The House Edge

Appearance

Chapter 5 — Research Note (Phase 1) ​

Status: RESEARCH (Phase 1 — written by agent, awaiting Nick's review) Date: 2026-05-14 Working chapter: 05 — The Validator (and the Builder) Carries forward from: SPEC.md (strawman with explicit content directive from Nick to weight the Solana validator client + scheduler ecosystem heavily)

How to read this note ​

Three things to know up front:

  1. The Solana client + scheduler subsection is the chapter's substantive heart. Per Nick's content directive, the chapter develops Agave / Jito-Solana / JitoBAM / Frankendancer / Firedancer as a stack with measurably different revenue profiles. The Firedancer documentation publishes two named scheduler modes — perf and balanced — with explicit trade-offs. Harmonic's "Performance" and "Balanced" buckets are visible in Syndica's March 2026 telemetry but not formally documented in Harmonic's overview. Figment's own published migration to Frankendancer (epoch 871, 30 October 2025) gives the chapter concrete dollar-anchored numbers: +18 bps SRR (peak +28 bps), 9× MEV/Jito tips, 2.5× priority fees.

  2. Voting is now a strategic surface. Solana's Timely Vote Credits (SIMD-0033, activated November 2024 epoch 703) made low-latency vote landing a competitive moat. Chorus One reports a 2.03% skip rate versus a 5.19% network average — that is the moat operationalised. Alpenglow (SIMD-0326, approved 30 August 2025 with 98.27% stake, testnet 11 May 2026) replaces TowerBFT + PoH with Votor + Rotor, removes vote transactions from blockspace (~70% of tx at peak), and drops the profitable-validator threshold from ~4,850 SOL (~$800K) to ~450 SOL (~$75K).

  3. On Ethereum, the validator does not build the block. Live relayscan.io snapshots for May 2026 show Titan at ~51.5% of blocks, BuilderNet at ~24%, Quasar at ~15.3% — top three builders ~91% of blocks. The validator's job is to sign a block constructed by one of these three firms. ePBS — the proposed protocol-level fix — has slipped past Glamsterdam's H1 2026 target; the EF Checkpoint #9 (10 April 2026) flags Glamsterdam-in-Q2 as unlikely. Coinbase Cloud, the chapter's recommended worked-example anchor, runs 4.5M ETH (12.17% of network) with 99.98% participation rate and zero slashing events since inception.

1. Key claims ​

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

The validator's role, in 2026, is not the role it was

  1. On Ethereum since the September 2022 Merge, the validator is the proposer but typically not the block constructor. The proposer's job in 2026 is, in nearly all cases, to sign a block that a specialist firm — a builder — has constructed on its behalf via the MEV-Boost relay infrastructure. (Ethereum.org PoS docs; already cited in Ch 1)

  2. On Solana, the validator is the slot leader — the actor scheduled to produce the current block — and runs a validator client whose internal scheduler decides which transactions enter the block and in what order. The choice of client and scheduler is now measurably the validator's largest single revenue lever. (Anza — Introducing the Central Scheduler; Syndica — Deep Dive: Solana Onchain Activity, March 2026)

  3. On Hyperliquid, the validator is part of a small permissioned-but-stake-elected set running HyperBFT consensus. The chain's matching engine sits inside consensus; validators do not run separate block-builder infrastructure because no equivalent role exists. (Hyperliquid Documentation — Staking)

The Solana client stack in 2026

  1. Solana's stake-weighted validator-client distribution in March 2026 was: Agave Jito 32%, Agave JitoBAM 28%, Agave Harmonic 17%, Agave Rakurai 6%, Frankendancer (Jump Crypto's hybrid) 12%, Firedancer (Jump's full rewrite) 2%. Roughly 86% of stake runs Agave-based clients. (Syndica — March 2026 deep dive)

  2. Agave, the reference Solana client maintained by Anza, ships a default scheduler — central-scheduler-greedy — that uses 4 worker threads, 1 scheduling thread, and 2 vote threads. The scheduler prioritises transactions by the formula ((Priority Fee × CU Requested) + base fee) / (1 + CU Requested). Anza's own early benchmarking showed up to 80% higher fee collection with the central scheduler versus the prior thread-pool design. (Anza — Introducing the Central Scheduler)

  3. Jito-Solana is Jito Labs' Agave fork; JitoBAM is the BAM-enabled variant that delegates transaction sequencing to BAM Nodes running inside AMD SEV-SNP TEE enclaves while the validator handles execution. The JitoBAM share (28% of stake in March 2026) is the cleanest measurement of how much of the network has adopted BAM at the client level. (Helius — BAM)

  4. Frankendancer is Jump Crypto's hybrid client: it replaces Agave's networking layer with Firedancer's lower-level C/C++ networking stack while keeping Agave's consensus and execution layers. Frankendancer ran approximately 21% of stake on 7 October 2025 (community measurement) and approximately 12% by March 2026 (Syndica); the relative decline reflects validators migrating onward to JitoBAM and Harmonic rather than away from the Jump stack. (BlockEden — Firedancer at 21% Stake on Solana Mainnet, October 2025; Syndica March 2026 as cited)

  5. Firedancer, Jump Crypto's full C/C++ client rewrite, ran approximately 2% of mainnet stake in March 2026. Adoption has been slower than expected, with most operators preferring the Frankendancer hybrid for production use. (Syndica March 2026)

Frankendancer's scheduler modes — what they are

  1. The Firedancer / Frankendancer client documentation publishes two named scheduler modes for the pack tile (the component that decides transaction ordering within a slot):

    • perf mode: "fill the block as fast as possible using the highest-paying transactions available." Produces "consistently 100% full blocks." Optimised for maximum throughput and priority-fee capture per slot.
    • balanced mode (default): "optimizes for revenue from priority fees, but can result in blocks that are not always 100% full, and can be poor at capturing MEV if using an external block builder." Docs explicitly recommend reverting to balanced under network congestion. (Firedancer Documentation — Configuring)

  2. A third mode — revenue — appears in older documentation and some operator-side references, described as "prioritizes MEV capture and bundle processing… sorts bundles and transactions by fee-per-compute-unit." Whether it remains in the current source tree or has been folded into Jito-specific builds is not clear from the publicly accessible Firedancer docs. The chapter will hedge this. ([Firedancer Documentation — same; older references via operator notes])

Harmonic's scheduler modes — observed in telemetry

  1. Harmonic is an aggregation layer (full mechanical treatment in Chapter 3) where validators express configurable preferences to select the most valuable block per slot from a pool of candidate builders (Jito, Temporal, JitoBAM, Paladin). The validator preferences shape what is selected. Syndica's March 2026 telemetry reports three distinct Harmonic configurations with measurably different per-block priority-fee captures versus the network median:

    • Frankendancer Harmonic Performance: +101% priority fees vs median
    • Frankendancer Harmonic Balanced: +39%
    • Agave Harmonic: +36%

    Block-time-adjusted: Agave Harmonic +33%, Frankendancer Harmonic Balanced +32%, Frankendancer Harmonic Performance +28%. The pattern strongly suggests Performance prioritises maximum-paying-transaction selection at cost of latency (~618ms blocks), while Balanced caps block time. Harmonic's public overview does not formally document the Performance/Balanced trade-off; the mode names are visible in operator telemetry. (Syndica March 2026; Harmonic — Overview)

  2. Agave Rakurai (which Nick has asked the chapter to skip) leads Jito tips at +250% versus median per Syndica's March 2026 data, illustrating that scheduler-level optimisation can shift different revenue streams disproportionately. (Cited for completeness; not developed in the chapter.)

The client choice as revenue lever — Figment's published migration

  1. Figment migrated its main Solana validator to Frankendancer at epoch 871 (30 October 2025). The published results from the firm itself:
    • Gross Staking Reward Rate up +18 bps (peak +28 bps)
    • MEV / Jito tips up 9×
    • Priority fees up 2.5×
    • Non-vote transactions per block up 57%
    • Total CU per block up 20%
    • Revenue per CU doubled (0.600 → 1.200)
    • Trade-offs: block duration up 18% (median 355.7ms → 398.4ms); failed transactions per block from 41 → 145. (Figment — Migration to Firedancer, late 2025 / early 2026)

Voting on Solana is now a strategic surface

  1. Timely Vote Credits (SIMD-0033) activated in November 2024 (epoch 703). Under TVC, validators earn up to 16 vote credits for votes confirmed within 2 slots of inclusion; one fewer credit per additional slot of latency; floor of 1 credit at ≥18 slots. The rule converted vote-landing-latency from a constant cost into a competitive moat: validators with better-positioned infrastructure and lower-latency clients now earn measurably more rewards. (Anza — Timely Vote Credits)

  2. The operational consequence of TVC is observable in skip rates. Chorus One's Solana skip rate is approximately 2.03%, versus a network average of 5.19% and a superminority-validator average of 5.68%. The gap is a function of hardware quality, geographic placement, and client configuration. (Chorus One — Metrics that Matter)

Alpenglow rewrites Solana's consensus and changes the validator economics

  1. Alpenglow (SIMD-0326) was approved by Solana's validator set on 30 August 2025 with 98.27% stake-weighted approval. The proposal replaces the existing TowerBFT + Proof-of-History consensus with a two-component design: Votor, which moves vote transactions off-chain using BLS-aggregated signatures, and Rotor, which replaces the existing block-propagation layer. Two-path finality: ≥80% stake confirmation in round 1 yields a Fast Finalization Certificate at approximately 100ms; otherwise ≥60% in round 2 yields a Finalized Certificate at approximately 150ms. Alpenglow's testnet activation was 11 May 2026; mainnet is expected in late Q3 or early Q4 2026. (SIMD-0326 — Alpenglow proposal; CoinDesk — Biggest consensus overhaul live for testing, 2026-05-11; Helius — Alpenglow)

  2. Alpenglow has three load-bearing consequences for validator economics. First, vote transactions are removed from blockspace — at peak load, vote transactions consume approximately 70% of all Solana transactions; their removal liberates that capacity for fee-paying transactions. Second, Alpenglow introduces a Validator Admission Ticket (VAT) of approximately 0.8 SOL per day (1.6 SOL per epoch), entirely burned, which is the first explicit per-epoch fixed cost on the network. Third, Helius estimates that VAT economics drop the profitable-validator stake threshold from approximately 4,850 SOL (~$800,000) to approximately 450 SOL (~$75,000), meaningfully widening who can run a validator profitably. (Helius — Alpenglow)

The Ethereum builder oligopoly is sharper than the Ethereum validator concentration

  1. Live MEV-Boost relay/builder telemetry for May 2026 (relayscan.io 24-hour snapshot):

    • Builder shares: Titan 51.53% (3,410 blocks, 57.46 ETH profit); BuilderNet 24.07% (1,593 blocks, 18.74 ETH); Quasar 15.34% (1,015 blocks); Beaverbuild 1.83%.
    • Top three builders = ~91% of MEV-Boost blocks. Beaverbuild's flow has migrated into BuilderNet, collapsing its standalone share.
    • Relay shares: Ultrasound 35.71%, Titan Relay 26.09%, BloXroute Max-Profit 13.38%, BloXroute Regulated 12.30%, Aestus 7.32%, Flashbots Relay 2.09%. (relayscan.io)

  2. Titan Builder is operated by Kubi Mensah (the publicly named lead) and a team described as ex-miners, ex-validators, ex-searchers, and ex-traders. Titan's published profit margins under exclusive flow arrangements (e.g., its agreement with the Banana Gun bot) have reached 17.75%, materially higher than Beaverbuild's pre-migration ~9% margin. Titan's launch public RPC went live on 17 April 2023. (Observers — How Two Block Builders Monopolized Ethereum Block Production; Frontier Research — Builder Dominance and Searcher Dependence; Scraping Bits #16 — Kubi Mensah: Titan's Takeover)

  3. BuilderNet launched on 18 November 2024 as a multi-operator builder protocol with founding operators Flashbots, Beaverbuild, and Nethermind. Each instance runs in a TEE; order flow is shared equally among operators; an "open source refund rule" redistributes MEV value back to order-flow providers based on participation in winning blocks. Onboarding is currently permissioned, which has prompted criticism that the protocol's decentralisation properties are partial. (Flashbots — Introducing BuilderNet; EigenPhi guest post — BuilderNet Infrastructure Monoculture)

ePBS has slipped

  1. ePBS (EIP-7732) — the protocol-level enshrinement of proposer-builder separation that would replace the out-of-protocol MEV-Boost relay infrastructure — was selected as the headliner for Ethereum's Glamsterdam upgrade, originally targeted at H1 2026. As of the Ethereum Foundation's April 2026 Checkpoint #9, Glamsterdam has not shipped: "Glamsterdam in Q2 seems to me to be unlikely." Implementation has proved trickier than anticipated, with partial blocks and two-party consensus coordination as documented bottlenecks. No firm replacement target date has been published. (Ethereum Foundation — Checkpoint #9: April 2026; QuickNode — Ethereum Glamsterdam)

Ethereum validator economics in 2026

  1. Figment's Q3 2025 Ethereum validator report (most recent published) gives:

    • Median consensus-layer reward per validator: 0.002029 ETH/day.
    • Median execution-layer reward when proposing: 0.0176 ETH (a single-day spike).
    • CL ~93% / EL ~7% of total annualised reward composition (the EL share is small because proposal is rare — roughly once every 150 days per 32-ETH validator).
    • Network SRR (Staking Reward Rate) average: ~2.94%.
    • 30% of ETH supply staked.
    • 56 slashing events network-wide in Q3 2025; zero on Figment. (Figment — Q3 2025 Ethereum Validator Report)

  2. Independent quant analysis from Block Scholes: without MEV-Boost, validator yield averages approximately 4% APY; with MEV-Boost, approximately 5.69% APY. Priority fees account for approximately 17% of total validator income; MEV-derived contribution is approximately 3% APY. (Block Scholes — Ethereum Staking Deep Dive)

  3. The Pectra upgrade activated on 7 May 2025 raised the maximum effective balance per Ethereum validator from 32 ETH to 2,048 ETH via EIP-7251. The change reduced network message overhead by approximately 40% and enabled large operators to consolidate stake into fewer, larger validators — which in turn allows them to capture MEV more efficiently per validator. (Consensys — Ethereum Pectra Upgrade; The Block — Ethereum Pectra Upgrade Activation)

Coinbase Cloud as the chapter's worked-example validator

  1. Coinbase Cloud runs approximately 4.5 million ETH of staked balance — 12.17% of the entire Ethereum network's stake — as of Q1 2026. The operation has a 99.98% participation rate versus a network average of 99.77%, zero slashing events since inception, and a self-imposed 30% network-share cap. Validators are distributed across Germany, Hong Kong, Ireland, Japan, and Singapore on AWS and GCP infrastructure. Optional OFAC SDN filtering is offered to enterprise clients. (gncrypto — Coinbase Validators 99.98% Uptime, Q1 2026)

  2. Kiln is the most-diversified institutional staking-as-a-service operator, running 50,000+ validators across 50+ networks with $14B+ assets under stake. Network shares: 6% of Ethereum, 2.5% of Solana. Kiln powers staking inside VanEck, 21Shares, CoinShares, Ledger, Crypto.com, and Trust Wallet; it is an approved node operator in Lido V3 stVaults. (Kiln — Staking Rewards profile; Kiln — Ethereum protocol page)

Hyperliquid's validator set

  1. Hyperliquid runs delegated proof-of-stake on HyperCore. The chain's active validator set is the top 21 by stake (transitioned from a smaller restricted set to a permissionless top-21 design). Self-delegation requirement is 10,000 HYPE locked for one year. Validator reward rate is inversely proportional to the square root of total HYPE staked; at ~400M HYPE staked, the APR is approximately 2.37%. Validator commissions cannot be raised more than 1% without a timelock. (Hyperliquid Documentation — Staking)

  2. HLP is structurally separate from the validator stack. HLP earns 3% of trading fees on the venue (the remaining 97% routes to HYPE buybacks); validators earn issuance and commissions on delegated stake but no direct HLP allocation. Recent institutional validator launches (e.g., Hyperliquid Strategies + Unit Labs, announced 7 May 2026, going live 11 May 2026) reflect rising institutional interest in the permissionless set. (DeFiLlama — Hyperliquid HLP; PRNewswire — Hyperliquid Strategies + Unit Labs validator launch)

Slashing — real but tail-risk

  1. Across both major proof-of-stake chains, slashing in 2025 was real but tail-risk. Fewer than 500 validators were slashed on Ethereum from the Beacon Chain's December 2020 launch through September 2025, against an active validator count over 1.2 million in 2025. Q3 2025 alone produced 56 slashing events. The single largest 2025 cluster was a 39-validator simultaneous slashing on 10 September 2025, traced to SSV Network distributed-validator-technology operator misconfiguration during Ankr maintenance. (CoinDesk — Ethereum Rare Mass Slashing Event, 2025-09-10; SSV Network — September 2025 Slashing Post-Mortem)

  2. Solana has no stake-slashing equivalent for vote-skipping. Penalties take the form of reduced vote credits → reduced inflation-share rewards — the economic discipline mechanism is opportunity cost rather than stake destruction. Alpenglow's VAT (1.6 SOL burned per epoch) is the first explicit per-epoch fixed cost on the network. (Anza — Timely Vote Credits docs)

2. Numbers to verify ​

#NumberSourceDateFlag
N1Solana client stake share (Mar 2026): Agave Jito 32%, JitoBAM 28%, Harmonic 17%, Rakurai 6%, Frankendancer 12%, Firedancer 2%; 86% on AgaveSyndica March 20262026-03Load-bearing for the chapter
N2Frankendancer Harmonic Performance +101% priority fees vs median; Balanced +39%; Agave Harmonic +36%Syndica March 20262026-03Cited in Ch 3; reuse here
N3Agave central-scheduler-greedy uses 4 worker + 1 scheduling + 2 vote threads; up to 80% higher fee collection vs prior schedulerAnza — Central Schedulerv1.18 era / 2024Primary
N4Frankendancer perf and balanced scheduler modes documented; revenue mode referenced in older docsFiredancer Configuring docscurrentPrimary; the revenue-mode question is flagged
N5Figment Frankendancer migration (epoch 871, 30 Oct 2025): +18 bps SRR (peak +28); MEV/Jito tips up 9×; priority fees up 2.5×; CU/block up 20%; block duration up 18%Figment migration report2025-Q4 / 2026-Q1Strong; published by the operator itself; the cleanest "client choice = revenue moat" stat
N6Timely Vote Credits: max 16 credits within 2 slots; floor 1 credit at ≥18 slotsAnza — TVC docsactivated Nov 2024 epoch 703Primary
N7Chorus One Solana skip rate 2.03% vs 5.19% network avg vs 5.68% superminorityChorus One — Metrics that Matter2025Strong primary; the moat operationalised
N8Alpenglow approved 98.27% stake-weighted, 30 Aug 2025; testnet 11 May 2026; mainnet late Q3/early Q4 2026SIMD-0326; CoinDesk 2026-05-112025-08 / 2026-05Strong
N9Alpenglow finality: ≥80% in round 1 = ~100ms; ≥60% in round 2 = ~150msHelius — Alpenglow2026Primary analysis
N10Alpenglow VAT: ~0.8 SOL/day (1.6 SOL/epoch), entirely burned; profitable-validator threshold drops from ~4,850 SOL (~$800K) to ~450 SOL (~$75K)Helius — Alpenglow2026Helius estimate
N11Vote transactions consume ~70% of Solana tx at peak; Alpenglow removes them from blockspaceHelius — Alpenglow2026Helius estimate
N12Ethereum builder share (May 14 2026): Titan 51.53%, BuilderNet 24.07%, Quasar 15.34%, Beaverbuild 1.83%; top 3 ~91%relayscan.io2026-05-14Live; will drift
N13Ethereum relay share (May 14 2026): Ultrasound 35.71%, Titan 26.09%, BloXroute Max 13.38%, Reg 12.30%, Aestus 7.32%, Flashbots 2.09%relayscan.io2026-05-14Live
N14Titan profit margin (exclusive Banana Gun arrangement) up to 17.75%; Beaverbuild ~9% pre-BuilderNetObservers; Frontier Research2025Strong but Beaverbuild now in BuilderNet; the standalone comparison is partly historical
N15BuilderNet launched 18 Nov 2024 with Flashbots, Beaverbuild, Nethermind as founding operatorsFlashbots — Introducing BuilderNet2024-11-18Primary
N16ePBS / Glamsterdam not shipped as of Apr 2026; "Q2 seems unlikely"Ethereum Foundation Checkpoint #92026-04-10Primary; the EF's own admission
N17Ethereum validator: median CL reward 0.002029 ETH/day; median EL reward when proposing 0.0176 ETH; CL/EL = 93%/7%; SRR ~2.94%; ~30% of ETH supply stakedFigment Q3 2025 Ethereum report2025-Q3Strongest published numbers we have
N18Ethereum validator APY: ~4% no MEV-Boost; ~5.69% with; priority fees ~17% of total income; MEV contributes ~3% APYBlock Scholes2025Quant primary
N19Pectra upgrade (May 7 2025) raised max effective balance to 2,048 ETH; reduced network message overhead ~40%Consensys2025-05-07Primary
N20Coinbase Cloud: 4.5M ETH (12.17% of network) staked Q1 2026; 99.98% participation; zero slashing since inception; 30% self-capgncrypto Coinbase report2026-Q1Strong; ideal worked-example anchor
N21Kiln: 50,000+ validators / 50+ networks / $14B+ AUM; 6% of Ethereum, 2.5% of SolanaKiln Staking Rewards profile2026Strong
N22Hyperliquid: top-21 active validator set; 10,000 HYPE self-delegation, 1-year lock; APR ~2.37% at 400M HYPE stakedHyperliquid Staking docscurrentPrimary
N23HLP earns 3% of trading fees (97% to HYPE buybacks); historical ~1.75%/month (~20% annualised); separate from validator stackDeFiLlama HLP2026Strong
N24<500 ETH validators slashed total since Dec 2020 launch (>1.2M active validators in 2025); 56 slashing events Q3 2025; 39-validator SSV cluster on 10 Sep 2025CoinDesk; SSV post-mortem2025-09Strong; the cleanest "slashing is tail-risk" cite
N25Jito Labs takes 6% of BAM tips; split 50/50 with Jito DAO; pending JIP would route 100% to DAO; Q2 2025 DAO share ~22,391 SOL (~$4M)Helius — BAM2025-Q2 onwardAlready in Ch 3; refresh here

Deliberately not pinned:

  • Exact validator/staker percentage split inside BAM-mediated MEV — Helius and Jito Labs disclose the 6% Jito Labs/DAO take but the validator share specifically is not in any source I could pin.
  • Mainnet stake share of pure Firedancer (not Frankendancer) as of May 2026 — Syndica's March 2026 puts it at 2%; no April/May 2026 update found.
  • Titan's full corporate structure beyond Kubi Mensah as named lead.

3. Contested or evolving claims ​

  • Whether the Firedancer revenue scheduler mode is current or deprecated. The current Firedancer docs page lists only perf and balanced. References to revenue exist in older documentation and operator notes but cannot be confirmed against the latest source tree. The chapter will describe perf and balanced precisely from the docs and note revenue as a mode that may exist in some Jito-integrated builds, with explicit hedge.

  • Harmonic's "Performance" / "Balanced" naming. Visible in Syndica's March 2026 telemetry as buckets that validators self-select. Their existence is observable in the data; Harmonic's public docs do not formally describe the trade-offs as named modes. The chapter will state what's observable (the +101% / +39% / +36% priority-fee captures and the implied latency/revenue trade-off) and note that the formal documentation has not caught up to the operator practice.

  • ePBS timeline. The Ethereum Foundation's April 2026 Checkpoint #9 explicitly says Glamsterdam Q2 is unlikely; no replacement target has been published. The chapter will describe what ePBS does and the structural reason it matters, but will treat its mainnet timing as open.

  • Alpenglow's pre-Alpenglow vs post-Alpenglow validator economics. The 0.36% MEV-as-share-of-staking-rewards figure from Figment Q1 2026 is the pre-Alpenglow number. Post-Alpenglow (expected late Q3 / early Q4 2026), the removal of vote transactions from blockspace and the introduction of the VAT will shift the economics — but the magnitude is not yet measurable. The chapter will state the pre-Alpenglow numbers clearly and frame Alpenglow's impact as a forward-looking but defined event.

  • Builder concentration vs builder competition. Titan's 51.5% share looks like dominance; BuilderNet's 24% share is multi-operator. The competitive equilibrium between these two designs (single-firm vs multi-operator-with-TEE) is the open structural question. The chapter will not predict the winner but will state the structural choice clearly.

4. Characters introduced ​

Two full "Meet the actor" sidebars:

The Validator — full sidebar, six fields per the Ch 2 (Market Maker) and Ch 4 (Searcher) template:

  • Who they are: An operator (often a firm; sometimes a single technically-sophisticated individual) who stakes the chain's native token, runs validator software, and participates in block production for revenue.
  • How they earn: Base issuance (chain-defined inflation share); priority fees from the transactions they include; MEV-derived revenue (MEV-Boost payments on Ethereum; Jito tips and BAM tips on Solana); validator commissions if running staking-as-a-service.
  • How they spend: Hardware (high-spec servers, multiple geographies); bandwidth (the modern bottleneck on Solana); engineering time on client choice, scheduler configuration, voting strategy; opportunity cost of locked stake; slashing-loss tail risk.
  • The moat: Increasingly, not raw uptime. The moat is the choice of validator client (Jito-derivative vs Frankendancer vs Firedancer on Solana; Lighthouse vs Prysm vs Teku on Ethereum); the choice of scheduler mode where the client supports configurable modes; the relationships with block-construction firms (Titan, BuilderNet, Quasar on Ethereum; Jito, Temporal on Solana); and the geographic placement of the infrastructure.
  • TradFi analogue: A specialist on the floor of the NYSE circa 1995 — the actor with privileged information access who handled order matching for one or more tickers, earning a share of the spread, with operational discipline as the visible criterion but information relationships as the actual moat.

The Builder — separate sidebar, parallel format:

  • Who they are: A firm whose software constructs the actual block content from the available transaction flow. On Ethereum, builders are specialised institutions (Titan, BuilderNet, Quasar, Beaverbuild) running low-latency hardware and dedicated relationships with searchers, RPCs, and order-flow providers. On Solana, the closest equivalent role is split between Jito's Block Engine (the bundle auction), the BAM Node operators (the TEE-enclave block constructors), and the Harmonic block-builder ecosystem.
  • How they earn: The wedge between the value of the block they construct (sum of priority fees + MEV captured) and the bid they pay the validator to win the validator's signature on that block. Profit margins on Ethereum reach 17.75% under exclusive flow arrangements; the median margin is harder to pin without primary disclosure.
  • How they spend: Low-latency hardware co-located with relays; in-house transaction simulators; engineering teams to write the bundle-merging and conflict-resolution logic; private order-flow relationships with searchers and RPCs.
  • The moat: Access to private order flow — Chapter 3's argument lands here. The builder that sees the most pending transactions earliest constructs the most valuable block.
  • TradFi analogue: A payment-for-order-flow wholesaler (Citadel Securities, Virtu Financial) — the actor that handles retail flow not because the broker is incompetent but because the wholesaler has scaled relationships with venues that retail brokers cannot match.

Returning named institutions:

  • Jito Labs (Solana; full treatment Ch 3)
  • Helius (Solana; full treatment Ch 3)
  • Anza (Solana; the foundation team maintaining Agave)
  • Flashbots (Ethereum; full treatment Ch 3)
  • BuilderNet (Ethereum; cameoed Ch 3)
  • Coinbase Cloud (Ethereum; introduced for the first time in this chapter; the worked-example anchor)
  • Kiln (Ethereum / Solana / multi-chain; cited in the institutional landscape paragraph)
  • Lido (Ethereum; cited as a curated-operator-set protocol)
  • Jump Crypto (Solana; introduced here for the first time as the Frankendancer / Firedancer team)
  • The Solana Foundation (already named Ch 3)
  • The Ethereum Foundation (named here)

Named individuals:

  • Kubi Mensah (Titan, the publicly named lead)
  • Tim Garcia (Solana Foundation; already named Ch 3)
  • Anatoly Yakovenko ("toly", Solana Labs; possibly cited if the chapter develops the leader-schedule design)
  • Brennan Watt (Anza; possibly cited on Alpenglow)
  • Vitalik Buterin (Ethereum Foundation; on ePBS in the forward-looking section)
  • Hasu (Flashbots / SMG; already cited Ch 3)
  • Justin Drake (Ethereum Foundation; on ePBS)

5. Worked example candidates ​

The SPEC offered three. The research narrows the choice.

Candidate A — A single Ethereum block, proposed by Coinbase Cloud, built by Titan ​

Recommended. Coinbase Cloud runs 12.17% of Ethereum's stake and is the cleanest non-Figment institutional validator to anchor a worked example to. Titan builds 51.5% of Ethereum blocks. The chapter walks a single block's economics: a searcher submits a bundle to Titan via Flashbots Protect; Titan constructs a block including the bundle, paying Coinbase Cloud (as proposer) the MEV-Boost bid; Coinbase Cloud signs; the block is finalised; Coinbase distributes the rewards to its staking clients after taking its commission. Each step has a published dollar number we can cite. Pros: Both actors named with primary-source data; the Coinbase Cloud worked-example preserves the Figment-named-only-in-epilogue rule; the four-actor flow (searcher → builder → validator → staker) lets the chapter develop PBS concretely. Cons: Ethereum-only; Solana and Hyperliquid cases need their own paragraph cameos.

Candidate B — Figment's published Frankendancer migration ​

The Figment migration report (epoch 871, 30 October 2025) gives the cleanest "client choice = revenue moat" empirical case the book has: +18 bps SRR, 9× MEV/Jito tips, 2.5× priority fees on a single client switch. This is rich, specific, and published.

The SPEC's editorial constraint, however, says: don't name Figment as the worked-example validator. The book's commercial framing depends on Figment's restraint about naming itself outside the epilogue.

Resolution: Use Figment's published migration as a named secondary cite in the Solana subsection — not as the worked-example anchor, but as a sourced data-point illustrating the client-choice-as-revenue-lever thesis. The chapter cites Figment's research the way it cites Helius's research: as institutional commentary in footnotes.

Candidate C — Side-by-side Coinbase Cloud / Solana client / Hyperliquid validator ​

Rejected as the anchor; partial elements absorbed. Candidate A handles Ethereum cleanly; the Solana subsection develops the client-stack story without needing an anchor trader; the Hyperliquid paragraph in §6 covers the architectural case briefly.

Agent's recommendation ​

Candidate A (Coinbase Cloud + Titan) as the worked-example anchor. Figment's migration appears as a sourced footnote citation in the Solana subsection. Solana and Hyperliquid get paragraph cameos in the chain-comparison box at §6.

6. Open questions for Nick ​

Q1 — Word budget. The user's content directive (full Solana client stack + scheduler modes + voting strategies) makes this another ~5,500-word chapter, like Ch 3. Acceptable, or do you want a tighter version with some material moved to footnotes?

Q2 — The revenue scheduler mode question. The Firedancer docs publish perf and balanced; older references include revenue. The chapter will describe what's documented and flag the revenue mode as potentially folded into Jito-integrated builds. Preference?

Q3 — Worked example. Candidate A (Coinbase Cloud + Titan, Ethereum) is the recommendation. Figment migration data appears as a secondary cite in Solana subsection footnotes. Confirm or override.

Q4 — Alpenglow framing. Solana's Alpenglow consensus rewrite is genuinely a major change — vote-tx removal from blockspace, profitable-validator threshold drop, new finality timeline. Should the chapter develop Alpenglow's validator-economics implications as a substantial subsection of the Solana coverage, or treat it more briefly with a forward-link to Ch 12? My recommendation: substantial coverage (~300 words). The shift is dated, sourced, and load-bearing for the chapter's "the validator role is changing again" thesis.

Q5 — Two sidebars or one. The SPEC scoped two: Meet the Validator + Meet the Builder. Given the chapter's length (~5,500 words) and the depth of the Solana client material, I think both sidebars are worth keeping — but if you want to shorten, the Builder sidebar can be folded into the Ethereum subsection prose. My recommendation: keep both.

Q6 — Tone on Coinbase Cloud + OFAC filtering. Coinbase Cloud explicitly offers OFAC SDN-list filtering as an optional service to enterprise clients. Chapter 3 already cited Base's OFAC filtering. Should the chapter develop the OFAC-and-censorship-resistance angle as a paragraph, or note it once and defer to Ch 11? My recommendation: one sentence in the worked example, with a forward-link to Ch 11.

Q7 — Slashing as a chapter beat. The 39-validator SSV slashing on 10 September 2025 is the cleanest concrete slashing-incident the chapter can cite. Should it appear as a sidebar moment, an inline paragraph, or just a footnote? My recommendation: an inline paragraph (~80 words) in the "operational excellence is still a moat" framing.

Sources cited ​

Primary research and protocol docs:

News and aggregators:

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