This episode features Eugene, CEO of Ellipsis Labs (known as Xerox ShitTrader on Twitter), discussing the evolution of market structure on Solana and the development of Phoenix perps.

Eugene and his co-founder Jerry started Ellipsis Labs three and a half years ago, building products exclusively on Solana including the Phoenix Spot order book and SolFi prop AMM.

The conversation covers Solana's infrastructure challenges, the paused Atlas L2 project, and how MCP (Multi-Client Protocol) represents the solution to current sequencing problems.

Eugene explains why Phoenix perps is launching natively on Solana rather than using sidechains, and discusses the broader implications of validator incentives and market structure evolution.

Solana's Market Structure Evolution and Infrastructure Improvements

When Phoenix Spot launched, Solana experienced a 19-hour chain outage during Jupiter integration testing in December 2022, highlighting early infrastructure risks

Spreads on major pairs like SOL/USDC improved dramatically: from 30 basis points initially, to 15 basis points after Phoenix, down to 5 basis points after SolFi

"The spreads on chain are actually very competitive with the spreads on centralized exchanges, and it's a very counterintuitive result" - Eugene

Market structure evolution happened because Jupiter was willing to integrate different liquidity primitives, enabling competition that doesn't exist on centralized venues

The Sequencing Problem and MCP as the Solution

Solana's sequencing layer operates on a scale from 0 (maximally good for users) to 10 (maximally extractive), currently trending from 4-5 toward 10

"MCP is the light at the end of tunnel for the sequencing layer on Solana" - Eugene, though it introduces new risks and research problems

Atlas L2 was paused indefinitely because MCP offers a path to fix sequencing while maintaining full on-chain benefits like easy integrations and composability

Current sequencing issues don't impact Phoenix until it reaches sufficient scale to attract sophisticated toxic traders or validator mods targeting the protocol

Phoenix Perps Strategy and Market Structure Design

Phoenix perps combines application, router, and liquidity layers, enabling deterministic on-chain routing that splits trades between order books and market maker splines

"We do not have a points program. We have no plan to have a points program. We also are not going to have a token and we're not going to have an airdrop" - Eugene

Perps aggregators are unlikely to succeed because different venues create non-fungible instruments that aren't convertible between platforms

Phoenix currently has a couple thousand users in private beta with referral codes available through their Discord for motivated users

Validator Incentives and Block Building Competition

Validators face incentives to be maximally extractive in the short term, offering better rewards by MEVing users and applications harder

"It's just people responding to the incentives of the situation in front of them" - Eugene, noting no moral judgment on validator behavior

Some teams are fighting MCP because it threatens their business models, but the costly signal is sacrificing short-term gains to grow the overall pie

Teams with deep protocol expertise like Jito and Harmonic will find valuable opportunities in an MCP world despite current business model disruption

DeFi vs TradFi Evolution and Regulatory Considerations

DeFi changes rapidly where "entrenched winners usually are not actually as strong as they seem," unlike TradFi where the same players dominate for years

"The barriers to entry have really never been lower, especially now that you can vibe code your way to a decent starting point" - Eugene

Mass crypto adoption could occur with crypto as an asset ledger while most coins go to zero and little actual DeFi activity remains

Regulatory intervention might be appropriate if markets get completely captured, but users can currently migrate to other chains if one becomes too extractive

Resources Mentioned

AI Agents with MCP Model Context Protocol for Building Clients, Services, and End-to-End Agents

the end of tunnel for the sequencing layer on Solana" - Eugene, though it introduces new risks and research problems Atlas L2 was paused indefinitely because MCP offers a path to fix sequencing whil

Sorry, It's Your Problem Now Because I'm Dead, Interesting Notebooks Workbooks Simple Business Office Books, Last Wishes and Emergency Binders for End-of-life Arrangements (Black)

you know, again, tons of risks involved with putting out MCP, a lot of unknown unknowns, plenty of research problems that are not totally solved yet. But there's one really, really big problem on mar