Appearance
Chapter 4 — Outline (Phase 2)
Status: OUTLINE (Phase 2 — written by agent, awaiting Nick's review) Date: 2026-05-13 Working chapter: 04 — The SearcherCarries forward from: RESEARCH.md
Working assumptions (carried over from Phase 1)
These are defaults I'm proceeding with. Any of them is cheap to override before Phase 3 — overriding after Phase 3 is not.
- Worked example: Candidate A from the research note. Alice's $10,000 USDC→SOL swap on Jupiter, sandwiched for $73, framed as an early-2025 retrospective — used to develop the sandwich mechanic in its purest form, then closed by the April 2026 Solana protocol patch + Jupiter MEV-Protect routing, which becomes the bridge into Chapter 5.
- The $370M–$500M Solana sandwich aggregate: skipped. The chapter uses the clean Helius/Vpe point-in-time data (one program, $13.4M in 30 days) and a hedged "hundreds of millions cumulatively" only if needed for context.
- Searcher gas-spend ratio: drop the unsourced "80–90%" figure. Make the cost-structure point via Paradigm's theoretical ceiling (Priority is All You Need) and Flashbots' Base examples (350 failed attempts per win; $0.12 profit, $0.02 fees per success).
- JIT liquidity: handled in a ~50-word footnote in the sandwich section. Full treatment in Chapter 2.
- TradFi analogue: the Flash Boys-era HFT-arbitrage desk (Citadel Securities, Jane Street, Knight Capital ca. 2010). PFOF / Robinhood is reserved for Chapter 7.
- Figment Q1 2026 validator report: fine to cite. The chapter is about the searcher, and the Figment number is being used to size MEV's share of validator economics (a context fact, not a load-bearing claim).
- Hyperliquid in the chain-comparison box: framed as "this surface doesn't exist here, and here is where the rents went instead (HLP, validator set)." Neutral, mechanical — not an endorsement.
Title and subtitle
Chapter 4 — The SearcherWhy someone is always watching your trade — and what they get paid for it.
Cold open
One paragraph, sourced to the Helius Q1 2025 Solana MEV Report.
Between early December 2024 and the first week of January 2025, a single anonymous program on Solana — known to on-chain analysts only as Vpe — executed 1.55 million transactions and walked away with $13.4 million in profit. According to Jito's own analysis, this one program was responsible for roughly half of all sandwich attacks on Solana during that window. It paid $4.6 million in fees to the validators who included its transactions. Its success rate was 89%. No one knows who runs it. This chapter is about what people like the operator of Vpe do for a living, why the role exists, and out of whose pocket the $13.4 million came.
(Why this open: it gives the reader a named, sized antagonist on page one; it establishes that the chapter is about a real business, not a theoretical possibility; and it pre-commits the chapter to the "name the loser" promise — by the end the reader needs to know whose $13.4M that was.)
What this chapter answers
The four questions, phrased the way the chapter spec template requires.
- What is a searcher, and why does the role exist on every chain that lets transactions queue?
- What are the three things searchers do — arbitrage, liquidations, sandwiches — and how are they economically different from each other?
- Where does the searcher's profit come from in each case, and whose pocket does it leave?
- Why is searching a real business with real costs, and why has it consolidated to a handful of firms?
Section list (one sentence each)
The chapter has eight required sections plus a footnotes block. Final structure:
- Cold open — the Vpe scene above.
- What this chapter answers — the four questions, as above.
- The setup (≈500 words). Defines MEV in plain language without using the acronym's expansion as the definition; introduces the searcher actor (the chapter's first "Meet the actor" sidebar); establishes — in two sentences only — the mempool/transaction-visibility prerequisite that Chapter 3 will develop. The TradFi analogue is planted here: what the HFT-arbitrage desk was to NASDAQ in 2010, the searcher is to on-chain markets in 2024.
- The worked example (≈350 words). Alice clicks "swap" on Jupiter in February 2025. Her $10,000 USDC→SOL transaction sits, briefly visible, before inclusion. The example is set up here but its resolution — what happens in the 400 milliseconds between click and confirmation — is held back until section 5c so the reader experiences the sandwich the way Alice did: through its consequences, not its mechanics.
- The mechanics, in detail (≈2,500 words, four H3 subsections — at the chapter spec template's hard limit):
- 5a. Arbitrage: closing a gap (≈600 words). The morally neutral case. The 2Fast $1.9M one-shot is the cameo. The CEX-DEX case is the structurally darker variant — the 19-month, $233.8M extraction across 7.2M trades is named. Who pays: in pure on-chain arb, whichever venue was slow to reprice; in CEX-DEX arb, passive LPs and slow market makers.
- 5b. Liquidations: forced repayment (≈500 words). The useful case. The March 10, 2026 Aave wstETH event ($27M liquidated within hours after a CAPO oracle misprice) is the named scene. Mechanics: liquidation bonus (5–15% on Aave V3, 5% on MarginFi), close factor, the searcher's race. Who pays: the borrower, who would have lost the position anyway under any healthy lending protocol — this is the case where "extraction" and "necessary maintenance" are the same act.
- 5c. Sandwiches: the pure-extraction case (≈900 words). Alice's transaction resumes here. The four-panel diagram is introduced. The mechanics are walked through in dollars: frontrun, fill at degraded price, backrun. The $73 is broken down (searcher net, Jito tip / priority fee, base fee, gas). The JIT liquidity footnote sits here. The architectural response — Jupiter MEV-Protect routing as of late 2024, Solana's April 8, 2026 protocol-level mitigation via Jito BAM private routing — is the section's final beat, and the chapter's bridge.
- 5d. Why this is a real business (≈500 words). The cost structure: gas on failed bids (Flashbots' 350:1 ratio on Base), the theoretical ceiling on net margin (Paradigm's "priority is all you need" — the winning searcher pays roughly the full extractable value as a priority fee, keeping only the wedge between their bid and the runner-up's), the consolidation (23→11 active Ethereum CEX-DEX searchers in Q1 2025; top three capture 90%). The point lands as: a searcher's competitive moat in 2026 is increasingly its relationship with a block builder, not its algorithm.
- How this plays out on each chain (≈350 words, three paragraphs). The required chain-comparison box.
- Solana: pre-patch, the most concentrated sandwich market — one program (Vpe) accounted for ~half. As of April 2026, simple-form sandwiches are mostly suppressed via private routing and Jito BAM; the live extraction surface is now CEX-DEX latency arb and MEV against long-tail memecoin pairs. Jito tips remain the dominant ordering signal.
- Hyperliquid: there is no equivalent public surface. The on-chain CLOB matches inside HyperBFT with ~0.07s finality; no mempool, no public next-block ordering view. The economic surplus that would accrue to external searchers on Ethereum-style architecture instead accrues to the HLP vault (which absorbs liquidation surplus) and to the validator set. MEV on Hyperliquid is not measured because the surface for permissionless extraction doesn't exist; the rents are simply collected closer to the protocol.
- Ethereum and L2s: extraction has fallen sharply on L1 — sandwich revenue dropped from $10M/month in late 2024 to $2.5M/month by Oct 2025 as order flow migrated to solver-mediated venues (CoW, UniswapX) and protected routing (MEV-Share, Flashbots Protect). On L2s (Base, Optimism, Linea), the surface is different — Flashbots measured >50% of Base gas consumed by spam bots paying <10% of fees, with two firms accounting for >80% of the spam.
- Who wins, who loses, why (≈400 words). The verdict. Structure as required by the chapter spec template:
- Winners: searcher firms with builder relationships ($233.8M extracted by 19 firms over 19 months, top three taking 90%); validators and Jito (per the Vpe sample, ~35% of sandwich gross revenue redirected to validators via tips); Jupiter (which monetizes routing flow at the wallet-app layer).
- Losers: the user being sandwiched (clean retrospective number — $73 of a $10K trade in unprotected early-2025 conditions); passive LPs on AMMs and CEX-DEX-arbed pools, who pay the silent tax that doesn't show up on a per-trade slippage line; slow market makers, who get picked off; and validators without exclusive infrastructure relationships, who increasingly do not see the order flow that matters.
- Is this bad?: clinical answer. Sandwiches are pure transfer with no useful function; the architectural response is real and is visible in the falling extraction numbers. Arbitrage is price discovery and the chain pays for it through validator economics. Liquidations are price discovery the protocol needed. The honest summary line for the reader: the dollar of slippage is not lost to corruption; it is lost to a market structure that has, until recently, paid more attention to throughput than to who pays for it.
- What changes when… (one paragraph). The chapter's transition into Chapter 5: what changes when the validator who built the block has an exclusive arrangement with one specific searcher? That is Chapter 5. (Plus a half-sentence reminding the reader that Vpe paid $4.6M to validators over 30 days — and we still don't know who they were.)
- Footnotes and sources — numbered, URL'd, with access dates per the agent prompt.
The worked example and where it threads through
Alice's $10,000 USDC→SOL swap on Jupiter, February 2025 (chosen to predate Jupiter's MEV-Protect default-on and Solana's April 2026 patch).
| Section | Beat |
|---|---|
| §3 (Setup) | Plant: "imagine a swap on Jupiter — we'll come back to it." |
| §4 (Worked example) | Setup only: Alice clicks swap; transaction is briefly visible in the public path; the reader sees the click and the result, but not what happened in between. |
| §5c (Sandwiches) | Payoff: the four-panel mechanic in dollars. The $73 is broken down. |
| §5d (Real business) | Pivot: a fraction of Alice's $73 became searcher net profit; most of it went to gas, priority fees, and the next-bid runner-up. |
| §6 (Chain comparison) | Counterfactual: the same Alice trade in May 2026, routed through Jupiter MEV-Protect, would lose closer to zero. |
| §7 (Verdict) | The dollar split is the centerpiece. Alice is the named loser; the named winners are the searcher, the validator, and Jito. |
Diagrams needed
Three. (The chapter spec template allows 2–5; the SPEC suggests 2–3 with the P&L table optional.)
D1 — The sandwich (sequence diagram). A four-panel Mermaid sequence diagram showing: (a) Alice's pending swap visible in the public transaction path (b) the searcher's frontrun buy executing first (c) Alice's swap executing at the now-worse pool price (d) the searcher's backrun sell capturing the difference Each arrow annotated with the dollar flow. Lives at the top of §5c.
D2 — Where the $73 went (horizontal stacked bar). Static SVG (or a clean Mermaid pie if SVG tooling is friction). Shows Alice's $73 broken down into: searcher net profit, validator/Jito tip, base fee + gas, runner-up bid spread. Lives at the bottom of §5c and is referenced again in §7.
D3 — Searcher P&L for a representative day (table). Markdown table, not a diagram strictly. Columns: opportunities seen / bids attempted / bids won / gross profit / gas paid on losers / gas paid on winners / Jito tips paid / net profit. Numbers derived from Flashbots' Base data + Vpe's monthly snapshot. Lives in §5d.
(Optional fourth, deferred unless §5 ends up shorter than expected: a 3-column comparison matrix of arbitrage / liquidations / sandwiches across "who pays / extractive? / victim aware? / scale.")
Glossary terms this chapter introduces
To be appended to book/glossary/GLOSSARY.md in Phase 3, in the template format. All written for the business reader; no jargon-defining-jargon.
Defined in full (this chapter is their first appearance):
- MEV — the profit a privileged participant can earn by reordering, inserting around, or censoring transactions inside a block. Always paid for, ultimately, by someone whose execution price got worse or whose opportunity was taken.
- Searcher — a firm or operator who runs latency-optimized software to find and capture MEV opportunities. The chapter's anchor actor.
- Sandwich attack — placing a buy immediately before a victim's swap and a sell immediately after, so the victim fills at a price the attacker just moved.
- Frontrun — placing a transaction ahead of a target transaction whose contents you have already seen.
- Backrun — placing a transaction immediately after a target transaction whose effect you have predicted.
- Arbitrage (on-chain) — closing a price gap between two on-chain venues, or between an on-chain venue and an off-chain one, by trading both sides in a single transaction or bundle.
- CEX-DEX arbitrage — the variant where the price gap is between a centralized exchange (Binance, Coinbase) and a DEX pool. The "darkest corner" because the inputs are not on-chain.
- Liquidation — forcing repayment of an under-collateralized loan by selling the borrower's collateral on behalf of the protocol, in exchange for a protocol-set bonus.
- Liquidation bonus — the premium (typically 5–15%) the searcher receives on the value of liquidated collateral, paid by the borrower's seized position.
- Priority fee — the per-transaction fee a user (or searcher) pays above the base network fee to influence ordering or inclusion. The dominant ordering signal on Solana and Ethereum.
- Jito tip — on Solana, a separate payment a transaction can include to be routed through Jito's block engine. Different from the priority fee in mechanism, similar in purpose.
Cameo only (defined inline in two sentences, full treatment in another chapter):
- Mempool — full treatment in Chapter 3.
- Bundle (Jito-style) — a group of transactions that must execute together, in order, or not at all. Full treatment in Chapter 6.
- Block builder — the actor who assembles the set of transactions in a block and submits it for inclusion. Full treatment in Chapter 5.
- Health factor — the on-chain ratio used by Aave-style lending protocols to determine when a position becomes liquidatable. One sentence inline; not a full glossary entry.
Mentioned but not glossary-worthy (single use, sufficient inline context):
- HyperBFT, HLP, CAPO oracle, close factor, JIT liquidity (footnote-only — full treatment in Chapter 2).
Forward and backward links
Backward (what this chapter builds on):
- Prologue (not yet drafted). Establishes Alice's $10,000 trade and the $73 question. If the prologue isn't ready when the draft lands, the worked example in §4 carries the setup inline with no loss.
- Chapter 2 — Where Liquidity Lives (not yet drafted). LPs as the silent counterparty to AMM-routed trades. Establishes who the "passive LP" loser in §7 is. Until Chapter 2 exists, §5a and §7 carry a 1–2 sentence inline definition of "passive LP."
- Chapter 3 — The Mempool and What Replaces It (not yet drafted). Transaction visibility is the prerequisite for sandwiches. Until Chapter 3 exists, §3 carries a single-sentence definition of the public transaction path.
Forward (what this chapter sets up):
- Chapter 5 — The Validator (and the Builder). The closing transition asks where Vpe's $4.6M in tips went. Chapter 5 names the recipients and explains the proposer/builder dynamics that make the relationship valuable.
- Chapter 6 — The Infrastructure Layer. Jito appears in cameo in this chapter; the firm gets full treatment in Chapter 6. So does the bundle mechanic.
- Chapter 7 — Exclusive Order Flow. §5d's "the moat is the relationship, not the algorithm" point is the seed for Chapter 7.
- Chapter 11 — Who's at a Disadvantage. §7's named losers (passive LPs, slow MMs, unsuspecting traders) are revisited as the four chronic losers in Chapter 11.
Self-check (preview of the Phase 3 REVIEW_NOTES.md questions)
I'll formally answer these in Phase 3. Logging them here so the structure of the outline can be tested against them now.
- Could a smart business reader with zero crypto background follow the outline? — Cold open is concrete and named; setup defines terms; worked example is dollar-anchored. Yes.
- Is every actor named, and is each one's income source clear? — Searcher (full sidebar), validator/builder (cameo, money source named: priority fees + tips), Jupiter (revenue source named: routing). Yes.
- Is the worked example threaded through? — Yes, six explicit beats above.
- Does the chapter end with a clear "who wins, who loses" verdict? — Yes, §7 is the required half-page.
- Will all numbers be sourced in footnotes? — Yes; all are in RESEARCH.md with URLs.
- Does the chain comparison box exist and contain real differences? — Yes; three distinct architectures, three distinct extraction surfaces.
- Banned moves avoided? — No "revolutionary," "the future of," "rugged," "Web 1.0," tribal chain endorsements. Will double-check at draft time.
- Would the Goldman MD finish it? — The Vpe cold open, the named March 2026 Aave event, the broken-down $73 — three reasons she keeps reading. Will read aloud at draft time to confirm.
Phase 2 is complete. Awaiting Nick's review before producing the Phase 3 draft, the Mermaid diagrams, and the glossary appends.