The House Edge

Appearance

Chapter 6 — Outline (Phase 2) ​

Status: OUTLINE (Phase 2 — written by agent, awaiting Nick's review) Date: 2026-05-14 Working chapter: 06 — The Infrastructure Layer Carries forward from: RESEARCH.md

Working assumptions (carried from Phase 1) ​

Defaults from RESEARCH.md §6, treated as binding for Phase 3 unless Nick overrides.

  1. Word budget: ~7,500 words including footnotes. Over the Bible's 6,000 target — the chapter has 15+ named firms to develop and the structural argument benefits from depth.
  2. JIP-24 framing: paragraph-level treatment in the Jito subsection. The Jito-Labs-to-DAO transition is the cleanest case in the book of an infrastructure firm voluntarily restructuring its revenue.
  3. Eden Network mention: one sentence as "the firms that did not survive the consolidation" beat.
  4. Ben Coverston / Temporal-Harmonic overlap: paragraph-level treatment in the Solana subsection. The CEO-of-two-firms fact is structurally striking.
  5. Quasar's unknown founders: cite the opacity explicitly. "A firm running 15% of Ethereum blocks with no publicly identifiable operators" is itself the chapter's structural point.
  6. Aggregate take rate: build bottom-up estimate with explicit methodology in the verdict.
  7. Back-edits to Chs 3 and 5: small JIP-24 corrections to footnotes only; flagged in REVIEW_NOTES.md.
  8. Worked example: paired Helius + Flashbots profiles as the anchor, with Jito Labs / Titan / BuilderNet / bloXroute developed in the subsection prose.

Title and subtitle ​

Chapter 6 — The Infrastructure LayerThe firms you've never heard of that route most of the order flow — and what they sell, to whom, and why.

Cold open ​

If a Solana retail trader opens her Phantom wallet, signs a swap, and watches the transaction confirm two seconds later, she has interacted with — at minimum — Phantom (the wallet), Helius (the wallet's RPC backend), Jupiter (the aggregator), Jito Labs (the bundle market the aggregator routed through), Jito-Solana (the validator client her chosen leader is running), and possibly BAM (the encrypted block-construction enclave). Five named firms touched her trade between her click and its inclusion. Phantom is the only one whose name appears on her screen. The other four are infrastructure firms: companies whose business is to route, sequence, build, or refund trading flow on behalf of the firms the trader does see. This chapter is about who they are, what they sell, and why their existence is — in the OUTLINE's framing of this chapter, written by a different person at a different time — "the most under-appreciated fact about on-chain trading."

(Why this open: continues the chapter's named-trader convention from prior chapters; lands the structural argument in the first paragraph; the five-named-firms framing is specific and the reader can verify each name in the chapter's prose.)

What this chapter answers ​

  • Which firms operate the major layers of the on-chain trading infrastructure stack on Solana and on Ethereum, and what does each one actually sell?
  • Who buys at each layer, and what is the take rate?
  • Why have these firms emerged on top of "permissionless" chains that, in theory, no one needed to operate?
  • How big is this infrastructure layer in aggregate — what fraction of every on-chain trading dollar passes through it?

Section list ​

  1. Cold open — the five-firms-touched-one-trade scene.

  2. What this chapter answers — the four questions above.

  3. The setup (≈500 words). What an "infrastructure firm" is. The four layers the chapter develops (RPC, builder, relay, OFA). The structural argument: each layer has a small number of firms; the firms aren't competing across layers but occupy specific positions. Plant: paired Helius / Flashbots profiles will anchor the chapter, with the other firms summarised against them.

  4. The worked example (≈500 words). Helius and Flashbots as the chapter's two anchor firms. Helius: founded 2022 by Mert Mumtaz, Liam Vovk, Nicolas Pennie; $67.6M raised across Series A and B; operates at six layers of the Solana stack (RPC, streaming, transaction submission, validator, LST, BAM Node); customers include Phantom, Backpack, Solflare, Bitwise. Flashbots: founded 2020 by Phil Daian and four co-founders; SUAVE archived in May 2025; current product stack is Protect (2.1M+ accounts, $43B+ shielded), MEV-Share, BuilderNet contribution. Both firms occupy multiple layers; layer occupation is the business.

  5. The mechanics, in detail (≈4,500 words; four H4 subsections — at the chapter spec template's hard limit):

    • The RPC providers (≈900 words). The layer at which wallets and aggregators talk to the chain. Helius (developed in full), Triton One (Solana, BAM Node operator), Astralane (Solana HFT-tier infrastructure), and the legacy multi-chain providers (QuickNode, Alchemy, Infura) in cameo. The structural observation: RPC is increasingly the layer where MEV rebates are first captured — Helius's 50/50 split with wallet customers is the canonical case. Helius's customer list (Phantom, Backpack, Solflare, Bitwise's exclusive SOL ETF) means that even the wallets that compete on user-facing UX share the same backend pipeline.

    • The block builders and bundle markets (≈1,300 words). On Ethereum: Titan (51.55% of blocks; operated by Gattaca; Kubi Mensah CEO; no public funding round disclosed; 17.75% margin under Banana Gun exclusive), BuilderNet (24.09% of blocks; multi-operator with Flashbots / Beaverbuild / Nethermind; Beaverbuild fully migrated 6 May 2025; permissioned onboarding criticised by EigenPhi), Quasar (15.29% of blocks; founders not publicly disclosed — the chapter develops this opacity as load-bearing). On Solana: Jito Labs as the firm with full Jito Block Engine + BAM treatment; JIP-24 routing 100% of Block Engine + BAM tips to the Jito DAO (changed from the 3%/3% split that operated through Q2 2025); Temporal as the firm reportedly behind HumidiFi (and Nozomi's transaction-submission service); Paladin as the validator-side anti-sandwich bot integrated by Chorus One. The Ben Coverston paragraph lands here — same individual is CEO of Temporal (HumidiFi + Nozomi) and Harmonic (block-building marketplace), one of the chapter's most structurally significant single-operator concentration facts.

    • The relays and routing infrastructure (≈900 words). The Ethereum relay landscape in May 2026: Ultrasound Money (~35% of payload), Titan Relay (~27%), bloXroute Max-Profit (~13%) + bloXroute Regulated (~12%) = ~25% combined, Aestus (~7%), Flashbots Relay (~2%). bloXroute developed as a firm: Uri Klarman CEO; founded 2018; $89.6M total funding; $70M Series B from SoftBank Vision Fund 2 (April 2022); operates both a regulated/OFAC-filtered relay and a neutral/profit-maximising relay — structurally unique among major infrastructure operators. Aestus as the non-profit community-run alternative serving 650K+ validators. Hasu's "neutral relay" critique from Chapter 3 lands here as institutional context.

    • The Order Flow Auction operators and refund services (≈800 words). Flashbots' OFA stack: Flashbots Protect (2.1M+ accounts), MEV-Share (selective disclosure), the SUAVE archival in May 2025. MEV-Blocker's transfer from CoW DAO to Consensys' Special Mechanisms Group (SMG) in January 2026; what SMG is (Consensys research division focused on blockchain microstructure); the 4.5M-user / 6,177-ETH-refunded inheritance. Kolibrio's RPC-layer OFA (Chapter 3). CoW DAO's solver-auction business: 73M+ transactions, hundreds of billions in protected volume, 50% price-improvement share + 0.1% volume fee, launched on Ink (Kraken's L2) on 17 February 2026.

  6. Meet the Infrastructure Provider sidebar (≈300 words). Instantiated against Helius — the cleanest single example of a firm spanning multiple layers of one chain's stack. Six standard fields: Job / How they earn / How they spend / The moat / TradFi analogue / Founded.

  7. How this plays out on each chain (≈400 words). Three paragraphs.

    • Solana: Helius at the RPC/streaming/validator layer + Jito Labs across Block Engine and BAM + Anza maintaining Agave + Jump Crypto maintaining Frankendancer/Firedancer + Harmonic as block-building marketplace + Rakurai + the prop-AMM operator firms (Temporal/HumidiFi). The structural observation: most Solana infrastructure firms span multiple layers; the same individuals appear in multiple firms; concentration runs across layers as much as within them.
    • Hyperliquid: structurally different. The chain runs its own validator set, its own matching engine, its own HLP vault — there is no significant third-party infrastructure layer the way Solana and Ethereum have. The infrastructure firms in the rest of the chapter do not operate on Hyperliquid because the architecture does not need them. Chapter 9 develops what that means for the chain's competitive position.
    • Ethereum: a wider, more functionally-specialised infrastructure layer than Solana's. Flashbots at the OFA/relay layer + Titan/Gattaca at the builder layer + BuilderNet as the multi-operator alternative + bloXroute spanning relay and BDN + Consensys/SMG at the OFA layer + CoW DAO at the solver-auction layer + the L2 sequencer firms (Coinbase, Offchain Labs, Optimism Foundation) reserved for Chapter 10.

  8. Who wins, who loses, why (≈500 words). Adversarial verdict.

    • Winners: the infrastructure firms themselves — Helius (private; revenue undisclosed but layered subscriptions + validator stake + rebate flows + ETF partnerships), Jito Labs / Jito Foundation (now with the entire 6% Block Engine + BAM take routed to the DAO post-JIP-24), Titan (51.55% of Ethereum blocks at ~17.75% margin under exclusive flow), bloXroute ($89.6M raised; 25% combined Ethereum relay share), Flashbots (2.1M+ Protect users; BuilderNet co-operator). The bottom-up aggregate: this layer captured somewhere in the mid-hundreds-of-millions to low-billions of US dollars across both chains in 2026.
    • Losers: the firms that did not survive the consolidation — Eden Network wound down on 12 August 2025; Beaverbuild's standalone business retired into BuilderNet on 6 May 2025. The retail traders, indirectly: every layer of the stack takes a cut, and the cut is what the trader pays at the venue level. The validator-as-a-service operators without infrastructure relationships, whose ability to compete is constrained by which builders and relays they can access (Chapter 5).
    • Is this bad?: clinical. The infrastructure firms emerged because there were real operational gaps on permissionless chains — running RPC at scale is expensive; constructing optimal blocks is hard; refunding MEV requires sophisticated routing logic. The firms that filled the gaps captured the resulting margin. The structural concern is concentration: Helius's six-layer Solana presence, Titan's 51%-of-Ethereum-blocks dominance, Ben Coverston's CEO-of-two-firms position, Quasar's anonymous 15%-builder share — all of these are products of the same dynamic, and all of them mean the trader's question "who am I trading against?" has an answer that the trader cannot easily discover.

  9. What changes when… (one paragraph). The transition. What changes when one of these infrastructure firms not only spans multiple layers but explicitly contracts with one specific searcher or trading firm for exclusive flow at one of those layers? When the take-rate becomes a captive economic rent rather than a market-clearing price? That is Chapter 7.

  10. Footnotes and sources — numbered with URLs and access dates. Approximately 25–30 footnotes.

The worked example and where it threads through ​

Helius (Solana) and Flashbots (Ethereum) as paired anchor profiles.

SectionBeat
§3 (Setup)Plant: "two firms anchor this chapter — Helius on Solana, Flashbots on Ethereum — and the other firms get summarised against them."
§4 (Worked example)Full firm profiles: founding, funding, product stack, customer base, revenue model.
§5 RPC providersHelius as the chapter's RPC-layer template; Triton, Astralane, multi-chain RPCs against it.
§5 Block builders + bundle marketsJito Labs (Solana) full treatment; Titan / BuilderNet / Quasar (Ethereum); the Ben Coverston paragraph; the JIP-24 paragraph.
§5 RelaysbloXroute full treatment; Ultrasound / Aestus / Flashbots Relay against bloXroute.
§5 OFA / refund servicesFlashbots Protect + MEV-Share + SUAVE archival; Consensys/SMG / MEV-Blocker; Kolibrio; CoW DAO.
§6 (Meet the Infrastructure Provider sidebar)Instantiated against Helius.
§7 (Chain comparison)Helius's six-layer span vs Flashbots' specialisation vs Hyperliquid's no-third-party-layer.
§8 (Verdict)The bottom-up aggregate; the named winners and losers.

Diagrams needed ​

Two diagrams.

  1. D1 — The infrastructure stack (Mermaid flowchart). Vertical layers, each labelled with the major firms operating it per chain. From the user side downward: User wallet (Phantom / Backpack / MetaMask) → RPC provider (Helius / Triton / Astralane on Solana; Infura / Alchemy on Ethereum) → OFA / refund service (Kolibrio / Helius rebates on Solana; Flashbots Protect / MEV-Blocker on Ethereum) → Block builder / bundle market (Jito Block Engine + BAM on Solana; Titan / BuilderNet / Quasar on Ethereum) → Relay (none on Solana; Ultrasound / Titan / bloXroute / Aestus on Ethereum) → Validator / Slot leader (operates the chosen validator client + scheduler) → Chain. The diagram lands the chapter's structural claim visually.

  2. D2 — Firm-by-layer table (markdown). Rows: each named firm. Columns: chain(s), founder/CEO, founded, funding raised, layer(s) of operation, what they sell, who buys, current scale (share / customers / stake / etc.). The table is the chapter's most compressed single unit of information — the reader can use it as a reference card.

Glossary terms this chapter introduces ​

To be appended to GLOSSARY.md in Phase 3.

Defined in full (first appearance):

  • RPC provider — a firm that operates the API gateway between user-facing applications and the chain itself; in 2026, the largest RPC providers also operate validator stake, MEV-rebate flows, and (in some cases) private-mempool surfaces.
  • Order Flow Auction (OFA) — a market in which a routing entity (Kolibrio, Flashbots-via-MEV-Share, CoW solver auction, MEV-Blocker / SMG) sells the right to fill a transaction to whichever searcher or solver bids the most, with a share of the bid refunded to the original submitter.
  • MEV refund — a payment from a successful searcher or solver back to the original transaction submitter, typically routed through a private-routing service or OFA. Helius's 50/50 split between wallet and provider is the canonical case.
  • JIP-24 — the 2025 Jito DAO proposal that routes 100% of Jito Block Engine and BAM fees to the Jito DAO treasury, replacing the prior 3%/3% split between Jito Labs and the DAO.
  • Infrastructure firm (B2B trading infrastructure) — a company whose business is to operate one or more layers of the on-chain trading infrastructure stack (RPC, OFA, builder, relay, block engine) and capture a share of the trading flow that passes through it.

Cameo only:

  • LaserStream / Sender (Helius-specific products; one inline mention each)
  • SUAVE (Flashbots' archived chain; one inline mention with date)
  • BDN (bloXroute Blockchain Distribution Network)
  • a4.astralane.io (Astralane's indexing platform)

Backward:

  • Chapter 3 (drafted): introduced many of these firms in the context of the mempool's replacement. Ch 6 develops them as institutions with revenue models, leadership, and funding histories.
  • Chapter 4 (drafted): the searcher pays the infrastructure firms; this chapter develops who they pay and how much.
  • Chapter 5 (drafted): the validator + builder roles. Ch 6 develops the firms that maintain the clients (Anza, Jump Crypto, Rakurai) and the firms that operate the block-construction stack (Jito Labs, BAM, Titan, BuilderNet).

Forward:

  • Chapter 7 (Exclusive Order Flow, not yet drafted): the next step — when an infrastructure firm not only operates a layer but contracts exclusively with one searcher or trading firm at that layer. The Titan + Banana Gun arrangement is the seed; Chapter 7 develops the full pattern.
  • Chapter 9 (Hyperliquid): the chain that does not have a significant third-party infrastructure layer. The contrast lands the structural argument.
  • Chapter 10 (Ethereum and L2s): the L2 sequencer firms (Coinbase Cloud at Base, Offchain Labs at Arbitrum, the Optimism Foundation's chosen sequencer for OP Mainnet) sit at the intersection of "infrastructure firm" and "L2 operator"; Chapter 10 develops them in that frame.
  • Chapter 11 (Who's at a Disadvantage): the structural concentration the chapter documents informs the chronic-loser thesis.
  • Chapter 12 (Where This Is Going): protocol-level alternatives — ePBS, Constellation, decentralised sequencers — that would re-shape or replace the current infrastructure-firm landscape.

Ch 3 and Ch 5 back-edits scoped for Phase 3 ​

The research surfaced one structural update that affects prior chapters: JIP-24's passage in 2025 routes 100% of Jito Block Engine + BAM fees to the Jito DAO, replacing the prior 3%/3% split between Jito Labs and the DAO. The chapters that cite the older structure:

  • Ch 3 footnote 8 (BAM revenue paragraph): cites "no direct user fees; revenue comes via a 6% tip fee split between Jito Labs and the DAO; Q2 2025 DAO share ~22,391 SOL." This is correct for Q2 2025 but does not reflect the current post-JIP-24 structure. Recommend: add a sentence noting JIP-24's effect on the current state.
  • Ch 5 footnote 25 (BAM revenue mention in chain-comparison): same. Add the JIP-24 note.

These are minimal patches — the chapters' arguments don't change, but the current-state framing should reflect the JIP-24 transition. I'll make the back-edits as part of Ch 6 Phase 3 and note them in REVIEW_NOTES.md.

Phase 2 is complete. Per the user's compressed-review pattern, Phase 3 begins immediately.