The intent thesis was compelling. Instead of telling the blockchain how to execute a transaction, you simply declare what you want. Sign a message. Let specialized actors compete to deliver the best outcome. MEV protection, better prices, gasless trades. The future of DeFi, distilled into a single paradigm shift.
Two and a half years after Paradigm formalized the concept, the verdict is in: intents solved the wrong problem.
Paradigm's June 2023 research on intent-centric architectures gave a name to something already brewing. CoW Protocol had pioneered the batch auction and solver model since 2020. 1inch Fusion launched in late 2022. UniswapX followed in mid-2023. The thesis was consistent across all of them: abstract away execution complexity, let professional "solvers" handle the routing, and users get better outcomes without thinking about gas, slippage, or MEV.
The pitch made sense. DeFi had become a maze of fragmented liquidity, siloed protocols, and predatory bots extracting value from every transaction. Intents promised to fix this by shifting the burden from users to a competitive marketplace of sophisticated executors.
And for simple token exchanges, it worked. Sort of.
Intent systems delivered incremental improvements to a narrow use case: single-asset exchanges.
CoW Protocol batches orders and finds "coincidences of wants" to reduce MEV exposure. UniswapX runs Dutch auctions where fillers compete for the right to execute your trade. 1inch Fusion lets resolvers wait until pricing hits their profitability threshold before filling.
These are engineering achievements. They make basic token exchanges marginally safer and sometimes cheaper. But they fundamentally misunderstand where DeFi's complexity problem actually lives.
The real complexity isn't in routing a single exchange. It's in executing multi-step strategies that span multiple protocols.
The intent thesis was compelling. Instead of telling the blockchain how to execute a transaction, you simply declare what you want. Sign a message. Let specialized actors compete to deliver the best outcome. MEV protection, better prices, gasless trades. The future of DeFi, distilled into a single paradigm shift.
Two and a half years after Paradigm formalized the concept, the verdict is in: intents solved the wrong problem.
Paradigm's June 2023 research on intent-centric architectures gave a name to something already brewing. CoW Protocol had pioneered the batch auction and solver model since 2020. 1inch Fusion launched in late 2022. UniswapX followed in mid-2023. The thesis was consistent across all of them: abstract away execution complexity, let professional "solvers" handle the routing, and users get better outcomes without thinking about gas, slippage, or MEV.
The pitch made sense. DeFi had become a maze of fragmented liquidity, siloed protocols, and predatory bots extracting value from every transaction. Intents promised to fix this by shifting the burden from users to a competitive marketplace of sophisticated executors.
And for simple token exchanges, it worked. Sort of.
Intent systems delivered incremental improvements to a narrow use case: single-asset exchanges.
CoW Protocol batches orders and finds "coincidences of wants" to reduce MEV exposure. UniswapX runs Dutch auctions where fillers compete for the right to execute your trade. 1inch Fusion lets resolvers wait until pricing hits their profitability threshold before filling.
These are engineering achievements. They make basic token exchanges marginally safer and sometimes cheaper. But they fundamentally misunderstand where DeFi's complexity problem actually lives.
The real complexity isn't in routing a single exchange. It's in executing multi-step strategies that span multiple protocols.
Consider what a sophisticated trader actually needs to do:
Open a delta-neutral position by going long on one protocol while shorting on another. Rebalance a portfolio across lending markets and liquidity pools. Execute a basis trade that involves spot purchases, perp positions, and collateral management. Move assets across chains while simultaneously entering positions on the destination network.
Try expressing any of these as an "intent" on CoW Protocol or UniswapX. You can't. These systems were built for token exchanges. Everything else still requires what we call the Transactional Loop: manually stitching together dozens of individual transactions across different protocols, watching prices drift between each step, and hoping nothing fails halfway through.
The limitations run deeper than feature gaps. The architecture itself creates constraints that prevent intent systems from scaling beyond simple use cases.
Auction Latency
Dutch auctions are elegant in theory. Prices start favorable to users and decay over time until solvers find them profitable. In practice, this means waiting. Solvers monitor orders until pricing hits their individual profitability thresholds. Your "instant" trade becomes a game of patience, and if market conditions shift during the auction window, you absorb that risk.
Real-time execution requires real-time routing. Auction mechanisms are fundamentally incompatible with strategies that depend on capturing specific market conditions before they change.
Solver Economics
Solvers are rational actors. They fill orders when profitable and ignore them when not. This creates a structural bias toward high-volume, low-complexity trades where the spread between user price and market price is easy to capture. Complex multi-leg strategies with thin margins? Not worth the solver's time.
Consider what a sophisticated trader actually needs to do:
Open a delta-neutral position by going long on one protocol while shorting on another. Rebalance a portfolio across lending markets and liquidity pools. Execute a basis trade that involves spot purchases, perp positions, and collateral management. Move assets across chains while simultaneously entering positions on the destination network.
Try expressing any of these as an "intent" on CoW Protocol or UniswapX. You can't. These systems were built for token exchanges. Everything else still requires what we call the Transactional Loop: manually stitching together dozens of individual transactions across different protocols, watching prices drift between each step, and hoping nothing fails halfway through.
The limitations run deeper than feature gaps. The architecture itself creates constraints that prevent intent systems from scaling beyond simple use cases.
Auction Latency
Dutch auctions are elegant in theory. Prices start favorable to users and decay over time until solvers find them profitable. In practice, this means waiting. Solvers monitor orders until pricing hits their individual profitability thresholds. Your "instant" trade becomes a game of patience, and if market conditions shift during the auction window, you absorb that risk.
Real-time execution requires real-time routing. Auction mechanisms are fundamentally incompatible with strategies that depend on capturing specific market conditions before they change.
Solver Economics
Solvers are rational actors. They fill orders when profitable and ignore them when not. This creates a structural bias toward high-volume, low-complexity trades where the spread between user price and market price is easy to capture. Complex multi-leg strategies with thin margins? Not worth the solver's time.
The intent model assumes solver competition will drive better outcomes. But competition only works when the economics align. For anything beyond straightforward token exchanges, they don't.
Fragmentation
Each intent protocol built its own format, its own solver network, its own auction mechanism. UniswapX intents can't interact with CoW Protocol. 1inch Fusion resolvers operate in their own silo. This fragmentation means solvers can't compose across systems, and users can't express intents that span multiple intent networks.
The irony is thick: a paradigm designed to abstract away complexity created an entirely new layer of it.
Intent systems answered a question nobody was urgently asking: "How do I exchange tokens with slightly better MEV protection?"
The question DeFi actually needed answered: "How do I execute complex strategies without spending hours clicking through protocols and watching my edge evaporate between transactions?"
This is the Transactional Loop. Users are stuck manually sequencing operations that should be unified. Every additional step introduces execution risk, market drift, and cognitive overhead. The protocols exist. The liquidity is there. But the execution layer forces users to become their own back-office, stitching primitives together one transaction at a time.
Intents didn't break this loop. They made one small corner of it slightly less painful.
The declarative philosophy at the heart of intents was correct. Users should define outcomes, not execution steps. The implementation was the problem.
What if instead of submitting orders to an auction and waiting for solvers to compete, you could define a target state and have it executed in real-time? Not "exchange A for B and hope a solver picks it up." Instead: "I want this portfolio configuration, across these protocols, on these chains. Execute."
This is what we call Declarative Transactions. Same declarative philosophy. Radically different execution model.
The intent model assumes solver competition will drive better outcomes. But competition only works when the economics align. For anything beyond straightforward token exchanges, they don't.
Fragmentation
Each intent protocol built its own format, its own solver network, its own auction mechanism. UniswapX intents can't interact with CoW Protocol. 1inch Fusion resolvers operate in their own silo. This fragmentation means solvers can't compose across systems, and users can't express intents that span multiple intent networks.
The irony is thick: a paradigm designed to abstract away complexity created an entirely new layer of it.
Intent systems answered a question nobody was urgently asking: "How do I exchange tokens with slightly better MEV protection?"
The question DeFi actually needed answered: "How do I execute complex strategies without spending hours clicking through protocols and watching my edge evaporate between transactions?"
This is the Transactional Loop. Users are stuck manually sequencing operations that should be unified. Every additional step introduces execution risk, market drift, and cognitive overhead. The protocols exist. The liquidity is there. But the execution layer forces users to become their own back-office, stitching primitives together one transaction at a time.
Intents didn't break this loop. They made one small corner of it slightly less painful.
The declarative philosophy at the heart of intents was correct. Users should define outcomes, not execution steps. The implementation was the problem.
What if instead of submitting orders to an auction and waiting for solvers to compete, you could define a target state and have it executed in real-time? Not "exchange A for B and hope a solver picks it up." Instead: "I want this portfolio configuration, across these protocols, on these chains. Execute."
This is what we call Declarative Transactions. Same declarative philosophy. Radically different execution model.
No auctions. No solver competition. No waiting. Dynamic routing calculates the execution path in real-time, and the entire strategy executes as a single, all-or-nothing transaction. Either your target state is achieved, or nothing happens. No partial fills, no stuck positions, no manual cleanup.
The difference is architectural. Intent systems outsource execution to third parties who may or may not find your order profitable. Declarative Transactions bring execution in-house, routing dynamically across 45+ protocols and 21 networks to find paths that intent solvers can't or won't pursue.
When execution isn't constrained by solver economics or auction latency, strategies that were impossible become routine.
A trader wants to open a delta-neutral position: long ETH on Aave, short ETH-perp on another protocol, funded by a USDC bridge from a third chain. In the Transactional Loop, this is an afternoon of clicking, monitoring, and praying nothing moves against you. With Declarative Transactions, it's one transaction. Define the target state. Execute.
A fund needs to rebalance across six protocols without moving markets. Sequential trading leaks information and creates slippage on every leg. Single-transaction execution collapses the entire rebalance into one operation, faster than arbitrageurs can react.
A protocol wants to offer sophisticated strategies to their users without building custom execution infrastructure. One API integration unlocks every strategy their users could want, without months of development.
This is where DeFi needs to go. Not marginally better token exchanges. Execution infrastructure that handles the strategies users actually need.
Intents were a step in the right direction. The insight that users should declare outcomes rather than manage execution was correct. But the implementation, built on auctions and solver marketplaces, couldn't scale beyond the simplest use cases.
The next generation of DeFi infrastructure needs to deliver on the original promise: any outcome, executed. Not just token exchanges. Not just simple routes. Complex, multi-protocol strategies that are impossible to execute manually, collapsed into single transactions.
Define the outcome. Execute the path.
That's what outcome execution means.
No auctions. No solver competition. No waiting. Dynamic routing calculates the execution path in real-time, and the entire strategy executes as a single, all-or-nothing transaction. Either your target state is achieved, or nothing happens. No partial fills, no stuck positions, no manual cleanup.
The difference is architectural. Intent systems outsource execution to third parties who may or may not find your order profitable. Declarative Transactions bring execution in-house, routing dynamically across 45+ protocols and 21 networks to find paths that intent solvers can't or won't pursue.
When execution isn't constrained by solver economics or auction latency, strategies that were impossible become routine.
A trader wants to open a delta-neutral position: long ETH on Aave, short ETH-perp on another protocol, funded by a USDC bridge from a third chain. In the Transactional Loop, this is an afternoon of clicking, monitoring, and praying nothing moves against you. With Declarative Transactions, it's one transaction. Define the target state. Execute.
A fund needs to rebalance across six protocols without moving markets. Sequential trading leaks information and creates slippage on every leg. Single-transaction execution collapses the entire rebalance into one operation, faster than arbitrageurs can react.
A protocol wants to offer sophisticated strategies to their users without building custom execution infrastructure. One API integration unlocks every strategy their users could want, without months of development.
This is where DeFi needs to go. Not marginally better token exchanges. Execution infrastructure that handles the strategies users actually need.
Intents were a step in the right direction. The insight that users should declare outcomes rather than manage execution was correct. But the implementation, built on auctions and solver marketplaces, couldn't scale beyond the simplest use cases.
The next generation of DeFi infrastructure needs to deliver on the original promise: any outcome, executed. Not just token exchanges. Not just simple routes. Complex, multi-protocol strategies that are impossible to execute manually, collapsed into single transactions.
Define the outcome. Execute the path.
That's what outcome execution means.
Haiku is the outcome execution layer for DeFi. Learn more at haiku.trade.
Haiku is the outcome execution layer for DeFi. Learn more at haiku.trade.
Subscribe to Haiku
Subscribe to Haiku
Share Dialog
Share Dialog
>1.2K subscribers
>1.2K subscribers
No comments yet