The House Edge

Appearance

Chapter 3 — Research Note (Phase 1) ​

Status: RESEARCH (Phase 1 — written by agent, awaiting Nick's review) Date: 2026-05-14 Working chapter: 03 — The Mempool and What Replaces It Carries forward from: SPEC.md (strawman drafted by agent; Nick gave content direction in a separate message to cover Jito mempool history specifically, plus BAM, Harmonic, and the multi-chain comparison)

How to read this note ​

This chapter is the book's hinge — adversarial content lands here. Two things to know up front:

  1. Solana's mempool history is the chapter's anchor. Jito Labs ran a private mempool from August 2022, shut it down on or about 8 March 2024, and built three distinct successors: Jito's Block Engine (bundle auction without the mempool surface), Jupiter MEV-Protect / Beam (wallet-level private routing), and BAM (TEE-based confidential block construction, live July 2025, shipping into broad adoption through 2026). The Solana Foundation removed approximately 32 validators (~1.5M SOL, ~0.5% of delegation-program stake) in June 2024 for running private mempools that enabled sandwich attacks. This is the cleanest single narrative the chapter has — a story with names, dates, and consequences.
  2. Ethereum's surface is now privately-routed by value, even if not by count. A peer-reviewed measurement from Wahrstätter et al. finds that approximately 30% of Ethereum transactions and over 50% of gas, but 54.6% of total block value, now flows through private channels. Titan alone built 47.6–50.6% of Ethereum blocks in 24-hour snapshots in May 2026; combined with BuilderNet, two firms now build ~73% of blocks. The public mempool is increasingly empty of valuable trades.

1. Key claims ​

Each numbered claim is something the chapter is allowed to state. Sources cited inline.

Solana's mempool history

  1. Jito Labs launched the Jito-Solana client into mainnet and testnet operation in late 2022 in validator-only form; the Block Engine — and with it broad searcher access to the mempool — opened to all Solana MEV searchers on 25 May 2023, after the prior validator-only cohort had processed over a million bundles. The mempool gave searchers an approximately 200-millisecond window to preview pending transactions before the slot leader included them. (Helius — Solana MEV Report; Jito Labs announcement / Solana Compass coverage of the 25 May 2023 broad-availability date)

  2. Jito Labs shut down the mempool on or about 8 March 2024, abruptly and within hours. The stated reason was to "prioritize the network's long-term growth and stability, sacrificing significant revenue in the process." Coverage at the time framed it as a response to evidence that the mempool surface was enabling sandwich attacks against retail traders. (CoinDesk — Solana Client Developer Jito Ends 'Mempool' Function, 2024-03-08; Decentralised.co — Why Jito Suspended Its Mempool on Solana, March 2024)

  3. The mempool shutdown removed the surface but did not remove the demand. Within months, at least three validator operators reported solicitations from searcher firms to run private (closed) mempools with profit-share terms — a black-market substitute for the eliminated public path. (Decentralised.co)

  4. The Solana Foundation responded by removing approximately 32 validator operators representing roughly 1.5 million SOL (~0.5% of delegation-program stake) from the foundation's delegation program in June 2024, citing participation in private mempools that enabled sandwich attacks. Tim Garcia, the foundation's validator-relations lead, described the action as ongoing rather than one-time enforcement. (CoinDesk — Solana Heavyweights Wage War Against Private Mempool Operators, 2024-06-10)

4a. The enforcement did not end the demand. Vpe — the canonical sandwich operator named in Chapter 4 — is itself operated by a specific Solana validator known as DeezNode (validator address HM5H6…jdMRA). DeezNode's stake nearly tripled across late 2024, from approximately 307,900 SOL (epoch 697, 13 November) to 802,500 SOL (epoch 709, 9 December), on the back of a published offer of a 50% MEV revenue share to opting-in validators. (New World Finance — This Solana validator extracted over $13M, 2025-03-24; Anarcaze — Spotting Shady Validators on Solana)

4b. The post-shutdown visibility surface fragmented across three layers underneath the formal architectural responses:

  • RPC layer: Kolibrio (VC-backed Order Flow Auction startup, founded early 2022, $2M Jump-led seed Feb 2023) operates a paid-tier version of what Jito's public mempool used to do, branded as "Meow RPC." Integrators route user transactions through Kolibrio; Kolibrio auctions visibility to searchers and rebates a share back. (Kolibrio Documentation; QuickNode — Feature Fridays: Kolibrio; AIN.ua, 2023-02-16)
  • Wallet infrastructure layer: Helius — the network's largest single validator at 15M+ SOL of stake — runs the RPC pipeline for Phantom, Backpack, Solflare, Bitgo, Ledger, Exodus, Squads, and Trust. Helius markets "SOL rebates from trades that create arbitrages" to its wallet customers. LaserStream's "shred-level ingestion" is the functional Solana equivalent of mempool visibility, gated to paid tiers ($499/mo Business; $999/mo Professional). (Helius — Wallets use case; Helius — Zero-Slot Execution with Sender and LaserStream)
  • Validator co-location layer: practitioner documentation describes positioning non-voting validator nodes "in the same rack" as paid RPCs, forwarding transactions to the current leader's TPU via QUIC ahead of the public gossip path. (RPC Fast — Co-location Strategies for Solana Trading in 2026)
  1. Despite the shutdown, Solana's dominant sandwich operator continued operating against the residual surface. The Vpe program — the same operator the book named in Chapter 4's cold open — ran approximately 1.55 million sandwich transactions over thirty days in December 2024 to January 2025 at an 88.9% success rate, generating about $13.4 million in gross profit. Jito's own analysis attributed roughly half of all Solana sandwich activity in that window to Vpe. (Helius — Solana MEV Report)

Jito BAM (Block Assembly Marketplace)

  1. Jito Labs launched BAM publicly on 21 July 2025. BAM separates the transaction-sequencing function (handled by BAM Nodes) from the transaction-execution function (handled by validators). BAM Nodes run inside hardware Trusted Execution Environments — specifically AMD SEV-SNP enclaves — that keep transaction contents encrypted until execution and produce cryptographic attestations of the ordering decision. (Jito Labs — Introducing BAM, 2025-07-21; Helius — Block Assembly Marketplace (BAM))

  2. The "confidential-execution surface" referenced in Chapter 4 is BAM's TEE enclave. The privacy guarantee depends on hardware attestation rather than on cryptographic primitives or time-locks — BAM is trust-minimized (depending on AMD's hardware) rather than trustless. TEE overhead is approximately 2–5% of normal validator load. (Helius — BAM)

  3. BAM has no direct user fees. Revenue flows from applications and searchers through BAM operators to validators, with a 6% tip split between Jito Labs and the Jito DAO. The Q2 2025 DAO share of this fee was approximately 22,391 SOL (~$4 million at then-prevailing prices). (Helius — BAM)

  4. BAM supports Application-Controlled Execution (ACE) plugins that let specific actors customise sequencing logic. Jump Crypto launched a "BAM Maker Priority" plugin in April 2026 that gives market makers deterministic top-of-batch execution every fifty milliseconds. Other plugin examples include oracle prioritization, liquidation-before-market sequencing, and LP-priority access in AMM batches. (Blocmates — BAM: Solana's Block Builder Era)

  5. As of March 2026, Solana's validator client stack is now multi-modal. The Syndica deep-dive on Solana on-chain activity for March 2026 reports the following stake-weighted shares: Agave Jito 32%, Agave JitoBAM 28%, Agave Harmonic 17%, Frankendancer Jito 12%, Agave Rakurai 6%, Firedancer 2%. Agave-based clients run 86% of stake. (Syndica — Deep Dive: Solana Onchain Activity, 2026-04-27)

Harmonic

  1. Harmonic is not a fork of Agave and not a single scheduler. It is an open block-building marketplace, launched November 2025 with a $6M Paradigm-led seed. Founded by Ben Coverston and Jakob Povšič (Povšič also co-founded Temporal, one of the builders Harmonic auctions for). A validator running Harmonic continuously receives candidate blocks from multiple independent builders — Jito, Temporal, JitoBAM, Paladin — and selects the most valuable candidate per slot, expressing configurable preferences (priority weighting, builder allow-lists, partner exclusivities) that shape how the selection is made. The architectural premise is that block construction should be a competitive market rather than a single-firm decision; the preferences architecture also creates a structural hook for partner-specific arrangements that Harmonic itself does not publicly advertise. Nasdaq-listed DeFi Development Corp adopted Harmonic in December 2025 to "optimize Solana validator revenue." (Harmonic — Overview; The Block — Paradigm leads $6M Harmonic seed, 2025-11-05; GlobeNewswire — DFDV partners with Harmonic, 2025-12-01)

  2. Harmonic ran approximately 17% of Solana network stake in March 2026. Validators running Harmonic capture meaningfully more priority-fee revenue than the median: Frankendancer Harmonic Performance earned +101% priority fees per block versus median, Frankendancer Harmonic Balanced +39%, and Agave Harmonic +36%. Block-time-adjusted figures are lower but still significantly above median: Agave Harmonic +33%, Frankendancer Harmonic Balanced +32%, Performance +28%. (Syndica — Deep Dive)

  3. Helius has published a proposal called Constellation — a Multiple Concurrent Proposers (MCP) design with 16 proposers, 256 attesters, and a leader that orders strictly by priority-fee per compute unit. The explicit goal is to replace off-protocol scheduler layers, including Jito's Block Engine, Harmonic, and BAM, by moving inclusion guarantees into the protocol itself. Constellation is targeted to follow Alpenglow's mainnet activation, which itself is expected in late Q3 or early Q4 2026. (Chainflow — Constellation and the Future of Block Production, Q1 2026)

Jupiter MEV-Protect / Beam in 2026

  1. Jupiter Ultra is the aggregator's default user-facing routing mode in 2026. Ultra uses Jupiter Beam, an in-house transaction-landing engine that provides complete transaction privacy until on-chain execution and reduces landing latency by 50–66% relative to the prior multi-provider stack. Jupiter Beam launched its Public Transaction Submission API in April 2026, allowing any signed Solana transaction to enter the protected route — not just Jupiter-routed swaps. (Jupiter — Ultra Swap API Documentation)

  2. Ultra V3 user trades show approximately +0.006% average slippage (positive — better than oracle mid), versus −0.14% on unprotected paths. Jupiter handles approximately 95% of Solana DEX aggregator volume and approximately 50% of total Solana DEX volume; over $700 million in daily swap volume routes through it. (Bravian Oyaro — Jupiter's MEV Protection Architecture, 2026)

Ethereum private routing

  1. Private order flow has become the value-weighted majority of Ethereum transaction activity. The most recent peer-reviewed measurement (Wahrstätter et al., updated through 2025) finds that private orders consume more than 50% of gas used on Ethereum and account for 54.59% of total block value, while comprising approximately 30% of transactions by count. The public mempool is increasingly empty of valuable trades, even when it appears full by transaction count. (Wahrstätter et al. — Private Order Flows and Builder Bidding Dynamics, arXiv:2410.12352)

  2. The three dominant private-routing services for Ethereum retail flow:

    • Flashbots Protect has served 2.1 million Ethereum accounts and shielded approximately $43 billion in cumulative DEX volume; third-party benchmarks estimate roughly 7% of Ethereum monthly transaction count (~3 million transactions per month) flows through Protect. (Flashbots — Flashbots Protect Documentation)
    • MEV-Blocker, transferred from CoW DAO to Consensys' Special Mechanisms Group in January 2026, has served over 4.5 million unique wallets and paid 6,177 ETH in cumulative rebates to users since launch. (CoW DAO — SMG Acquires MEV Blocker RPC, January 2026)
    • MEV-Share is the Flashbots-operated mechanism that lets users selectively share transaction details with searchers in exchange for a share of extracted value. Adoption has been slower than the simpler Flashbots Protect path.

MEV-Boost and builder concentration in 2026

  1. The MEV-Boost relay/builder stack is now sharply concentrated. Live relayscan.io snapshots for the 24 hours ending 14 May 2026 show:

    • Relays: Ultrasound 35.71%, Titan 26.09%, BloXroute Max-Profit 13.38%, BloXroute Regulated 12.30%, Aestus 7.32%, Flashbots 2.09%.
    • Builders: Titan 50.60%, BuilderNet 24.55%, Quasar 15.68%, Beaverbuild 1.83%.
    • The top three builders together account for approximately 91% of all Ethereum blocks. (relayscan.io)

  2. Flashbots itself stopped operating a centralised block builder in December 2024, migrating its building operation into BuilderNet (a decentralised, multi-operator builder protocol). A February 2026 Flashbots writing frames the open question of the era: "80%–90% of Ethereum blocks are produced by the largest two block builders." (Flashbots — Migrating to BuilderNet, 2024-12-09; Flashbots — decentralized building: wat do?, 2026-02-14)

L2 sequencer behavior

  1. Arbitrum activated Timeboost in April 2025. Timeboost replaces the prior first-come-first-served sequencing with an off-chain sealed-bid auction granting the winner a 200-millisecond "express lane" advantage per round. Proceeds flow to the Arbitrum DAO. The Arbitrum sequencer remains operated by Offchain Labs. In February 2026, an academic paper documented that searchers were routing around the primary auction via a third-party resale market called Kairos, prompting Arbitrum to raise and then lower the auction's reserve price. (Arbitrum Docs — Timeboost Gentle Introduction; arXiv 2509.22143)

  2. Base runs a single sequencer operated solely by Coinbase. The sequencer respects OFAC SDN-list filtering. Base's sequencer revenue for H1 2025 was approximately $42.4 million, accounting for 87.2% of total Optimism Superchain sequencer revenue; daily priority fees alone average approximately $156,138, or roughly 86% of daily sequencer revenue. Coinbase has stated that it does not directly run sandwich-attack strategies, but the priority-fee auction permits any third party to backrun visible transactions. (The Block — Base Defends Moving Sequencer Fees; Messari — State of the Superchain H1 2025)

  3. Optimism's Unichain, a member of the Superchain, deploys Flashbots-built Flashblocks with 200-millisecond block times and returns MEV to users via revenue-share. This is the most aggressive MEV-returns-to-user design currently live among major L2s. OP Mainnet itself contributes 100% of sequencer profit back to the Optimism Collective. (Optimism Foundation — Optimism Partners with Flashbots)

  4. No major L2 (Base, Arbitrum, Optimism) had migrated to a decentralised sequencer as of May 2026. All three still run a single operator-controlled sequencer that has full visibility into the L2's pending transactions before ordering them. The leading decentralised-sequencing services are:

    • Espresso Systems — mainnet went generally available on 12 February 2026, finalising blocks in approximately 6 seconds. (Caldera × Espresso Vienna integration)
    • Astria — mainnet alpha launched 28 October 2024 on Celestia data availability.

Hyperliquid as the architectural alternative

  1. Hyperliquid's HyperBFT consensus processes transactions and order-book actions inside consensus itself — there is no public mempool and no pre-execution visibility layer. The chain's matching is semantically aware (non-order actions process first, then cancels, then GTC/IOC sends within a block), which structurally eliminates the surface that simple sandwich attacks require. Chapter 2 already cited the architectural choice; Chapter 3 develops what it eliminates. (Hyperliquid Documentation — HyperCore Overview)

Censorship resistance vs MEV resistance

  1. The Ethereum protocol-level response to the visibility-versus-censorship trade-off is FOCIL (Fork-Choice enforced Inclusion Lists), specified in EIP-7805. Under FOCIL, a 16-validator committee per slot publishes inclusion lists; attesters only vote for a proposer's block if it includes the transactions on the stored inclusion lists, with conditional omission permitted only when gas or block space is unavailable. FOCIL has been selected as the consensus-layer headliner for Hegotá, the post-Glamsterdam Ethereum upgrade. Glamsterdam is targeted for H1 2026 with ePBS as its headline; Hegotá is now framed as a late-2026 cleanup-and-hardening fork. No firm FOCIL mainnet date has been published. (EIP-7805; Ethereum Foundation — Checkpoint #9, 2026-04-10; Ethereum Foundation — Protocol Cluster Updates: May 2026)

  2. Flashbots has framed FOCIL as the consensus-layer half of a two-part answer to the trade-off, with their own Flashnet (a network-anonymized mempool, published February 2026) as the submission-layer half. The argument: privacy at submission plus enforced inclusion at consensus addresses both extraction and censorship risk in different layers. (Flashbots — Network Anonymized Mempools (Flashnet), 2026-02-17)

2. Numbers to verify ​

#NumberSourceDateFlag
N1Jito-Solana on mainnet/testnet (validator-only) from late 2022; Block Engine + broad searcher access opened 25 May 2023; ~200ms preview window for searchers; ~10 months of broad availability before March 2024 shutdownHelius MEV Report; Jito Labs / Solana Compass2022–2024Strong; the 25 May 2023 broad-availability date is load-bearing for the chapter's "~10 months" framing
N2Jito mempool shut down on or about 8 March 2024CoinDesk; Decentralised.co2024-03-08Two sources slightly disagree on date (8 vs 9 March); hedge as "on or about"
N3On 8 March 2024, daily Solana MEV tips exceeded 10,000 SOL (~$1.5M)Decentralised.coMarch 2024Sets the scale of what the shutdown gave up
N4June 2024: Solana Foundation removed ~32 validators (~1.5M SOL, ~0.5% of delegation-program stake) for private-mempool participationCoinDesk2024-06-10Concrete enforcement number
N5BAM launched 21 July 2025; uses AMD SEV-SNP TEE enclaves; TEE overhead ~2–5%Jito Labs — Introducing BAM; Helius — BAM2025-07-21Primary source
N6BAM revenue: 6% tip split between Jito Labs and Jito DAO; Q2 2025 DAO share ~22,391 SOL (~$4M)Helius — BAMQ2 2025
N7March 2026 Solana client-stack stake share: Agave Jito 32%, Agave JitoBAM 28%, Agave Harmonic 17%, Frankendancer Jito 12%, Agave Rakurai 6%, Firedancer 2%; 86% on AgaveSyndica2026-03The chapter's load-bearing Solana market-share number
N8Harmonic priority-fee capture vs median (Mar 2026): Frankendancer Harmonic Performance +101%, FH Balanced +39%, Agave Harmonic +36%Syndica2026-03Strong
N9Vpe sandwich program: 1.55M sandwich txns / 30 days; 88.9% success rate; $13.4M gross / $4.6M Jito tips (Dec 2024 – Jan 2025)Helius MEV ReportQ1 2025Already in Ch 4; reuse here as continuity
N10Jupiter Ultra V3: +0.006% avg slippage vs −0.14% unprotected; ~95% of Solana DEX aggregator volume; ~50% of total Solana DEX volume; >$700M daily swap volumeBravian Oyaro2026Bold improvement vs Ch 2's Jupiter framing
N11Ethereum private order flow (Wahrstätter et al.): ~30% of transactions by count, >50% of gas, 54.59% of total block valuearXiv 2410.123522025 (v3)Peer-reviewed; the chapter's strongest single Ethereum stat
N12Flashbots Protect: 2.1M+ Ethereum accounts, $43B+ cumulative DEX volume shielded, ~7% of Ethereum monthly tx countFlashbots Protect Docs2026Self-reported by Flashbots
N13MEV Blocker: 4.5M+ unique wallets, 6,177 ETH cumulative rebates; transferred CoW → Consensys SMG in Jan 2026CoW DAOJan 2026
N14Ethereum relay share (May 14 2026, 24h): Ultrasound 35.71%, Titan 26.09%, BloXroute Max 13.38%, BloXroute Reg 12.30%, Aestus 7.32%, Flashbots 2.09%relayscan.ioliveWill drift
N15Ethereum builder share (May 14 2026, 24h): Titan 50.60%, BuilderNet 24.55%, Quasar 15.68%, Beaverbuild 1.83%; top 3 ≈ 91%relayscan.ioliveWill drift
N16Flashbots stopped operating a centralised builder in Dec 2024, migrated to BuilderNetFlashbots — Migrating to BuilderNet2024-12-09Foundational
N17Arbitrum Timeboost: sealed-bid auction granting 200ms express-lane advantage; activated April 2025; proceeds to Arbitrum DAOArbitrum DocsliveStrong primary
N18Base sequencer revenue H1 2025: $42.4M (87.2% of Superchain sequencer revenue); daily priority fees avg ~$156K (~86% of daily revenue)Messari — State of the Superchain H1 2025; The BlockH1 2025
N19Espresso Systems mainnet GA: 12 February 2026; ~6s block finalityCaldera × Espresso Vienna2026-02-12
N20No major L2 (Base / Arbitrum / Optimism) had migrated to a decentralised sequencer as of May 2026Composite of L2 docs2026-05Strong negative claim
N21FOCIL (EIP-7805): 16-validator IL committee per slot; selected as consensus-layer headliner for Hegotá (post-Glamsterdam fork); no firm mainnet dateEIP-7805; Ethereum Foundation Checkpoint #92026The protocol-level answer to the visibility/censorship trade-off
N22Flashnet (network-anonymized mempool) published February 2026Flashbots2026-02-17

Deliberately not pinned:

  • Exact SIMD / consensus-layer mechanism of the April 8, 2026 Solana sandwich-mitigation patch — multiple secondary sources confirm the date and effect but no canonical SIMD or release-tag is publicly indexed. The chapter will cite the patch as having shipped without overcommitting to the technical mechanism. Flag for verification before final publication.
  • Exact % of Ethereum L1 user count (vs gas-weighted or value-weighted) routed through private channels in 2026 — the Wahrstätter paper is the cleanest decomposition we have.

3. Contested or evolving claims ​

  • The April 8, 2026 Solana protocol-level sandwich mitigation. Chapter 4 cites it. Chapter 3 will need to cite it. The exact technical mechanism is not publicly documented in a way I could verify — multiple secondary sources confirm the effect (simple validator-leader sandwiches are largely suppressed) but the SIMD, the release tag, and the precise scope of the fix have not surfaced in a canonical primary source. The chapter will state the effect without overcommitting to the mechanism.

  • Whether the public Ethereum mempool is still load-bearing. The Wahrstätter et al. measurement — 54.6% of block value is now privately routed — suggests the public mempool is increasingly empty of valuable trades. The chapter's structural claim is that searchers reading the public mempool in 2026 are reading the leftovers. This is the chapter's most novel argument; it should be developed carefully, since it would undercut the simple "public mempool → predators" narrative the SPEC's strawman language implied.

  • Constellation as a Solana long-term answer. Helius's MCP proposal is well-documented but unbuilt; Anza has not committed to it as a roadmap item; Alpenglow is the immediate post-2026 priority. The chapter will name Constellation as one possible future direction but won't treat it as a certainty.

  • Whether Jupiter's Beam and the April 8 patch make MEV-Protect redundant. Jupiter continues to default to Ultra (with Beam routing) in 2026. Whether Beam still adds value over the post-patch baseline for clean major-pair swaps is unclear in the public record; for long-tail tokens and unprotected wallets, Beam remains the recommended path.

  • The censorship-vs-MEV-resistance trade-off as a settled discourse. It isn't. Hasu (Flashbots), Vitalik, and Justin Drake have published positions; the FOCIL / Flashnet / ePBS package is the current consensus answer, but the mainnet timelines are uncertain and Vitalik's "Big FOCIL plus encrypted mempools" framing is a moving target. The chapter will describe the trade-off and the current direction without claiming a winner.

4. Characters introduced ​

No new "Meet the actor" sidebar. The reader has met the relevant actors: the searcher in Chapter 4, the market maker in Chapter 2, the trader Alice across multiple chapters. Chapter 3's structural job is to develop the substrate (the visibility surface and its replacements), not introduce a new actor type.

Returning named actors:

  • Vpe — the Solana sandwich program named in Chapter 4. Returns here as the canonical example of what the post-shutdown surface still enables.
  • Jito Labs — the firm that built and then shut down the public mempool; the firm now building BAM. Named institution.
  • The Solana Foundation — the actor that did the June 2024 enforcement action (Tim Garcia named as the foundation's spokesperson).
  • Jupiter — appears as the routing layer with Beam.
  • Flashbots — the Ethereum private-routing institution.
  • CoW DAO and Consensys' Special Mechanisms Group — operators of MEV-Blocker.
  • Titan, BuilderNet, Beaverbuild, Quasar — Ethereum block builders.
  • Offchain Labs — Arbitrum's sequencer operator.
  • Coinbase — Base's sequencer operator.
  • Espresso Systems, Astria — decentralised-sequencing services.

Named individuals appearing for the first or second time:

  • Tim Garcia (Solana Foundation validator-relations lead, 2024)
  • Hasu (Flashbots; quoted on the neutral-relay problem)
  • Vitalik Buterin (ePBS, FOCIL, encrypted mempools)
  • Brennan Watt (Anza, Alpenglow lead — likely cited in the forward-looking section on Constellation)

5. Worked example candidates ​

The SPEC offered three options. The 2026 research reshapes the call.

Candidate A — Alice's $10,000 USDC→SOL swap, submitted through public RPC vs Jupiter Beam ​

Alice from the prologue and Chapters 1 and 4. The chapter walks both paths through the visibility layer: public RPC (her transaction observable to anyone watching the leader's pre-inclusion path until ~April 2026, when the protocol-level patch substantially compressed that window); Jupiter Beam (delivered as a private bundle, no public exposure, fills at +0.006% slippage). The reader sees what private routing actually does at the visibility layer, in the chain whose mempool history the chapter spent the most time on.

  • Pros: continuity with all three drafted chapters; lets the chapter's most novel content (Solana's mempool wars) carry the worked example; the +0.006% number is striking and well-sourced.
  • Cons: requires the chapter to make the Ethereum case in a one-paragraph cameo rather than a worked example.

Candidate B — Ethereum trader's $10K trade through Flashbots Protect vs public mempool ​

A different trader (call him Carlos, an institutional user — the chapter has reused Alice enough). Carlos changes his wallet's RPC endpoint to Flashbots Protect's URL; his next transactions are bundled with refunds from any backrunning value; he pays slightly higher priority fees but sees no sandwich slippage. The chapter develops Flashbots' mechanism in detail (the RPC override, the bundle delivery, the refund flow), and uses the trade as the anchor.

  • Pros: lets the chapter develop Ethereum's private-routing mechanism concretely; introduces a new named character (cleaner narrative arc; the book has used Alice across three chapters now); Flashbots is the most-explained-but-least-understood institution in the book.
  • Cons: makes Solana the secondary case, against Nick's stated direction.

Candidate C — Side-by-side: Alice on Solana via Beam, Carlos on Ethereum via Flashbots Protect, plus the L2 sequencer paragraph ​

Two named traders, three architectural responses, one anchor narrative.

  • Pros: maximum coverage of the chapter's content directive; lets each architecture get its own worked-example beat.
  • Cons: complex to thread; risks the chapter feeling like a survey rather than an argument.

Agent's recommendation ​

Candidate A (Alice + Beam) as the anchor, with Carlos (Candidate B's framing) returning as a named secondary example in the Ethereum subsection. This carries the Solana-heavy content directive the user gave, while letting the Ethereum subsection develop its own named-trader walkthrough. The L2 subsection and the Hyperliquid cameo in §6 use neither Alice nor Carlos — they're described in general terms because the visibility surface there is structural rather than tied to a specific trader's experience.

6. Open questions for Nick ​

Q1 — Word budget. The user's content directive expands what the chapter must cover (full Jito mempool history, BAM, Harmonic, Constellation, ETH private routing, MEV-Boost, L2 sequencers, Hyperliquid cameo, censorship trade-off). My estimate is ~5,500 words including footnotes, near the top of the Book Bible's 4,000–6,000 target. Is that acceptable, or do you want a tighter chapter (say 4,500 words) with some material moved to footnotes or a sidebar?

Q2 — The April 8, 2026 Solana patch mechanism. Chapter 4 already cites the patch. Chapter 3 needs to as well, but the exact technical mechanism isn't documented in a primary source I could find. Options: (a) describe the effect ("simple validator-leader sandwiches are largely suppressed") without committing to mechanism; (b) describe the BAM-routing-as-default hypothesis with explicit hedge; (c) skip the patch and treat BAM as the substantive response. My recommendation: (a). The patch is real and worth naming; the mechanism is a flag for a later research pass.

Q3 — Worked example. Candidate A (Alice + Beam, with Carlos as the Ethereum secondary) is the recommendation. Confirm or override.

Q4 — How adversarial is the verdict. The OUTLINE says this is the hinge chapter — adversarial content lands here. The verdict in §7 will be sharper than Chapters 1 or 2. Should it match Chapter 4's tone (clinical but explicitly damning of sandwiches as pure transfer) or stay slightly softer (the visibility surface is structural; what's done with it is the moral question)? My recommendation: match Ch 4's tone. The chapter has earned it.

Q5 — Constellation as forward-looking content. The chapter could end with a paragraph on Constellation (Solana's proposed MCP design) and FOCIL (Ethereum's inclusion lists) as the architectural answers being built. This is forward-looking, which the Bible says to use sparingly. Acceptable as a closing beat, or hold for Chapter 12 (Where This Is Going)?

Q6 — The Solana Foundation enforcement action. The June 2024 removal of ~32 validators is the cleanest "actor accountability" beat the book has in any chapter so far — a named foundation, a named spokesperson, a specific number of validators and SOL. Should the chapter develop this as a named-incident sidebar, or keep it as a paragraph inside the Solana subsection? My recommendation: paragraph inside the subsection. A full sidebar would imply ongoing enforcement importance the book doesn't claim.

Q7 — Hyperliquid weight. The SPEC scoped Hyperliquid as a cameo in the chain-comparison box (§6). The user's content directive named Hyperliquid explicitly as one of the chains to explain. Should Hyperliquid get more weight than the cameo (say a half-page subsection on "the architectural elimination of the mempool"), or does the chain-comparison treatment plus Chapter 9's full coverage suffice? My recommendation: keep as chain-comparison-box treatment, since Hyperliquid's structural choice is a one-paragraph point ("no public surface at all") and overdeveloping it here would steal from Chapter 9.

Sources cited ​

Primary:

News and aggregators:

Phase 1 is complete. Per the user's compressed-review pattern, Phase 2 (OUTLINE.md) follows immediately.