Rebuilding Crypto's Market Structure: Eugene Chen's Vision for Atlas, Phoenix, and Solana's Future

The conversation around decentralized exchange infrastructure has entered a new phase. While the crypto industry continues to debate the merits of automated market makers versus central limit order books, a quieter revolution is taking place in how professional traders think about on-chain market structure. At the center of this evolution stands Eugene Chen, co-founder of Ellipsis Labs, whose work on Phoenix, Atlas, and the recently revealed SolFi is reshaping how liquidity functions on Solana.

Top Projects

Lulo

Score: 7,586.13

Bonk

Score: 4,806.12

Sphere

Score: 2,544.42

Moonwalk

Score: 2,026.11

In a wide-ranging conversation on the Lightspeed podcast, Chen articulated a vision that challenges many assumptions held dear by crypto purists. His perspective, shaped by a background in mathematics and a deep understanding of traditional finance market mechanics, offers a pragmatic blueprint for what he calls "verifiable finance"—a framework that prioritizes user outcomes over ideological commitments to maximum decentralization.

The stakes of this conversation extend far beyond technical architecture. As Solana positions itself as the home for "internet capital markets," the question of whether its market structure can support sophisticated financial activity becomes paramount. Chen's insights illuminate both the current limitations and the potential pathways toward a future where on-chain trading can genuinely compete with centralized venues.

The Evolution of Ellipsis Labs

Ellipsis Labs has undergone significant evolution since its founding, building multiple products that address different aspects of the on-chain trading problem. Phoenix, their first major product, launched as a fully on-chain limit order book exchange on Solana and has since facilitated approximately $75 billion in lifetime notional volume. This remarkable figure demonstrates that order book infrastructure can achieve meaningful scale in decentralized environments.

Beyond Phoenix, the team developed Gavil, a token origination platform designed to facilitate fully on-chain token launches without the traditional dependencies on market makers or centralized exchanges. Chen describes this as addressing a fundamental problem in crypto: "There are all these people you have to bribe, all these centralized exchange listing teams. They technically don't take listing fees, and so you have to put your coin on their launch pad or whatever. You have to pay off the market makers. And these things add up to sometimes more than 10% of your total supply."

Most recently, Ellipsis Labs revealed that they are the developers behind SolFi, which Chen describes as "the largest on-chain market maker on Solana." This revelation connected dots that only the most dedicated observers of Solana market structure had begun to notice—the emergence of proprietary AMMs capturing significant order flow for major trading pairs.

The common thread connecting these products is Chen's conviction that on-chain markets can and should compete with their centralized counterparts on execution quality. This belief drives every architectural decision the team makes.

Why Phoenix Led to Atlas

Phoenix was always intended as a proof of concept rather than a final destination. Chen explains that the exchange launched during a period when "the total on-chain volume on Solana was like $20, $30 million per day. And today that number is like a hundred times higher." This dramatic growth in activity revealed limitations in Solana's base layer that weren't apparent at lower volumes.

The fundamental challenge Chen identifies is what he calls "the proposer monopoly"—the fact that for each 1.6-second period on Solana, a single validator has complete control over transaction ordering within their block. "Just like some random guy around the world is running your matching engine for 1.6 seconds, they can fuck with everything they want to fuck with. And then even if they're nice to you, 1.6 seconds later, there's somebody else doing it."

This architectural reality creates problems for anyone attempting to run an efficient exchange. Market makers need predictable execution to provide tight spreads. When any validator can reorder transactions to their advantage, the entire market-making operation becomes more expensive and less efficient, costs that ultimately get passed to end users through wider spreads.

Atlas represents Ellipsis Labs' attempt to solve this problem by building a layer two blockchain with an "extremely opinionated" design optimized specifically for high-frequency trading applications. Rather than competing as general-purpose block space, Atlas will be purpose-built for exchange infrastructure.

The Centralized Sequencer Trade-off

One of the most striking aspects of Chen's approach is his explicit rejection of decentralization as a primary design goal for Atlas. The blockchain will run with a centralized sequencer operated by Ellipsis Labs, a choice that might seem heretical in a space that often treats decentralization as an unqualified good.

Chen's reasoning is characteristically pragmatic. "We're not trying to compete on being the most decentralized. And the actual property that we think is important is you don't need to trust the operator or the sequencer for everything. The system will guarantee that the state transitions are all valid. The system will guarantee that if you want to take your money out and the operator is being malicious, you can still take it out."

This framework distinguishes between different types of trust assumptions. A centralized exchange requires users to trust that the operator won't steal their funds and will execute the matching engine fairly. A layer two like Atlas still requires trust in the sequencer's benevolence regarding transaction ordering, but the underlying cryptographic guarantees mean users can always withdraw their assets regardless of operator behavior.

Chen introduces the term "verifiable finance" to describe this middle ground: "It would be disingenuous to call it decentralized when there's a single permission operator who's going to be running the sequencer, which is us. And it definitely grinds my gears when I see other teams with some sort of clearly centralized dimensions products, labeling it as decentralized."

Multiple Concurrent Leaders and Solana's Path Forward

Solana's development roadmap includes multiple concurrent leaders (MCL), a proposed change that would distribute the proposer monopoly across multiple validators simultaneously. Chen acknowledges this addresses the problem "to some degree" but remains unconvinced it provides the same level of control as a centralized sequencer.

The key insight is that MCL forces transaction ordering to be determined by the protocol rather than by individual leaders. "You can already do this with a single leader where the leader chooses which transactions go in, but they don't get to control the ordering. And then in that world, it turns out they can still do a lot of things to mess with the contents of a block to make money."

With multiple concurrent leaders, censorship becomes more difficult, and the ability for any single validator to manipulate execution decreases. However, the fundamental abstraction of discrete block times remains, whereas a centralized sequencer can implement "truly continuous block building."

For applications like Phoenix that require the tightest possible spreads and most reliable execution, this difference matters. Chen suggests that different applications have different requirements along the decentralization-efficiency spectrum, and the market should offer options across this range.

Designing Exchanges for Market Maker Welfare

A recurring theme in Chen's analysis is the importance of market maker welfare to overall market quality. This perspective might seem counterintuitive—shouldn't exchanges prioritize the traders who are actually buying and selling? Chen argues that this framing misunderstands the dynamics of efficient markets.

"The reason you want to bias towards the market makers, rather than to the takers... you have to think back to what is the actual purpose of the exchange. And for us, for Phoenix, it's for retail users to get the best possible prices. And you want market makers to be competing on how good of a price I can give to the user at whatever depth."

The problem with toxic flow—high-frequency traders who specialize in picking off stale quotes—is that they force market makers to protect themselves by widening spreads. If the HFT takers are "doing really well, that forces the market maker on average to provide liquidity slightly wider." This cost ultimately falls on retail users who receive worse execution.

Atlas addresses this through specific transaction ordering mechanisms. Oracle updates should land "really, really reliably," meaning the sequencer will prioritize them even when bottlenecked. Similar priority can be given to cancellations and order placements. These design choices systematically favor liquidity provision over toxic extraction.

The Perpetuals Advantage for New Chains

When asked whether Atlas needs to prioritize perpetual contracts to compete as an exchange, Chen offers a nuanced explanation of why perps represent an ideal starting point for new blockchain infrastructure.

The key advantage is that perpetuals abstract away the asset bridging problem. "For a spot [market], I need to bridge over USD and then I need to bridge over those 10 coins as well. And so you're going to have like 11 different assets on the chain, whereas if all we're doing is perps, then I need to bridge over a collateral asset, which is probably some form of USD. And then the way we trade the perps is through an oracle."

This architectural simplification has meaningful implications for bootstrapping liquidity on a new chain. The oracle "effectively is a thing that's bridging the pricing information over," while spot markets require physical bridging of each underlying asset. For a layer two attempting to establish itself, this reduction in complexity can be decisive.

Chen also notes that perpetuals have "very clear product market fit" as "the best known form of leverage... just significantly better than options as a product for retail traders."

Why Ethereum for Settlement

The decision to settle Atlas on Ethereum rather than Solana might surprise observers given Ellipsis Labs' deep roots in the Solana ecosystem. Chen's explanation reveals a sophisticated understanding of what properties matter for different layers of the blockchain stack.

"The properties you care about for a settlement layer in theory are very different from the properties you care about in an execution layer," Chen explains. Settlement layers need strong censorship resistance—the ability to force through transactions even if the sequencer goes down—along with robust revert protection and fast finality for bridging.

Ethereum has explicitly optimized for these settlement layer properties. "Totally separate question [whether] that's like a good thing for Ethereum to do or not, but they've done it. And yeah, I think Ethereum's a pretty good settlement layer. And basically all the other blockchains do not care about it."

This pragmatic assessment sidesteps tribal allegiances in favor of matching infrastructure to use case. Chen emphasizes that "the importance of the settlement layer is like a little bit overrated," noting that "most of the layer twos [are] not doing any sort of true settlement anyway."

The State of Layer Two Decentralization

Chen's critique of existing layer two claims extends beyond terminology to fundamental properties of these systems. He points to L2Beat.com as a valuable resource for understanding the actual trust assumptions of various rollups, noting that "those guys are really great. They're nits."

The core problem is that many systems "are just lying to the users about what properties that they have." A common example: "People say these systems are decentralized or non-custodial. And it turns out just like one guy has the upgrade key, which means they can upgrade the bridge contract to take all the money out. So seems kind of custodial."

For Chen, the truly non-custodial version requires that "even if they went rogue and wanted to take your money out, they can't do it." Most layer two chains, "even some things that are calling themselves layer one chains today," fail to meet this standard.

This concern about honest labeling connects to Chen's broader philosophy about crypto's relationship with users. "It's one of those things where 99% of the time it doesn't matter. And then sometimes the retail guy gets hurt. And if this industry is going to sell for a [certain vision], I think like definitions is a pretty easy place to start."

SolFi and the Rise of Proprietary AMMs

The revelation that Ellipsis Labs built SolFi illuminates a significant shift in Solana's market structure. Beginning in late 2024 and early 2025, observers noticed proprietary AMMs emerging that captured substantial order flow for major pairs like SOL/USDC. These venues offered better prices than traditional AMMs, but operated with less transparency about their internal mechanisms.

Chen contextualizes SolFi within the evolution of market making on Solana. Ellipsis Labs was already "one of the market makers, sometimes largest, not always at the time, probably the largest market maker on Phoenix." They observed the competitive dynamics—toxic flow hitting the exchange, competition for block space among market makers, high compute requirements translating to higher tips for transaction landing.

"In the same way that Atlas came from our understanding of the Solana market structure, SolFi was our other response. So Atlas is our way to fix it and SolFi is our way to... if this is the game that is being played, we're going to play it the best."

The technical interface is straightforward: "At any time, Jupiter or anyone else can query the SolFi contract and say, 'Hey, I want to take some amount of Sol and I want to receive as much USDC as you'll give me, how much can I get?'" This swap interface matches what any other DEX exposes, but SolFi consistently provides more competitive prices than Phoenix or other venues.

Jupiter's Role in Shaping Market Structure

A fascinating aspect of Chen's analysis is the credit he gives to Jupiter for enabling the proprietary AMM phenomenon. The aggregator's commitment to routing flow to venues offering the best prices creates the competitive environment that makes SolFi viable.

"This market structure is induced by Jupiter caring about the quality of their fills. Because I think, yeah, I think it's like quite noble from them, to be honest, where clearly their end users on the website just like do not care if they get like half a basis point of price improvement. There's no way they care. There's no way they know."

Despite user insensitivity to marginal execution quality, Jupiter has chosen to make price the primary criterion for order routing. "If Jupiter made other decisions, then I think the market structure will look very different."

This highlights a crucial point about ecosystem design: the choices made by key infrastructure providers shape the incentives for everyone building on top. Jupiter's commitment to execution quality enables business models like SolFi that compete purely on the quality of their pricing.

The Sandwich Attack Problem and Gavil

Gavil emerged from Ellipsis Labs' observation that sandwich attacks were "really taking off on Solana, primarily because there was a lot of trading in these low liquidity meme coin pools." The team concluded that the problem couldn't be solved at the infrastructure layer because "the validator does have full agency over what they put into their own blocks."

The sandwich-resistant AMM design represented an application-layer solution. Chen's team wrote a reference implementation and approached order flow originators—meme coin launch pads and similar platforms—with a compelling pitch: "If you take your coin and instead of graduating it to a regular AMM, you graduate to the sandwich-resistant AMM, then your users are literally going to make more money. The coins are almost certainly going to have higher market caps because you don't have a sandwicher just taking money out of the LP."

The response was disappointing. "We thought it was like a very clear win, and maybe we didn't sell it well. Maybe this is just not a top five problem for any of these guys."

This experience informs Chen's broader understanding of the crypto market: user insensitivity to execution quality creates market structures that prioritize other concerns. The parallel to traditional finance is striking—sports betting companies charge wide spreads not because users demand them, but because users don't resist them, and the resulting margins fund customer acquisition instead.

Token Launch Infrastructure and Market Quality

Gavil's mission extends beyond sandwich resistance to address broader problems in token origination. Chen observes that most existing products for launching tokens "basically exist to fleece the end user as much as possible, whether it's for the sake of the platform or for the sake of the team."

The central issue with sniping is that "the initial price was way below the market price." This happens by design: teams want charts that "look good," which typically means starting low and rising. But this creates extraction opportunities that benefit sophisticated actors at the expense of retail participants.

Chen proposes an alternative: "If you want to facilitate good price discovery, the expected future price of the token might actually just be the opening price of the token." Technical mechanisms like Dutch auctions can achieve this, after which tokens can graduate to sandwich-resistant AMMs and eventually to more sophisticated venues like Phoenix.

The deeper critique targets the hypocrisy of an industry that "likes to talk about, 'oh, we're so transparent and everything goes on the blockchain.' And then you have this whole centralized exchange business where it seems like everyone who's launching a token is signing 12 of these side deals with some really, really shady actors."

Even basic applications of blockchain technology, like token vesting contracts, frequently occur inside centralized custody protocols rather than on-chain. "That is the canonical example of the type of thing that you can do on a blockchain, some sort of conditional escrow. And even those are done inside custody protocols. They're not done on-chain."

Payment for Order Flow in Crypto

The comparison between crypto payment for order flow and its traditional finance counterpart reveals regulatory gaps with real consequences for users. Chen explains that while front-running is "generally quite illegal in traditional finance," there is regulated payment for order flow where "the prices that you get that the end user gets have to be as good or maybe even strictly better than the prices on the lit books."

In crypto, no such protections exist. A wallet could receive a signed transaction from a user and, instead of broadcasting it to the network, "sell it to a MeV shop. Or maybe I sell it in bulk. I'm going to sell you a whole month of transactions for a million dollars a month."

The MeV shop can then land the transaction while adding components "in front and behind to force the users to get the worst possible fill." This can happen "at any part of the stack once the transaction has been signed. So wallet and beyond, wallet could do it, RPC could do it, validators who are receiving transaction flow, they can do it."

The result is that "payment for order flow in crypto is much more about giving the user the worst possible fill and price versus price improvement." Traditional finance's regulated version improves user outcomes; crypto's unregulated version extracts from them.

The Rakes and Rebates Dynamic

Chen identifies a structural similarity between crypto trading apps and other industries that monetize user insensitivity. Credit card points provide a familiar example: users are relatively insensitive to fees but responsive to rewards, so optimal strategy involves maximizing fees while returning value through points.

"The user is insensitive to the fee, but somewhat sensitive to the incentive. And so then you want to maximize fee and maximize incentive." This creates an "inefficient market structure" where users pay more in fees than they would in a more transparent pricing regime, but feel better about the transaction because of rebates.

Sports betting demonstrates the same dynamic. "The spreads on sports bets are extremely, extremely wide. It doesn't mean that the sports betting companies just make all that money. In fact, it just all goes back into customer acquisition and customer retention."

For crypto trading apps, this means charging maximum fees—or equivalently, providing the worst fills users will accept—and then competing on rebates and promotions rather than on execution quality. Users who don't understand the underlying economics become the product rather than the customer.

Regulatory Implications and Industry Self-Regulation

Chen expresses concern about how regulators might respond once they fully understand crypto market structure. "Regulators once they like really understand what is happening in these types of systems, you know, part of their job is to ensure integrity of the markets and also to ensure that retail finance users are not getting boned. And yeah, some of them are probably getting boned in crypto today."

The question of what should be done remains open. Chen doesn't advocate for simply copying traditional finance regulations but acknowledges that the current state—where validators are incentivized "to be maximally greedy on behalf of their stakers" and "the incentive is just to maximally fuck the user whenever you can"—is incompatible with the vision of bringing serious financial activity on-chain.

Ethereum's approach of private channels to block builders, with trust relationships enforced by reputation or cryptography, represents "a step in the right direction" even if imperfect. Solana's market structure has evolved differently, with recent improvements in landing rates potentially reducing the incentive to broadcast transactions widely and thus limiting front-running opportunities.

Building Products Users Actually Need

Throughout the conversation, Chen returns to a central tension: building technically superior products doesn't guarantee adoption if users don't care about the relevant dimensions of quality. Coinbase fees haven't been competed down despite years of predictions; Robinhood fees have actually increased over time.

This observation might seem discouraging for someone building more efficient market infrastructure. Chen's response reveals his longer-term thinking: "I think in the traditional world, we see tons and tons of demand when, you know, professionals are trading the assets that actually matter, you know, revenue generating and revenue collecting assets where people have real risks to manage."

The current state of crypto trading, dominated by meme coins and retail speculation, represents a bootstrapping phase rather than the destination. "If what crypto delivers to the world is, you know, this shit coin game, maybe even gets like 10 times bigger or 100 times bigger, then I would view that all as just like a total failure."

Real financial flows—corporate treasuries, institutional asset management, sophisticated risk transfer—require execution quality that current on-chain infrastructure often can't provide. Chen's conviction is that this demand will materialize as crypto matures, and Ellipsis Labs is positioning Atlas and Phoenix for that future.

Market Quality as a Prerequisite for Institutional Adoption

The connection between market structure and institutional adoption runs in both directions. Sophisticated traders require high-quality execution to participate, but their participation also improves market quality through competitive pressure.

"You're gonna need the market quality to be high for, you know, the really sophisticated traders to show up," Chen explains. This creates a chicken-and-egg problem that must be solved through intentional infrastructure design rather than organic evolution.

Current on-chain token launches exemplify the problem. Chen argues that "some of these internet capital markets products have been designed in a way such that like no halfway sensible, halfway serious team would ever choose to launch something on them." This design is often intentional—extractive platforms profit from the extraction—but it prevents the ecosystem from maturing.

Atlas represents a bet that building infrastructure optimized for professional traders will eventually attract the flow that justifies the investment. Asset selection and quality matter as much as execution quality: Phoenix on Atlas won't succeed by trading the same meme coins currently dominating Solana volume.

The Technical Architecture of Atlas

While Chen keeps some details close to the chest, he reveals several design principles guiding Atlas development. The layer two is "extremely opinionated" about its use case: exchange infrastructure and similar high-frequency applications.

Oracle updates receive highest priority from the sequencer, ensuring that price feeds remain current even under heavy load. This matters because stale oracles create arbitrage opportunities that cost liquidity providers money. Cancellations and new orders can also receive priority treatment, biasing the system toward market maker welfare.

The centralized sequencer enables "truly continuous block building," eliminating the discrete block time abstraction that limits even Solana with multiple concurrent leaders. For a limit order book where microseconds matter, this represents a meaningful improvement.

Settlement on Ethereum provides the security guarantees that a layer two requires: if the Ellipsis Labs sequencer becomes malicious or goes offline, users can still withdraw their assets through the underlying layer one. This design acknowledges that most users, most of the time, will trust the sequencer—but builds in guarantees for the cases where trust breaks down.

The Path to Atlas Launch

When asked about timeline, Chen indicates that Atlas is "coming soon. We can see the light at the end of the tunnel in the next few months." Phoenix will become "the flagship product on top of Atlas," representing the next evolution of the limit order book exchange.

The relationship between Phoenix on Solana and Phoenix on Atlas remains somewhat undefined. Chen notes that "we haven't done too much active maintenance of the current product for a while," suggesting resources have shifted toward the layer two build.

For the Solana ecosystem, this represents both a loss and an opportunity. Losing a high-quality exchange to another execution environment isn't ideal, but the architectural learnings from Atlas could eventually inform Solana's own development. And SolFi's continued presence demonstrates Ellipsis Labs' ongoing commitment to improving Solana market structure.

Honest Assessment of Crypto's Current State

Chen's willingness to acknowledge crypto's shortcomings reflects a maturity often absent from industry discourse. His assessment of order book versus AMM debates: "The current discourse, as usual, is just like missing all of the nuance... there's this attempt to latch on to, oh, there's this current thing that's doing extremely well, which is Hyperliquid. And how do we mold that into, you know, my narrative for my token."

This critique extends to the broader "internet capital markets" narrative. While Chen shares the vision of bringing real financial activity on-chain, he recognizes that current infrastructure often fails the test: "What is crypto delivers to the world is, you know, this shit coin game... I would view that all as just like a total failure."

Yet this honest assessment coexists with genuine optimism. The problems are solvable. The demand exists, even if currently latent. Teams willing to build with precision and maintain honest communication about trust assumptions can create infrastructure worthy of the vision.

SolFi's Future and the Proprietary AMM Landscape

The proprietary AMM model that SolFi represents is likely to proliferate. Chen notes that the market structure "is not necessarily a new trend"—Lifinity pioneered similar approaches in 2021—but current conditions have made it more prominent.

Asked whether SolFi will eventually provide tighter spreads through Phoenix on Atlas than through its current Solana integration, Chen offers a nuanced response: "In some equilibrium, yes, but in equilibrium that we're going to push towards, likely not. I think things will not be, you know, like so aggregator-based, will not be so fragmented because, you know, we're like driving the whole thing."

This suggests Ellipsis Labs envisions a future where their various products integrate more tightly, with less reliance on third-party aggregators to route flow between fragmented venues. The competitive advantage shifts from winning the aggregator's routing algorithm to controlling the entire stack.

Lessons for Builders in Crypto

Chen's approach offers several implicit lessons for crypto builders. First, understand the game being played before trying to change it. Ellipsis Labs built SolFi to win within current market structure while building Atlas to change that structure.

Second, be honest about trust assumptions. Users may not care about decentralization, but lying about what properties your system actually has eventually damages the entire industry's credibility.

Third, target products at specific users with specific needs. "When we think about building product, the product is always for some user. So Phoenix is for traders and Atlas is for applications that look very similar to Phoenix." Products without clear target audiences struggle to make coherent design decisions.

Fourth, recognize that the most important users might not be the current users. Chen's focus on institutional-quality execution targets traders who don't yet participate on-chain but whose eventual arrival will define crypto's long-term relevance.

The Broader Vision for Verifiable Finance

Chen's concept of verifiable finance represents a middle path between crypto's maximalist dreams and traditional finance's regulatory apparatus. Not every system needs Bitcoin's level of censorship resistance. Not every system needs Coinbase's convenience. The design space includes many points between these extremes.

For trading infrastructure specifically, verifiability means users can confirm state transitions are valid and can exit the system regardless of operator behavior. Centralized sequencing can coexist with these properties while enabling execution quality that decentralized alternatives can't match.

This framework may disappoint those who hoped crypto would revolutionize finance through maximum decentralization of everything. But it offers a realistic path toward meaningful improvement: lower costs than traditional finance, better transparency than centralized exchanges, and execution quality that serves real economic activity rather than just speculation.

Conclusion: Building for the Future Crypto Deserves

Eugene Chen's work at Ellipsis Labs embodies a particular philosophy about crypto's development. Technical excellence matters. Honest communication about trade-offs matters. Building for future use cases, not just current ones, matters.

Atlas represents a bet that the market for professional-grade on-chain trading infrastructure will eventually justify the investment required to build it. Phoenix demonstrated that limit order books can work on-chain; Atlas aims to prove they can compete with centralized venues on every dimension that matters.

SolFi and Gavil address more immediate problems: how to provide competitive prices within Solana's current constraints, how to launch tokens without the extractive mechanisms that currently dominate.

Together, these products outline a vision where verifiable finance serves real economic activity—treasury management, institutional trading, sophisticated risk transfer—rather than just meme coin speculation. Whether that vision materializes depends on many factors outside Ellipsis Labs' control. But if it does, the infrastructure will be ready.

Facts + Figures

• Phoenix lifetime volume: Approximately $75 billion in total notional lifetime volume traded on the fully on-chain limit order book exchange on Solana.

• Solana volume growth: When Phoenix launched in early 2023, total on-chain volume on Solana was $20-30 million per day; today that number is approximately 100 times higher.

• Proposer monopoly duration: On Solana, a single validator controls transaction ordering for 1.6-second periods, creating the "proposer monopoly" problem that Atlas aims to solve.

• SolFi revelation: Ellipsis Labs announced they are the developers behind SolFi, described as "the largest on-chain market maker on Solana."

• Token launch costs: Market maker payments, centralized exchange listing arrangements, and other token launch requirements can add up to "more than 10% of total token supply."

• Atlas settlement: Atlas will settle on Ethereum rather than Solana, based on assessment that Ethereum has optimized for settlement layer properties including censorship resistance and fast finality.

• Atlas timeline: Atlas is expected to launch "in the next few months" according to Chen's comments during the interview.

• Layer two trust assumptions: Chen notes that most layer twos "are not doing any sort of true settlement anyway" and many have single upgrade keys that make them effectively custodial despite marketing claims.

• Verifiable finance framing: Ellipsis Labs explicitly rejects the "decentralized" label for Atlas, instead using "verifiable finance" to describe their approach of centralized sequencing with cryptographic guarantees.

• Order flow monetization: Wallets, RPCs, and validators can all potentially sell user transaction flow to MEV shops, with crypto lacking the regulatory protections that exist in traditional finance payment for order flow arrangements.

• Lifinity precedent: The oracle-based proprietary AMM model that SolFi employs was pioneered by Lifinity in 2021, though the vocabulary to describe it didn't exist at the time.

• Phoenix development status: Ellipsis Labs has not done "too much active maintenance of the current product for a while," with resources shifting toward Atlas development.

• Multiple concurrent leaders: While Solana's MCL proposal addresses the proposer monopoly "to some degree," Chen argues it still doesn't provide the same control as a centralized sequencer.

Questions Answered

What is Atlas and why is Ellipsis Labs building it?

Atlas is a layer two blockchain being built by Ellipsis Labs specifically optimized for high-frequency trading and exchange infrastructure. The team decided to build Atlas after discovering that Solana's base layer limitations—particularly the "proposer monopoly" where a single validator controls transaction ordering for 1.6-second periods—prevented them from building the exchange quality they envisioned. With a centralized sequencer operated by Ellipsis Labs, Atlas can implement transaction ordering mechanisms that prioritize market maker welfare, ensure oracle updates land reliably, and enable truly continuous block building rather than discrete block times. Phoenix V2 will be the flagship application on Atlas.

What is SolFi and how does it work?

SolFi is an on-chain market maker on Solana developed by Ellipsis Labs, recently revealed as the largest such market maker on the network. It functions by exposing a swap interface that aggregators like Jupiter can query—asking how much USDC they'll receive for a given amount of SOL, for example. Unlike traditional AMMs that execute passively according to bonding curves, SolFi represents active market making strategy deployed on-chain, enabling more competitive pricing than Ellipsis Labs could achieve through their Phoenix order book positions. The proprietary AMM model allows them to play the current market structure game optimally while building Atlas to change that structure.

Why does Atlas settle on Ethereum instead of Solana?

Atlas settles on Ethereum because the properties required for settlement layers differ from those needed for execution layers. Settlement layers need strong censorship resistance—ensuring users can withdraw even if the sequencer goes down or becomes malicious—along with robust revert protection and fast finality for bridging. Chen argues that Ethereum has explicitly optimized for these properties while "basically all the other blockchains do not care about it." He also notes that the importance of the settlement layer is "a little bit overrated" since most layer twos aren't doing true settlement anyway, making the choice somewhat less consequential than it might appear.

Why doesn't Ellipsis Labs call Atlas decentralized?

Ellipsis Labs avoids calling Atlas decentralized because doing so would be dishonest—a centralized sequencer operated by the company will control transaction ordering. Instead, they use the term "verifiable finance" to describe the actual properties: cryptographic guarantees that state transitions are valid and that users can withdraw funds regardless of operator behavior. Chen expresses frustration with other teams marketing clearly centralized products as decentralized, noting that accurate definitions matter because "99% of the time it doesn't matter. And then sometimes the retail guy gets hurt." The layer two architecture provides verifiability without requiring decentralization.

What's wrong with current token launch infrastructure in crypto?

Chen argues that most existing token launch products "exist to fleece the end user as much as possible." Teams launching tokens must navigate centralized exchange listing teams who "technically don't take listing fees" but require participation in launch pads, pay market makers substantial sums, and sign numerous side deals with "really shady actors"—costs that can exceed 10% of total token supply. Additionally, token launches are designed to make "the chart look good" by starting prices artificially low, which creates sniping opportunities that benefit sophisticated actors at retail expense. Gavil attempts to address this through mechanisms like Dutch auctions that enable fair price discovery without extraction.

How does payment for order flow differ between crypto and traditional finance?

In traditional finance, payment for order flow is regulated such that retail users must receive prices "as good or maybe even strictly better than the prices on the lit books." Front-running is illegal, and robust monitoring ensures compliance. In crypto, no such protections exist. Wallets can sell signed transactions to MEV shops that add components "in front and behind to force the users to get the worst possible fill." This can happen at any point in the stack—wallet, RPC, or validator level. As Chen puts it