The House Edge

Appearance

3. The Mempool and What Replaces It ​

A surface, three responses, and the actor whose job is to read it. ​

Cold open ​

On Friday, 8 March 2024, Jito Labs — the firm whose software runs on most of Solana's stake — shut down its private mempool within hours of an internal decision, with no advance notice to the searchers who had built businesses on it. The Block Engine that exposed the mempool to all searchers had been live for roughly ten months; Jito-Solana validators had been running it in more limited mainnet and testnet operation for about six months before that. At the run rate of the days leading up to the shutdown, MEV tips on Solana were generating approximately ten thousand SOL per day — over half a billion dollars annualised at the prevailing price.[1] Jito's stated reason was to "prioritize the network's long-term growth and stability, sacrificing significant revenue in the process."[2] The reason that did not need to be stated was that sandwich attacks during the memecoin frenzy had made the mempool's negative externalities untenable for the firm to host. This chapter is about the surface Jito turned off, the surfaces other chains have built or refused to build, and the actor whose job is to read whichever surface is available.

What this chapter answers ​

  • What is the public mempool, and why does its existence make sandwich attacks and most other extractive techniques possible?
  • What did Jito's mempool do, and why did Jito shut it down — and what replaced it on Solana?
  • How do Ethereum and its L2s handle the same surface, and why has private order flow become the value-weighted majority of Ethereum traffic?
  • What is the trade-off between MEV-resistance and censorship-resistance, and what is being built about it?

The setup ​

A mempool is the queue of pending transactions a node holds before they are included in a block.

On Ethereum, every node maintains its own local mempool and gossips its contents to peers across the public peer-to-peer network. Anyone running a node — or paying for an RPC provider that exposes pending transactions — can read the queue. The queue is the public mempool.

On Solana, there is no shared global mempool: Gulf Stream, the protocol's pre-inclusion transmission layer, forwards transactions directly to the current and next slot leaders' transaction-processing units. The leader sees the transactions; everyone else sees only what makes it into a block.

On Hyperliquid, the chain matches transactions inside consensus itself; there is no pending state to observe.

These three architectural choices set the floor of what is possible to do with transaction visibility. The ceiling is set by what each chain's operators have built on top of the floor — and that is where the chapter spends most of its pages.

The worked example ​

Alice opens her wallet on a Wednesday morning in February 2026. She has $10,000 of USDC and she wants SOL. Her wallet is configured, by default, to broadcast her transactions through a stock Solana RPC — the kind any wallet ships with, the kind anyone can read. Jupiter, the aggregator she uses to route the swap, gives her a choice on the way to confirmation: she can submit through the default path, or she can use Ultra, which routes her transaction through Jupiter Beam — Jupiter's private transaction-landing engine.

Her transaction reaches Jito's block engine via Beam as a private bundle. It is included in a block within roughly six hundred milliseconds. The pool slippage she paid was, on aggregate, somewhere between positive a few hundredths of a basis point and zero — Jupiter's published Ultra V3 numbers report an average of approximately +0.006% on protected paths, versus −0.14% on unprotected ones.[3]

What just happened, at the visibility layer, is the subject of the rest of this chapter. A couple of years earlier, an Alice making the same trade would have had a meaningfully different experience — and the difference traces, in a straight line, back to a Friday afternoon in March 2024.

The mechanics, in detail ​

Solana: the mempool that wasn't, the mempool that was, and what replaced both ​

Solana technically never ran a public mempool in the Ethereum sense. The chain's design, going back to 2020, used a transmission layer called Gulf Stream that forwarded transactions directly to the slot leader rather than gossiping them across the network. The architecture was sometimes promoted as a structural defence against the kinds of extraction Ethereum was experiencing.

In practice, the dominant validator client — Jito-Solana, run by Jito Labs — built a private mempool on top of the network's base layer. Jito-Solana validators began running the software on mainnet and testnet in late 2022, processing over a million bundles in the months that followed. On 25 May 2023, Jito Labs opened the Block Engine — and with it the mempool — to all Solana MEV searchers. From that point until the shutdown roughly ten months later, the Jito mempool gave searchers an approximately two-hundred-millisecond window to preview pending transactions before the slot leader included them.[4]

What grew on top of that surface was a sandwich-attack industry. A sandwich attack is the on-chain version of seeing a customer's market order before it trades, buying just ahead of it, letting the customer's order push the price up, and selling just after it. The trader gets a worse execution price; the attacker keeps the difference.

The largest known Solana operator was Vpe, an anonymous program that ran approximately 1.55 million sandwich transactions over the thirty days from 7 December 2024 to 5 January 2025, with an 88.9% success rate and approximately $13.4 million in gross profit. Jito's own analysis attributed roughly half of all Solana sandwich activity in that window to Vpe.[5] Vpe's existence pre-dated the shutdown in the cold open. What it needed was visibility into pending swaps, and the Jito mempool provided exactly that.

On Friday, 8 March 2024, Jito Labs shut down the public Jito mempool within hours. The announcement was abrupt; the firm cited network long-term stability and revenue sacrifice in roughly equal measure.[2:1] The day before the shutdown, daily Solana MEV tips had exceeded ten thousand SOL.[1:1] In removing the surface, Jito gave up a substantial fraction of its own revenue stream — the company kept the bundle infrastructure that delivered atomic transactions to validators, but turned off the pre-inclusion preview that searchers had used to construct sandwich attacks.

The shutdown did not end the sandwich industry; it sent it underground. Within months, independent reporting documented at least three Solana validator operators receiving solicitations from searcher firms to run private mempools — closed, paid, contract-based services that would re-create the visibility surface Jito had eliminated, for the searchers willing to pay.[6] The Solana Foundation responded in June 2024 by removing approximately thirty-two validator operators, representing roughly 1.5 million SOL or 0.5% of the foundation's delegation-program stake, for participating in such arrangements. The foundation's validator-relations lead, Tim Garcia, framed the action as ongoing rather than a single enforcement event: "Enforcement actions are ongoing as we detect operators participating in mempools which allow sandwich attacks."[7] The book has not, until this point, described an enforcement action of this clarity by any chain-affiliated body. It is the cleanest example the book has of a foundation acting against participants whose business depended on a specific extraction surface.

The enforcement did not end the demand for visibility. It fragmented the supply. Vpe — the program the chapter has been citing as the canonical sandwich operator — has itself been linked, in independent on-chain analysis, to a specific Solana validator known as DeezNode (validator address HM5H6…jdMRA). DeezNode's stake nearly tripled across late 2024, from approximately 308 thousand SOL in November to over 800 thousand SOL by mid-December, on the back of a published offer to opting-in validators of a fifty-percent revenue share from the private-mempool flow it routed.[8] The Vpe program is, in this telling, less an anonymous bot than the customer-facing arm of a validator-side business that grew up directly inside the surface Jito's shutdown had eliminated.

The chapter's structural point lands in the surrounding ecosystem rather than in DeezNode specifically. The visibility surface re-emerged, in the eighteen months after the Jito shutdown, at three layers underneath the formal architectural responses Jito itself was building.

RPC providers sell pre-leader transaction visibility to paying customers — Kolibrio, a VC-backed Order Flow Auction startup founded in 2022 (with a $2 million Jump-Crypto-led seed in February 2023), runs precisely this product: integrators route user transactions through Kolibrio's "Meow RPC," and Kolibrio auctions visibility to searchers, rebating a share back.[9]

Wallet infrastructure carries the same dynamic at the user interface: Helius — the network's largest single validator at fifteen million SOL of stake, and the RPC pipeline behind Phantom, Backpack, Solflare, and several smaller wallets — markets the resulting MEV-rebate flow back to its wallet customers as part of the service ("SOL rebates from trades that create arbitrages").[10]

Validators themselves position non-voting infrastructure next to the leader's Transaction Processing Unit, or TPU: the validator-side intake pipe that receives transactions for the slot the validator is about to produce. Because Solana has no public mempool, pending transactions do not sit in one shared public waiting room. They concentrate, briefly, around the current and next leaders' intake paths. For the slot in which a validator is leader, that intake becomes the mempool-like surface. If a paid RPC or non-voting validator is placed next to it, it can observe transactions and forward them to specific searcher partners before the rest of the market sees the resulting block. [11]

On 8 April 2026, Solana shipped a protocol-level change that further compressed the surface for simple validator-leader sandwich attacks. The exact mechanism is not publicly documented in any source the chapter could verify — the change is widely reported, but the specific consensus-layer mechanism, the validator-client release tag, and the precise scope of the fix have not surfaced.[12]

The simple sandwich attack of 2023 and 2024 is, in 2026, largely a historical phenomenon on Solana; the residual extraction has shifted to long-tail tokens, unprotected wallets, and CEX–DEX latency arbitrage across a wider range of blocks, see https://sandwiched.me/wide_sandwiches.html.

A protocol-level redesign is in the works — Constellation, a Multiple Concurrent Proposers architecture with 16 proposers, 256 attesters, and a leader that orders strictly by priority fee per compute unit — which is positioned to replace the off-protocol scheduler stack entirely. The argument is that inclusion guarantees should live inside the protocol rather than in private contracts between searchers and validators. Constellation is a proposal, not a commitment; it follows Alpenglow on Anza's roadmap and is unlikely to ship before late 2026 at the earliest.[13]

Taking control over Solana mempool ​

Jupiter — Solana's dominant aggregator — built Beam, an in-house transaction-landing engine that delivers signed transactions directly to Jito's block engine as private bundles, bypassing the public RPC paths entirely. Beam launched as part of Jupiter Ultra, which became the aggregator's default routing mode in 2026 and which now handles approximately 95% of Solana's aggregator-routed flow and roughly half of Solana's total DEX volume.[3:1]

In April 2026 Jupiter opened Beam as a Public Transaction Submission API — any signed Solana transaction can enter the protected path, not just Jupiter-routed swaps. The published Ultra V3 numbers — +0.006% average slippage versus −0.14% unprotected.

At the block-construction layer, Jito Labs launched BAM — the Block Assembly Marketplace — on 21 July 2025. 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. The privacy guarantee depends on hardware attestation rather than on cryptographic primitives or time-locks. BAM splits a function that had previously been monolithic:

  • The Sequencing — the decision of which transactions go into which slot, and in what order — is delegated to BAM Nodes.
  • The Block Execution — the actual production of the block — remains with the validator.

BAM has no direct user fees; its revenue model is a 6% tip split between Jito Labs and the Jito DAO, with the Q2 2025 DAO share alone approximately 22,391 SOL (~$4 million at then-prevailing prices).[14] BAM also supports Application-Controlled Execution plugins: in April 2026, Jump Crypto launched a "BAM Maker Priority" plugin giving registered market makers deterministic top-of-batch execution every fifty milliseconds. Other plugins handle oracle prioritisation, liquidation-before-market sequencing, and LP-priority access inside AMM batches.[15]

Harmonic is an open block-building marketplace, launched November 2025 with a $6 million Paradigm-led seed. A validator running Harmonic continuously receives order flow from Harmonic builders — that are supposed to capture order flow from Jito's Block Engine, Temporal's Nozomi, Harmonic Bundles — 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 across multiple builders, not a single-firm decision, but so far it seems that the only company running Harmonic block builders is Harmonic.

The preferences architecture also creates a hook for partner-specific arrangements that Harmonic itself does not publicly advertise but that the marketplace structurally supports. As of March 2026, Harmonic ran approximately 17% of Solana network stake; Nasdaq-listed DeFi Development Corp adopted Harmonic in December 2025 to "optimize Solana validator revenue."[16]

Validators running Harmonic variants captured meaningfully more priority-fee revenue than the median: Frankendancer Harmonic Performance earned +101% priority fees per block versus the network median, Frankendancer Harmonic Balanced +39%, Agave Harmonic +36% which is explained by Harmonic bundles requiring PF for faster inclusion rather than relying on Jito MEV contracts.[17]

Ethereum: the public surface and its private replacement ​

Ethereum's public mempool exists in a way Solana's never did. Every Ethereum node — every one of the roughly seven thousand publicly-discoverable mainnet nodes, plus the larger universe of private nodes operated by RPC providers, validators, and searchers — maintains its own local pool of pending transactions and gossips its contents across the peer-to-peer network. Anyone running a node sees every transaction that has been broadcast publicly but not yet included in a block. The information is not just available to chain insiders; it is available to anyone who downloads the software.

This is the surface the searcher industry was built on. It is also, in 2026, the value-weighted minority of the network's transaction flow.

Carlos — an institutional Ethereum trader, a new character for this chapter — is making a $10,000 USDC→ETH swap. He is technically sophisticated and aware of the visibility issue. Before submitting the swap, he changes the RPC URL in his wallet's settings to the Flashbots Protect endpoint. From the wallet's perspective, nothing changes: it constructs a signed transaction and sends it to whatever RPC URL is configured. From the network's perspective, everything changes: the transaction does not enter the public mempool. Instead, it goes directly to Flashbots' relay infrastructure, which forwards it as a private bundle to block builders, who include it in blocks they submit to validators. If Carlos's transaction creates a backrunning opportunity — say, a price gap a searcher could close immediately after his fill — the searcher's bid for that opportunity is split, with a share refunded back to Carlos through the Protect mechanism.[18]

Flashbots Protect has served approximately 2.1 million Ethereum accounts and shielded $43 billion or more in cumulative DEX volume since launch; third-party benchmarks place its share at roughly 7% of Ethereum's monthly transaction count, or approximately 3 million transactions per month.[18:1] MEV-Blocker, the alternative service originally operated by CoW DAO and transferred to Consensys' Special Mechanisms Group in January 2026, has served over 4.5 million unique wallets and paid 6,177 ETH in cumulative rebates back to users since launch.[19] MEV-Share — Flashbots' selective-disclosure mechanism, which lets users opt into letting searchers see specific parts of a transaction in exchange for a share of the extracted value — is the smaller third path; its adoption has been steady but never matched the simpler Flashbots Protect URL-swap path.

The structural observation, from a peer-reviewed measurement by Wahrstätter and co-authors updated through 2025, is that private orders consume more than half of all gas used on Ethereum, account for approximately 30% of transactions by count, and represent 54.59% of total block value.[20] The public mempool is increasingly empty of valuable trades, even when it appears full by transaction count.

The searcher who reads the public mempool in 2026 is reading the leftovers.

Above the private-routing layer sits the MEV-Boost stack — the Ethereum proposer-builder-relay separation that has been the network's standard infrastructure since the Merge. A validator (the proposer) outsources block construction to a specialised builder via a relay. The builder constructs the most profitable block it can find, including bundles from searchers and private routing services like Flashbots Protect. The relay forwards the block to the proposer. The proposer signs it.

The structural concentration in 2026 is sharp. Live snapshots from relayscan.io for the 24 hours ending 14 May 2026 show one builder, Titan, producing 50.60% of all Ethereum blocks, with BuilderNet at 24.55% and Quasar at 15.68% — the top three builders together accounting for approximately 91% of mainnet blocks.[21] On the relay side, Ultrasound (35.71%), Titan Relay (26.09%), BloXroute Max-Profit (13.38%), and BloXroute Regulated (12.30%) handle the great majority of payloads.[21:1] Flashbots itself, which once operated the dominant centralised builder, stopped doing so in December 2024 and migrated its building infrastructure into BuilderNet, a decentralised multi-operator builder protocol. The migration was both an acknowledgment of and a response to the firm's concern that builder centralisation had become the network's central structural risk.[22]

The downstream observation that lands the chapter's argument: when 91% of Ethereum blocks are built by three firms, the question of what is visible in the public mempool matters less than the question of which firms have the relationships to put a transaction in front of those builders. Carlos's choice to route through Flashbots Protect is, from the public-mempool perspective, an act of privacy; from the builder-concentration perspective, it is a vote for one set of plumbers over another. Chapter 7 develops the consequences.

L2s: the sequencer as a one-firm private mempool ​

The L2 surface is structurally different from both Solana and Ethereum L1. The major L2s — Base, Arbitrum, Optimism — each run a single sequencer, an off-chain ordering service operated by one firm. The sequencer receives transactions from users (or from RPC providers acting on users' behalf), orders them according to its policy, batches the ordered transactions, and posts the batches to L1 for canonical settlement. The sequencer sees every transaction before ordering it. From the user's perspective, an L2 sequencer is a one-firm private mempool by construction.

Base — the largest of the three by transaction count — runs a sequencer operated solely by Coinbase. The sequencer respects OFAC SDN-list filtering, which has been the subject of recurring debate inside the Ethereum community about whether L2 sequencers should be censorship-resistant by default. Base's sequencer revenue for the first half of 2025 was approximately $42.4 million, accounting for 87.2% of total Optimism Superchain sequencer revenue; daily priority fees alone averaged roughly $156,138, or about 86% of daily sequencer revenue.[23] Coinbase has stated publicly that it does not directly run sandwich-attack strategies through the Base sequencer.

Arbitrum — the second-largest L2 by activity — runs a sequencer operated by Offchain Labs. In April 2025, Arbitrum activated Timeboost, which replaced the prior first-come-first-served sequencing with an off-chain sealed-bid auction. Searchers bid for a 200-millisecond "express lane" advantage in each auction round; the winner can submit transactions that are guaranteed to be ordered ahead of any submitted through the default path during the express-lane window. Proceeds from the auction flow back to the Arbitrum DAO.[24]

The Timeboost design is the cleanest example of an L2 monetising its sequencing privilege at the protocol level rather than at the operator level — the value of priority ordering becomes a protocol revenue stream. In February 2026, an academic paper documented that searchers had begun routing around Timeboost's primary auction through a third-party resale market called Kairos, prompting Arbitrum to raise and then lower the auction's reserve price.[24:1]

Optimism's Unichain — a member of the Optimism Superchain — runs Flashblocks, a Flashbots-built sub-block protocol producing 200-millisecond block intermediaries that return MEV to users via revenue-share.[25] Unichain is the most aggressive MEV-back-to-user design currently live among major L2s. The OP Mainnet sequencer contributes 100% of its profit back to the Optimism Collective; other OP Chains contribute either 2.5% of revenue or 15% of profit, whichever is greater.

The decentralised-sequencing alternative exists but has not yet been adopted by any major L2. Espresso Systems went generally available on 12 February 2026, finalising blocks in approximately six seconds and offering a shared sequencing layer that any rollup can plug into.[26] Astria's mainnet alpha launched in October 2024 on Celestia data availability. As of May 2026, no major L2 — Base, Arbitrum, or Optimism — had migrated to a decentralised sequencer. All three still run a single operator-controlled sequencer that has full pre-execution visibility into the chain's pending transactions. The visibility surface on Ethereum L2s, in 2026, is a one-firm private mempool. Whether that is better or worse than Ethereum L1's public mempool depends on whose risk the reader is weighing.

How this plays out on each chain — and Hyperliquid as the alternative ​

On Solana, the surface has been compressed in three steps over two years. The Jito mempool shut down in March 2024. Beam at the router layer and BAM at the block-construction layer absorbed most of what the mempool had done. The protocol-level patch in April 2026 closed the residual public surface for simple sandwich attacks. Sandwich activity persists on long-tail tokens and unprotected wallets; the major surface is now privately routed. Roughly half of Solana's stake runs a Jito-derived client; a quarter runs JitoBAM specifically.

On Hyperliquid, there is no public surface at all. The chain's HyperBFT consensus matches transactions and order-book actions atomically inside consensus itself, with semantic ordering inside each block (non-order actions process first, then cancels, then GTC/IOC sends).[27] The mempool problem the chapter has spent most of its pages on does not exist on Hyperliquid; the structural choice eliminates the surface rather than managing it. The trade-off is that Hyperliquid's architecture is far less flexible than Solana's or Ethereum's — it works extremely well for what the chain is designed to do (a perpetuals exchange with a CLOB), and less well for everything else.

On Ethereum and its L2s, the picture splits. The L1 public mempool still exists but holds the value-leftovers; 54.6% of block value flows through private channels. The L2 sequencers are one-firm private mempools by construction — Coinbase for Base, Offchain Labs for Arbitrum, the Optimism Foundation's chosen sequencer for OP Mainnet. FOCIL (EIP-7805) is the protocol-level answer being built for the censorship-resistance side of the trade-off; Flashnet is the submission-layer answer for the privacy side. Neither is live on mainnet as of mid-2026. FOCIL has been selected as the consensus-layer headliner for Hegotá, the post-Glamsterdam Ethereum upgrade, which is now framed as a late-2026 cleanup-and-hardening fork.[28]

ChainWhat the visibility surface looks like in 2026Who operates itStructural note
Ethereum L1Public mempool exists; 30% of txns by count, 54.6% of block value privateNo single firm; ~91% of blocks built by Titan + BuilderNet + QuasarSurface still public; value increasingly private
BaseOne-firm sequencer; no public pre-execution viewCoinbaseOFAC-filtered; permissionless backrunning via priority fees
ArbitrumOne-firm sequencer + Timeboost express-lane auctionOffchain LabsAuction proceeds to Arbitrum DAO
Optimism UnichainOne-firm sequencer + Flashbots FlashblocksOptimism Foundation / FlashbotsMEV returned to users via revenue-share
SolanaGulf Stream (Leader Validator) + RPCs + Jito Block Engine / BAM (TEE-encrypted) + Harmonic BuildersValidators, RPC providers, Jito, HarmonicJito Public mempool eliminated 2024; private routing dominant
HyperliquidNo public surface at allHyperliquid validatorsArchitectural elimination, not management

Who wins, who loses, why ​

Winners. Jito Labs operates the dominant formal private-routing infrastructure on Solana — the Block Engine, BAM, the bundle market — and takes a 6% tip on BAM-mediated value. Jupiter has captured the wallet-level routing position; Beam is the default for the chain's largest aggregator. Helius captured the wallet-RPC pipeline; its position as a validator and the RPC business used by the popular wallets lets it monetise both sides of every retail trade. Kolibrio captured the same role at the integrator-RPC layer. Harmonic is taking over a sizeable chunk of Solana blockspace.

On Ethereum, three firms — Titan, BuilderNet, Quasar — build approximately 91% of mainnet blocks; their relationships with the relays and the proposer set are the structural moat. On L2s, Coinbase has built a half-billion-dollar-a-year sequencer business on Base; Offchain Labs runs both the Arbitrum sequencer and the Timeboost auction whose proceeds flow back to the Arbitrum DAO; the Optimism Foundation has positioned Unichain as the MEV-back-to-user alternative. None of these firms is a bad actor in any individual transaction; collectively, they are the new oligopoly the chapter's predecessor — the public mempool — has been replaced by.

Losers. Retail traders who continue to submit transactions through default public RPCs without protection still pay a slippage tax — smaller than in 2023 and 2024, but not zero, especially for long-tail tokens. The searcher firms whose business models depended specifically on the public-mempool surface that Jito shut down have largely consolidated, gone underground, or pivoted to the validator-side surface that DeezNode-style operators rebuilt; the approximately 32 validators the Solana Foundation removed in June 2024 are the cleanest named casualties of the architectural transition, though they represent the visible fraction of a larger underground that the formal enforcement was never going to fully eliminate. And the public-goods narrative of "anyone can run a node and see the network" has been quietly traded for the institutional reality of "submit a transaction visible to one firm — your wallet's RPC provider, or your aggregator's routing service, or your validator's preferred searcher partner — who chooses who else to share it with."

Is this bad? The dollar of slippage the chapter's worked example tracked has materially shrunk. In 2024, an unprotected version of Alice's trade could lose meaningful dollars to sandwiching; on the protected path in this chapter, the loss is somewhere in the cents-to-pennies range.

But the visibility-as-product market did not go away — it fragmented into the informal surfaces the previous paragraphs developed. The shutdown of the Jito mempool eliminated one centrally-managed surface and produced, in its place, a dozen smaller surfaces operated by RPC providers, wallet backends, validator partnerships, and Order Flow Auction startups.

The extraction is harder to count because no single dashboard tracks it; the dollar amounts per transaction are smaller because Jupiter Beam, Solana April 8 patch genuinely did compress the simple-sandwich path; but the market for visibility — the willingness of searchers to pay for pre-leader transaction sight, and of intermediaries to sell it — is alive in a way Jito's 2024 statement implied it would not be.

Whether that trade is good depends on which risk the reader weighs more heavily: the chronic transfer of value from retail traders to extractive bots, which the new architecture has measurably reduced; or the chronic transfer of trust from a public network to a small number of private firms operating fragmentary visibility surfaces, which the new architecture has measurably increased.

What changes when… ​

What changes when one of the firms that handles the private routing also runs the validator that includes the block — when the relationship between who submits the bundle and who chooses what goes in the block becomes a single-firm decision? That is Chapter 7's question. Constellation on Solana, FOCIL and Flashnet on Ethereum, are the protocol-level proposals being built against it.

Footnotes and sources ​

  1. Decentralised.co, Why Jito Suspended Its Mempool on Solana — What's Next?, March 2024, https://www.decentralised.co/p/mev-on-solana. Accessed 2026-05-14. The same source documents that searcher solicitations to validator operators began within months of the shutdown. ↩︎ ↩︎

  2. CoinDesk, Solana Client Developer Jito Announces End of Mempool Function, 8 March 2024, https://www.coindesk.com/business/2024/03/08/solana-client-developer-jito-announces-end-of-mempool-function. Decentralised.co dates the formal shutdown to 9 March 2024; the difference is a time-zone artefact. Accessed 2026-05-14. ↩︎ ↩︎

  3. Jupiter, Ultra Swap API Documentation, https://developers.jup.ag/docs/ultra; Bravian Oyaro, Jupiter's MEV Protection Architecture (2026), https://medium.com/@bnyamosie/jupiters-mev-protection-architecture-how-solana-traders-get-safer-onchain-execution-34a5bccb33be. The +0.006% (protected) and −0.14% (unprotected) figures are Jupiter's own published Ultra V3 measurements. Jupiter handles approximately 95% of Solana aggregator-routed volume and roughly 50% of Solana DEX volume; over $700 million in daily swap volume routes through it. Accessed 2026-05-14. ↩︎ ↩︎

  4. Helius Research, Solana MEV Report, https://www.helius.dev/blog/solana-mev-report, documents the approximate 200-millisecond preview window. The timeline — Jito-Solana on mainnet and testnet in late 2022 (validators only), Block Engine and broad searcher access opening 25 May 2023 (after the prior software cohort had processed over a million bundles) — is documented by Jito Labs' own announcement and by Solana Compass coverage. Accessed 2026-05-14. ↩︎

  5. Helius Research, Solana MEV Report, op. cit. Vpe's December 2024 – January 2025 window: 1.55M sandwich transactions, 88.9% success rate, approximately $13.4M gross profit, approximately $4.6M paid to validators via Jito tips. Chapter 4 returns to these figures from the searcher's side. Accessed 2026-05-14. ↩︎

  6. Decentralised.co, op. cit. Solicitations from searcher firms to validator operators are documented in the same source. ↩︎

  7. CoinDesk, Solana Heavyweights Wage War Against Private Mempool Operators, 10 June 2024, https://www.coindesk.com/business/2024/06/10/solana-heavyweights-wage-war-against-private-mempool-operators. Tim Garcia is identified as the Solana Foundation's validator-relations lead. The ~32-validator / ~1.5M-SOL / ~0.5%-of-delegation-program-stake numbers are from the same source. Accessed 2026-05-14. ↩︎

  8. New World Finance, This Solana validator extracted over $13M from users in one month using sandwich attacks, 24 March 2025, https://www.newworldfinance.io/solana-validator-sandwich-attacks-mev/, citing Jito's internal analysis. DeezNode's validator address is HM5H6…jdMRA; its associated sandwich program is vpeNAL…oax38b ("Vpe"). Validator stake rose from 307,900 SOL (epoch 697, 13 November) to 802,500 SOL (epoch 709, 9 December). The 50% revenue-share offer to opting-in validators is documented in independent on-chain analysis at Anarcaze, Spotting Shady Validators on Solana, https://anarcaze.medium.com/spotting-shady-validators-on-solana-a-friendly-guide-f39ef8b32a00. Note: the Vpe / DeezNode linkage is sourced to secondary on-chain analysis, not a Jito Labs direct disclosure. Accessed 2026-05-14. ↩︎

  9. Kolibrio Documentation, Orderflow Service, https://docs.kolibrio.xyz/; QuickNode, Feature Fridays: Kolibrio, https://blog.quicknode.com/feature-fridays-kolibrio/; AIN.ua, Ukrainian crypto startup Kolibrio raises $2M, 16 February 2023, https://en.ain.ua/2023/02/16/ukrainian-crypto-startup-kolibrio-raises-2m-to-build-web3-service/. Kolibrio operates an RPC-layer Order Flow Auction: integrators (wallets, trading bots, solvers) route user transactions through Kolibrio's RPC; Kolibrio auctions pre-leader transaction visibility to searchers and rebates a share back to the integrator. Founded early 2022 by Anatolii Padenko (CEO) and Alex Starikov; $2 million seed in February 2023 led by Jump Crypto, with Delta Blockchain Fund and Everstake Capital. Distinct from the validator-side operators the Solana Foundation targeted in June 2024 — Kolibrio's surface is at the RPC layer, not the validator layer. Accessed 2026-05-14. ↩︎

  10. Helius, Infrastructure and APIs for Wallets, https://www.helius.dev/use-case/wallets; Helius, Zero-Slot Execution with Sender and LaserStream, https://www.helius.dev/blog/zero-slot. Helius runs the RPC pipeline for Phantom, Backpack, Solflare, Bitgo, Ledger, Exodus, Squads, and Trust wallets, among others. The company markets "SOL rebates from trades that create arbitrages" and operates as the network's largest single validator by stake (15M+ SOL). LaserStream's "shred-level ingestion" is the functional Solana equivalent of mempool visibility, gated to paid tiers ($499/mo Business; $999/mo Professional). The chapter does not assert that Phantom, Backpack, or Solflare themselves directly sell their users' pre-broadcast transaction flow to searchers; the assertion is that the wallet → Helius → MEV-rebate pipeline exists publicly and is the load-bearing wallet-infrastructure-level visibility surface in 2026. Accessed 2026-05-14. ↩︎

  11. RPC Fast, Co-location Strategies for Solana Trading in 2026, https://rpcfast.com/blog/solana-trading-co-location-low-latency. The "non-voting validator node alongside an RPC in the same rack, forwarding transactions to the current leader's TPU via QUIC" pattern is described as a standard practitioner pattern. The phrase "next to the TPU port" is the author's framing of this architectural choice. Accessed 2026-05-14. ↩︎

  12. Coverage of the 8 April 2026 Solana protocol-level patch: Edgen, Solana Network Shuts Down Sandwich Attacks on April 8, https://www.edgen.tech/news/post/solana-network-shuts-down-sandwich-attacks-on-april-8-boosting-user-security. Multiple secondary sources confirm the date and the effect; the specific SIMD, the validator-client release tag, and the precise consensus-layer mechanism are not publicly documented in any source the chapter could verify. The chapter cites the patch's effect without committing to mechanism, and flags this as an item for verification before publication. Accessed 2026-05-14. ↩︎

  13. Chainflow, Constellation and the Future of Block Production on Solana, Q1 2026, https://chainflow.io/constellation-and-the-future-of-block-production-on-solana-what-validators-need-to-know/. Constellation is a proposal from Helius; not officially adopted on Anza's roadmap. Mainnet activation is not before the post-Alpenglow timeframe — Alpenglow itself targets late Q3 / early Q4 2026 per Chapter 1, footnote 10. Accessed 2026-05-14. ↩︎

  14. Jito Labs, Introducing BAM: The Future of Block Building on Solana, 21 July 2025, https://bam.dev/blog/introducing-bam/; Helius Research, Block Assembly Marketplace (BAM), https://www.helius.dev/blog/block-assembly-marketplace-bam. AMD SEV-SNP enclave details, the 6% tip split, the Q2 2025 DAO share figure (~22,391 SOL ≈ $4M), and the TEE overhead range (~2–5%) are all drawn from these two sources. The privacy guarantee depending on hardware attestation rather than cryptography is the chapter's framing, supported by the Helius technical description. 2025 update via JIP-24 (passed by Jito DAO in 2025; see Jito Forum — JIP-24): the entire 6% Block Engine and BAM tip take now routes to the Jito DAO treasury; Jito Labs (the company) no longer takes a direct cut of these flows. The Q2 2025 numbers cited above were pre-JIP-24. Chapter 6 develops the firm vs. DAO restructuring. Accessed 2026-05-14. ↩︎

  15. Blocmates, BAM: Solana's Block Builder Era, https://www.blocmates.com/articles/bam-solana-s-block-builder-era. Jump Crypto's BAM Maker Priority plugin launch (April 2026), and the description of ACE plugins generally, are drawn from this source. Accessed 2026-05-14. ↩︎

  16. Harmonic, Overview, https://harmonic.gg/posts/overview; The Block, Paradigm leads $6 million seed round for Harmonic, 5 November 2025, https://www.theblock.co/post/377791/paradigm-harmonic-funding-solana-nasdaq-speed; GlobeNewswire, DeFi Development Corp. Partners with Harmonic, 1 December 2025, https://www.globenewswire.com/news-release/2025/12/01/3196994/0/en/DeFi-Development-Corp-Partners-with-Harmonic-to-Optimize-Solana-Validator-Revenue-and-Network-Participation.html. Harmonic founded by Ben Coverston and Jakob Povšič (Povšič also co-founded Temporal, one of the builders Harmonic auctions). The "configurable preferences" framing is from Harmonic's overview page; the partner-exclusivity inference is the chapter's reading of how that preference architecture can be used, not a documented practice by Harmonic itself. Accessed 2026-05-14. ↩︎

  17. Harmonic, Open block building infrastructure for Solana, https://harmonic.gg/posts/overview (closed beta; landing page accessible). Stake-share and priority-fee-capture numbers are from Syndica's March 2026 deep dive (see footnote 11). Accessed 2026-05-14. ↩︎

  18. Flashbots, Flashbots Protect Documentation, https://docs.flashbots.net/flashbots-protect/overview. 2.1M+ accounts, $43B+ cumulative shielded DEX volume, ~7% of Ethereum monthly transaction count are Flashbots' self-reported figures. The Protect mechanism — wallet RPC URL change, transactions submitted to the Flashbots relay as private bundles, refund-share if backrunning value is captured — is documented in the same source. Accessed 2026-05-14. ↩︎ ↩︎

  19. CoW DAO, SMG Acquires MEV Blocker RPC, January 2026, https://cow.fi/learn/special-mechanisms-group-acquires-mev-blocker-rpc-to-advance-state-of-the-art-backrunning-auction-infrastructure. 4.5M+ unique wallets served, 6,177 ETH cumulative rebates paid since launch. The MEV Blocker RPC was transferred from CoW DAO to Consensys' Special Mechanisms Group in January 2026. Accessed 2026-05-14. ↩︎

  20. Wahrstätter, Pierro, et al., Private Order Flows and Builder Bidding Dynamics, arXiv:2410.12352 (v3, 2025), https://arxiv.org/abs/2410.12352. Approximately 30% of Ethereum transactions by count, more than 50% of gas, and 54.59% of total block value flow through private channels. Accessed 2026-05-14. ↩︎

  21. Flashbots / relayscan.io, MEV-Boost Relay & Builder Statistics (24-hour snapshot ending 14 May 2026), https://www.relayscan.io/. Relay shares: Ultrasound 35.71%, Titan 26.09%, BloXroute Max-Profit 13.38%, BloXroute Regulated 12.30%, Aestus 7.32%, Flashbots 2.09%. Builder shares: Titan 50.60%, BuilderNet 24.55%, Quasar 15.68%, Beaverbuild 1.83%. These are live numbers and will drift. Accessed 2026-05-14. ↩︎ ↩︎

  22. Flashbots, Migrating to BuilderNet, 9 December 2024, https://writings.flashbots.net/migrating-to-buildernet; Flashbots, decentralized building: wat do?, 14 February 2026, https://writings.flashbots.net/. The 80–90% top-two builder share quote is from the February 2026 piece. Accessed 2026-05-14. ↩︎

  23. Messari, State of the Superchain H1 2025, https://messari.io/report/state-of-the-superchain-h1-2025; The Block, Base Defends Moving Sequencer Fees to Coinbase Custody, https://www.theblock.co/post/339698/base-defends-moving-sequencer-fees-to-coinbase-custody-for-security-and-audit-reasons. Base sequencer revenue H1 2025: $42.4M (87.2% of Superchain sequencer revenue). Daily priority fees average ~$156,138, ~86% of daily sequencer revenue. Coinbase has publicly stated that it does not directly extract sandwich MEV via the sequencer; the priority-fee competition surface remains permissionless. Accessed 2026-05-14. ↩︎

  24. Offchain Labs, Timeboost: A Gentle Introduction, https://docs.arbitrum.io/how-arbitrum-works/timeboost/gentle-introduction; Express Lane to Spam: A Study of Arbitrum's Timeboost Auction, arXiv:2509.22143, https://arxiv.org/abs/2509.22143. Timeboost activated April 2025. The 200ms express-lane advantage, sealed-bid auction, and Arbitrum DAO proceeds are documented in the protocol docs. The February 2026 Kairos resale-market dynamic and the subsequent reserve-price adjustment are documented in the arXiv paper. Accessed 2026-05-14. ↩︎ ↩︎

  25. Optimism Foundation, Optimism Partners with Flashbots, https://www.optimism.io/blog/optimism-partners-with-flashbots. Unichain runs Flashblocks with 200ms block-intermediary times and returns MEV to users via revenue-share. OP Mainnet contributes 100% of sequencer profit to the Collective; other OP Chains contribute either 2.5% of revenue or 15% of profit, whichever is greater. Accessed 2026-05-14. ↩︎

  26. Caldera, Caldera and Espresso Systems Unveil First Optimistic Rollup Using Espresso's Decentralized Sequencer, mid-2025, https://blog.caldera.xyz/caldera-and-espresso-systems-unveil-first-optimistic-rollup-using-espressos-decentralized-sequencer/. Espresso's full general-availability launch was 12 February 2026, with ~6-second block finality. Astria's mainnet alpha launched on Celestia DA on 28 October 2024. As of May 2026, no major L2 had migrated to either. Accessed 2026-05-14. ↩︎

  27. Hyperliquid Documentation, HyperCore Overview, https://hyperliquid.gitbook.io/hyperliquid-docs/hypercore/overview. Order-book semantics inside HyperBFT consensus: non-order actions process first within a block, then cancels, then GTC/IOC sends. Accessed 2026-05-14. ↩︎

  28. EIP-7805 (Fork-Choice enforced Inclusion Lists), https://eips.ethereum.org/EIPS/eip-7805; Ethereum Foundation, Checkpoint #9, 10 April 2026, https://blog.ethereum.org/2026/04/10/checkpoint-9; Flashbots, Network Anonymized Mempools (Flashnet), 17 February 2026, https://writings.flashbots.net/network-anonymized-mempools. FOCIL was selected as the consensus-layer headliner for Hegotá (the post-Glamsterdam Ethereum upgrade); Glamsterdam targets 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. Accessed 2026-05-14. ↩︎