Jul 1, 2025

Greener Pastures: For CoW Swap, A Better DEX Starts with Intent

Head of BD Chen Magen on how batch auctions, solver ecosystems, and sustainable practices make for better harvests in DeFi

If 2024 seemed to be all about user interface and abstraction – “surface” issues, in a way, albeit super critical for adoption – then this year, we’ve gone back under the hood. DeFi infrastructure is evolving from the inside out, thanks to exchanges and protocols with first-hand knowledge of the needs and pain points of an ultra-sophisticated user base that happens to be at the forefront of onchain capital growth.

Fluid (FKA Instadapp) got a head start in November 2024, making a splash with their hybrid model melding previously siloed lending and trading functionalities; in early 2025, Uniswap v4 raised the bar for DEX modularity and customization.

Founded in 2021, CoW Swap has mostly eschewed hyped up new versions, favoring relatively quiet updates that target the structural problems hindering growth across the board.

For CoW Swap’s business development lead, Chen Magen (that’s “ch” as in “Bach,” not “cheddar”), that means taking aim at MEV exploitation, fragmented liquidity, and the false simplicity of yield.

The meta-DEX’s solutions — batch auctions, solver competition, and an intent-based trading model — began as an answer to poor execution and unfair pricing. Now they’re used by some of DeFi’s most respected institutions. Curve uses CoW Swap in DAO operations. Aave named it their preferred swap routing solution. Lido taps it for treasury management. Sky uses it to power on-chain swaps.

These aren’t one-off integrations — they’re signs of deep protocol trust. Chen tells Coinshift just how it was earned.

CoW Swap's Chen Magen

CS: We like to start at the beginning: how did you come to Web3 and DeFi?

CM: My background is in mechanical engineering. Around the time that I was doing my master’s in that field, a close relative of mine got into Bitcoin — it was very early then — and talked about it often. Like many others, I didn’t listen early enough or translate it into actually buying Bitcoin. But that’s how I started learning the concepts, especially the idea of a global, neutral, and independent form of money. That really fascinated me.

I continued working in mechanical engineering in the medical device industry, manufacturing infusion pumps. Eventually I came across an opportunity with a company that had deep expertise in options trading and had pivoted into crypto. That was my first full-time role in the crypto space, though not yet in DeFi.


CS: What year was that?

CM: Around 2018, a year or so after the ICO boom. We provided services like market-making to various projects. This was during the early days of Ethereum, when infrastructure was minimal. Wallets were basic, and there were virtually no DEXs.

I started working with Gnosis, who we helped market-make one of the earliest Ethereum DEXs using auction mechanisms, which was core to what they were working on. I eventually joined Gnosis around 2020, right as COVID began. I didn’t meet my co-workers in person for more than a year. At the time, we were one team within Gnosis, building decentralized exchange infrastructure for prediction markets. I was specifically tasked with onboarding market makers to the new DEX. It was quite a challenge. 


CS: How so?

CM: The first design for the DEX was based on Dutch auctions: six-hour-long, falling price auctions. The UX wasn’t good enough to onboard users, though interestingly, the concept is still relevant today and is used by other DeFi projects. After that, we moved to batch auctions under what we called Gnosis Protocol. It required users to deposit funds before trading — more like a centralized exchange, but non-custodial.A key challenge was bootstrapping liquidity. Without traders, liquidity providers won’t come; without LPs, traders won’t either. The turning point came when MEV (Miner Extractable Value) became a real-world problem. 


CS: What changed? CM:One example was UMA launching a token on Uniswap. Bots exploited price uncertainty and front-ran retail users.

CM: Martin Köppelmann, Gnosis’ co-founder, wrote about this, and we began attracting projects wanting to launch tokens without falling victim to that MEV vulnerability. Gnosis Protocol’s batch auctions fit this need well, and the protocol ended up hosting token launches (IDOs) that raised millions. But our vision wasn’t fundraising — we wanted a trading protocol.

That led to CoW Swap. 

CS: How does CoW Swap evolve the model?

CM: We kept the batch auction mechanism but made key changes: running the batch auction off-chain while settling on-chain, and accessing all Ethereum liquidity for settlement.

This made it easy for people to use and for the protocol to bootstrap liquidity. We have used our original smart contracts ever since – that’s four years, almost 2 million transactions, and counting.


CS: No way.

CM: Of course it has evolved over time. We now call it CoW Protocol, and improvements are continuously being made. But the fundamental protocol design and smart contract are still in use today.


CS: What issues have come up that made those updates necessary?

CM: One major challenge we’ve seen is composability across DeFi protocols, particularly when factoring in cross-chain or L2 fragmentation.

A pivotal point came with the introduction of intent-based trading. This shift represented a fundamental change from traditional DEX aggregator models. A user defines their intent — for example, “swap token A for token B at this price” — but without specifying a route. It’s a generic but powerful idea.

This model broke expectations built around traditional DEX aggregators. Previously, you’d get a quote and full calldata for a transaction you’d execute on-chain, which made it easy to compose DeFi actions, like swapping and lending in one step. Intent-based trading made that harder, so we’ve had to build tooling to improve composability and help partners integrate.


CS: Can you give us an example of that pain point in action?

CM: Think of it like booking a complex trip. In the old model, you'd get a detailed itinerary upfront — every flight, connection, and timing — which made it easy to book a hotel at your destination in the same transaction. With intent-based design, you just say "get me to Paris by Friday for under $500," but the actual route isn't determined until later, making it harder to coordinate that hotel booking.

In DeFi terms, imagine a user holds USDT and finds attractive yield in DAI or USDS on a lending platform. With the older model, they could construct a single transaction that both swaps the asset and deposits it into the lending pool — a seamless, one-click flow. That is, with the disadvantage of users being required to handle the complexity of onchain transaction execution, gas costs, and transaction ordering.

With intent-based design, the route and execution details aren’t defined at the time the user expresses their intent. That means it’s harder to bundle actions across protocols or assets. This is further compounded when assets and protocols are spread across multiple chains or L2s, each with different liquidity, tooling, and wallet constraints.

We’ve been working on tooling to bridge that gap. CoW Protocol’s Hooks allows users and developers to provide additional execution instructions that will get executed once the route has been defined. Interoperability and composability — particularly in an L2-fragmented environment — remain a real challenge.

“Liquidity providers were leaking value every time they traded at stale prices. CoW AMM was designed to stop that.”

CS: Is intent-based architecture just where DeFi is headed?

CM: In some areas, yes. Bridges have already moved to intent-based models. Some DEXs, like Uniswap X and 1inch Fusion have also adopted it, though they still support legacy methods. They’ve seen the potential but haven’t fully supported the tooling needed.


CS: Let’s talk about milestones to date. Where has CoW Swap been getting traction in your view?

CM: First, we started to see volumes growing, CoW Protocol had recently surpassed $100B in cumulative trading volume. Another big milestone was our fee model. During the auction process, solvers compete to give users the best price. If the user gets a better price than expected, we call that “surplus.” And we only charge fees when that surplus exists. This aligned our incentives with the user and made the protocol cashflow positive.

But I’m most proud of the traction we’ve achieved among DeFi protocols. In my view, they represent the most advanced and demanding DeFi users and developers.


CS: What recent developments have strengthened that position with protocols?

CM: One major advancement we've made is launching combinatorial auctions — one of our biggest protocol upgrades since our inception. The change fundamentally improves how we handle competing solutions for user orders.

Previously, solvers would submit one solution attempting to settle an entire batch optimally. Now, they must submit multiple bids: individual solutions for each order, plus additional combined solutions that group orders together. Each user order gets a reference price from the best individual bid, and any batched solution can only win if it improves upon that reference.

This creates stronger execution guarantees for users and opens up opportunities for smaller, specialized solvers to compete. Instead of needing to optimize for entire batches, they can focus on individual orders or specific combinations where they have advantages.


CS: How does this change the competitive dynamics within the solver ecosystem?

CM: It democratizes solver participation. Previously, large, well-capitalized solvers would have an advantage for winning large batches. Now, a solver specializing in, say, stablecoin swaps can win those specific orders — even if they can't optimize the entire batch. This increased competition ultimately benefits users through better execution.

The upgrade also improves throughput. By allowing partial batch settlements and more granular optimization, we can process more orders efficiently while maintaining our MEV protection guarantees.


CS: With so much attention on DeFi UX and capital efficiency, what do you see as the direction DEXs are heading in?

CM: One area we’ve focused on is helping users and liquidity providers generate yield more safely and efficiently. Until CoW AMM, CoW Protocol didn’t cover the liquidity provision side — we were primarily focused on aggregation and trade execution. But we saw a growing need to support LPs wanting to deploy capital while avoiding common risks. CoW AMM is designed to address a major issue: LVR — loss versus rebalancing. The problem is that when you provide liquidity in most AMMs, you're often trading at stale prices because blockchains only update data in discrete time intervals. That’s every 12 seconds on Ethereum. That delay opens the door for arbitrageurs to extract value, and LPs end up losing out.

What made this especially relevant for us is that CoW Protocol already had an edge, using batch auctions and a competitive solver environment to constantly find the best execution. So we took those mechanics and applied them to a new kind of AMM — one that ensures liquidity is traded at prices that better reflect current market conditions. That way, LPs don’t leak value just because of timing.

As the first AMM design to address LVR, we’ve taken a significant step toward making yield-generation in DeFi more sustainable. But there’s still plenty to be desired. Together with Balancer, we’re working on a new version to address that.

CS: When we spoke to Fluid Protocol’s founders in early 2025, they said it’s going to be the “year of the DEX wars.” Between what they’re building, CoW AMM, and even Uniswap’s newer models, it seems like there is indeed some productive competition around these liquidity layers.

CM: Definitely. There's a lot of energy in the space, both in AMM design and in how protocols handle routing, fees, and MEV protection. What sets CoW Protocol apart is our focus on solving the fundamental problem of MEV extraction, rather than just optimizing existing AMM designs. While others are innovating on liquidity mechanisms, we're addressing the core issue of fair price discovery and execution.

But, alongside that innovation, the challenge of fragmentation across chains and L2s remains unresolved. It’s really hard to deliver seamless experiences, especially for things like batch auctions or composable trading strategies, when liquidity and tooling are scattered.

Solutions might happen at the chain level, say with Ethereum introducing new interoperability standards or native communication layers. That removes some of the burden from protocols like ours. There's been more discussion around this topic recently in the Ethereum community, and solutions like based rollups are being proposed. But for now, cross-chain liquidity unification remains one of the hardest engineering problems in the space — and it’s not clear yet what the best path forward is.


CS: Are you optimistic?

CM: In general?

CS: Yes.

CM: I am.


CS: And about batch auctions?

CM: Settling many orders across chains in one batch remains complex, and the solutions might not always be perfect. But I know we’ll continue to find them.

Never miss an update with The Value Add newsletter