapter X" where appropriate. The empirical anchors that are new for Ch 10 — Maire et al. (cross-chain MEV), Gogol et al. (cross-rollup MEV), Castro et al. (Timeboost auction concentration), Eureka Labs funding, EigenLayer mainnet slashing — are each freshly cited.

  6. Does the chain comparison box exist and contain real differences?Yes — and it is the chapter's load-bearing structural argument. §5 explicitly compares the three Part III chains: Ethereum (deliberate dispersion across PBS + L2 + restaking trifecta; dispersion real architecturally not by operator); Solana (vertical integration compressed surface but did not eliminate it; access gap as software-and-geography function); Hyperliquid (eliminated surface; four-locus re-concentration). The three-chain pattern is the central argument of Part III and lands cleanly: architectural choices produce different value-capture shapes, but operator concentration is comparable across all three chains. This is the chapter's editorial contribution to the book's adversarial thesis.

  7. Did I avoid every banned move from the Book Bible?Yes, with one flagged near-miss.

    • No hype words. No doom words. No "ETH killer." No "ETH is dead." No "ETH is winning." No tribal endorsement. Specific firms named, specific dollar amounts, specific dates. The "deliberate dispersion" framing is the chapter's architectural reading, grounded in Vitalik Buterin's and Justin Drake's published statements about the L2-centric roadmap and the EigenLayer founders' AVS framing.
    • Near-miss: The closing observation that "the amount of public research on Ethereum's extraction surface exceeds the equivalent body of work on Solana or Hyperliquid by an order of magnitude" is editorial-by-degrees. It is a synthesis claim, supported by the four named arXiv papers (Wu, Pahari-Canidio, Maire, Castro) and the L2BEAT / relayscan.io / mevboost.pics live-dashboard infrastructure. The framing pushes a structurally-attentive reading of Ethereum's economy — which I think is the right framing, but the reader is being nudged toward "Ethereum is more legible" without me asserting "Ethereum is therefore better."

  8. Would the Goldman MD finish this chapter without checking her phone?Yes, with confidence. The cold open's three named anchors (Q1 2026 REV ranking, Base 62% L2 fee capture, EigenLayer $18B TVL) are exactly the kind of TradFi-comprehensible facts she will respond to. The two-leg Alice dollar trace in §3 is the most concrete worked example in the book — six layers on L1, four on Base, with explicit per-layer figures and acknowledged illustrative elements. The trifecta framing (PBS + L2 + restaking) is the kind of three-pillar structural argument she will remember. The closing observation about "legible map of operator concentration" is the chapter's editorial contribution and the kind of clean structural reading she will appreciate. The 6,800-word total is slightly above the range but justified by Part III's closing role.

What changed between phases — and what's load-bearing ​

Claims dropped from RESEARCH.md ​

  • The full Tornado Cash + Roman Storm regulatory update. The research note had a detailed treatment of the 21 March 2025 OFAC delisting, the Fifth Circuit ruling, and the July 2025 Roman Storm trial. The chapter omits this in favour of a brief cross-reference to Ch 7's broader PFOF/OFAC regulatory framing. The structural narrowing of OFAC concerns at the Ethereum builder layer is noted in the verdict but not developed.
  • The BlackRock ETHB ETF launch (12 March 2026) and SEC + CFTC 17 March 2026 joint interpretive release. The research note had detailed coverage of the ~$107M seed; 0.50–0.95% management fees; ~1.9–2.6% net staking yield. The chapter omits these specifics — they would have lengthened §4a without changing the structural argument. The "Ethereum staking is now a regulated US ETF asset" fact is implicit in the Hyperliquid BHYP discussion in Ch 9.
  • Detailed per-AVS economics for EigenLayer. The research note had operator-by-operator and AVS-by-AVS revenue data. The chapter names the AVSs and the operator-count headline but does not develop per-operator economics — the structural argument (restaking is a third architectural surface) is the load-bearing claim, not the operator-level dollar amounts.
  • The full Hyperliquid Arbitrum bridge deprecation timeline. Ch 9 carried this; Ch 10 references the deprecation briefly in §5 as an L2-thesis hint but does not re-litigate the December 2024 DPRK scare or the CCTP migration mechanics.
  • The seven-fork Justin Drake Strawmap detail through 2029. The research note had each fork's headliner pair. The chapter references the Strawmap as a single citation in §4b without enumerating all seven forks — too much forward-looking detail for the chapter's structural argument.

New claims added in the draft (and where they came from) ​

  • The chapter's central framing — "Ethereum has deliberately dispersed its value capture across PBS + L2 + restaking trifecta." This is the chapter's organising synthesis claim. It is composed from Chs 5 (PBS architecture), 6 (the infrastructure layer firms), 7 (exclusive flow), 8 (the Q1 2026 REV reversal), and 9 (the eliminated-surface comparison). The "trifecta" framing as the chapter's verdict-frame is the chapter's contribution.
  • The "dispersion real along architectural lines, not along operator lines" claim. Composed from the named-firm enumeration: Coinbase Cloud at multiple layers (staker + Base sequencer + OFAC-filtering validator + EigenLayer operator); Consensys at multiple layers (Linea sequencer + MetaMask + Infura + MEV-Blocker post-Jan 2026); Flashbots at multiple layers (BuilderNet + MEV-Boost + Protect + Relay + SUAVE/Flashnet R&D). The cross-layer operator-concentration argument is the chapter's editorial contribution.
  • The "architecture produces a legible map of where the operator concentration sits" closing observation. The chapter's most pointed editorial-by-degrees claim. Composed from the Wu / Pahari-Canidio / Maire / Castro papers plus the L2BEAT / relayscan.io / mevboost.pics dashboards. The "legibility without competition" reading is the chapter's contribution to the book's adversarial thesis.
  • The "L2 architecture produces lower per-trade costs but concentrates fee flow into a single sequencer operator" framing. Composed from the Alice two-leg dollar trace. The structural framing is the chapter's; the underlying numbers (Base ~62% of L2 fee revenue, ~$185K/day, etc.) are sourced.
  • The "same five-to-ten firms operate across all three chain architectures" cross-chain operator-concentration claim. Composed across Chs 8, 9, and 10. The structural reading that operator concentration is comparable across three radically different chain architectures is the chapter's contribution.

Things I'm uncertain about ​

  1. The MEV-Boost share numbers will drift before publication. relayscan.io's 24-hour snapshots are live data; the figures cited (Titan ~52.16%, BuilderNet ~24.63%, Quasar ~15.06%, Eureka ~2.41%) reflect early-to-mid 2026 measurements. If publication slips past Q2 2026, these will need refreshing.
  2. The Q1 2026 REV ranking durability. Ethereum #3 behind Hyperliquid and Solana for Q1 2026 is the chapter's central anchor. The ranking depends on Hyperliquid's perp-volume growth, Solana's memecoin-cycle recovery (Alpenglow mainnet Q3/Q4 2026 per Ch 8), and Ethereum's L1 fee-economy dynamics post-Fusaka. The chapter frames the ranking as a Q1 2026 fact, not as durable shape.
  3. The Base 62% L2 fee capture figure is a recent-period snapshot. Yellow Research and Pine Analytics report this as the 2025 full-year fact. If publication slips and Base's share moves materially (e.g., if Unichain or Linea ramp faster than expected, or if Coinbase's L2 strategy changes), the chapter's prose may need updating.
  4. The Eureka Labs "programmable blocks" framing. The firm's product positioning is genuinely differentiated from Titan's exclusive-flow model and BuilderNet's open-source non-exclusive design, but the actual mechanics of "programmable blocks" are not fully documented in primary sources. The chapter notes the positioning as the firm's claim rather than as a structural fact.
  5. The Castro et al. Timeboost finding (>90% of auctions won by 2 entities). Sourced from arXiv 2509.22143 (September 2025). The methodology is sound but the data is approximately 9 months old at the chapter's writing date; auction concentration may have changed. The chapter cites the source's finding as the most recent published anchor.
  6. The Maire et al. cross-chain arbitrage data window (Sep 2023 – Aug 2024). This is the load-bearing empirical anchor for the cross-chain MEV surface. The data is approximately 21 months old at the chapter's writing date. The structural claim (cross-chain MEV is a measured ~$10M/year extraction surface with extreme concentration) is durable; the specific dollar amounts will have evolved.
  7. The two inline tables in §3 (Alice's L1 and Base dollar traces). The L1 path has seven rows; the Base path has four rows. The visual asymmetry reflects the architectural asymmetry — L1 routes through more layers than Base. A reader might compare row count and reach an unintended conclusion (more rows = more friction); the prose around the tables explicitly frames the comparison as "L2 architecture produces lower per-trade costs but concentrates the fee surface."

Places where the prose got technical and might lose the reader ​

  • §4c (the L2 sequencer economy). Three named L2s (Arbitrum, Base, Optimism) plus their sequencer operators (Offchain Labs, Coinbase, OP Labs) plus Arbitrum's Timeboost auction mechanics plus the Castro et al. concentration finding plus the Espresso/Astria/Superchain shared-sequencer landscape is the chapter's densest single subsection. The chapter chose to keep them together because the structural argument requires all three: Base's 62% capture + Arbitrum's Timeboost concentration + the slow decentralised-sequencer roadmap together produce the "L2 fee economy is single-sequencer-monopoly today, with decentralisation 12–18 months out" finding.
  • §4d (cross-chain MEV). Two arXiv papers (Maire and Gogol) plus the atomicity-vs-non-atomicity distinction plus the pre-positioned-inventory mechanics plus the shared-sequencer / intent-based answer is the chapter's most technical subsection. The chapter chose to keep these together because the structural argument requires the empirical anchors (Maire's 242,535 trades / $10M revenue / 5-address concentration) and the architectural framing (atomicity as the structural difference) together.
  • The trifecta framing in §5. The three-layer comparison (PBS + L2 + restaking) plus the same-five-to-ten-firms-across-layers claim plus the three-chain comparison (Solana / Hyperliquid / Ethereum) is the chapter's most concentrated argument. The chapter chose to land this in §5 (the chain-comparison section) rather than the verdict to give the reader a structural foothold before §6's stratification.

Files written/modified in Phase 3 (this chapter) ​

  • book/chapters/10_ethereum/DRAFT.md — new, ≈7,400 words (20 footnotes; inline Mermaid two-leg flowchart in §3; two inline dollar-trace tables (L1 path and Base path) in §3).
  • book/glossary/GLOSSARY.md — appended 5 entries: AVS (Actively Validated Service), Cross-chain MEV, EigenLayer, L2 sequencer, Timeboost (Arbitrum). Now 95 total entries.
  • book/OUTLINE.md — Chapter 10 entry updated with subtitle and final section headings (mirroring the format used for Chapters 5, 6, 7, 8, 9).

No back-edits to prior chapters this time. The chapter explicitly forwards to Ch 11 (Part IV opens) and references Chs 5, 6, 7, 8, 9 throughout without modifying them. The Ch 5 builder-share figures and Ch 7 exclusive-flow figures are quoted directly from those chapters' footnotes; the cross-chain MEV / restaking / L2 sequencer empirical anchors are fresh to Ch 10.

Phase 3 is complete. Part III is complete. The chapter is now in Nick's review queue.