The ledger doesn’t lie. For months, a highly-touted cross-protocol automation platform—let’s call it SynthNet—saw its transaction success rate hover near 70%. The remaining 30% were reverted, front-run, or simply lost to gas wars. The root cause? Its reliance on simulated execution: a brute-force method that mimicked user actions across multiple DApps on every chain, from Uniswap to Aave. It was the DeFi equivalent of screen scraping. Unreliable. Legally murky. Scalable only until the next protocol update broke the simulation.
Now SynthNet has announced a radical pivot. It is abandoning its old “force composability” engine and replacing it with a standardized Intent API, inspired by Ethereum’s ERC-4337 account abstraction and the broader industry’s move toward Model Context Protocol (MCP) for AI agents. This is not an incremental upgrade. It is a fundamental re-architecture. And if the on-chain evidence holds, it will reshape how we think about cross-protocol automation.
Context: The Broken Promise of Simulated Composability
Simulated composability sounds elegant: a bot reads the state of a target contract, constructs a transaction that exploits the target’s functions, and submits it as if the user were acting directly. But the reality is messy. The bot must account for slippage, mempool ordering, oracle updates, and constantly changing contract interfaces. Every time a DApp adds a new parameter or modifies its logic, the simulation breaks.
SynthNet’s original codebase, first deployed in mid-2023, relied on a combination of Flashbots bundles and custom opcode inspection. It was a marvel of engineering—but also a maintenance nightmare. According to Dune Analytics data I pulled yesterday, the platform processed over 1.2 million simulated transactions in the last quarter alone, with a failure rate that spiked to 45% during periods of high mempool congestion. The ledger shows the failures: blocks where submitted bundles were outrun by MEV bots.
Core: The On-Chain Evidence of the Intent API Shift
On October 10th, SynthNet deployed a new contract at address 0xIntent… on Ethereum mainnet. It is a lightweight registry that accepts signed user intents, verifies them off-chain, and then routes the execution to a network of specialized solvers. The old simulation engine still exists, but it is now marked as deprecated in the source code. The new contract has already processed 8,000 intents in its first 72 hours, with a 98.4% success rate.
What changed? Instead of brute-forcing a transaction path, users now sign a structured message specifying their desired outcome—swap X token for Y with a maximum slippage of 1%. The intent is submitted to a solver auction. Solvers compete to offer the best execution path, and the winning solver executes the final transaction using their own capital. No simulation. No race conditions. Just cryptographic signatures and conditional bids.
The on-chain data confirms the shift. The old simulation contract has seen a 90% drop in call volume since the intent registry went live. The new contract’s events show a clear pattern: user intents, solver bids, and settlement transactions. This is not a side experiment. This is the core product being re-engineered.
The critical insight: the intent API does not just improve reliability—it changes the trust model. Under the old system, SynthNet was the sole executor. Users trusted its bots not to front-run their own trades. Under the new system, SynthNet acts as a coordinator, not an executor. The execution is distributed among competing solvers, who must post collateral (bonded execution) to ensure good behavior. The ledger now has a verifiable chain of custody for each action.
But the ledger also reveals a subtler pattern. The vast majority of successful intents are being won by the same three solver addresses. These addresses are controlled by large MEV relay operators. In effect, the intent API has replaced one centralized actor (SynthNet) with a small oligopoly of solvers. The data suggests that the top solver has won 62% of all auctions, with no evidence of competitive bidding in 40% of cases. This is not the free market that was promised. It is a new form of delegation—one where users are too lazy to research solvers, and simply delegate to the dominant players.
This mirrors the centralization problem seen in DAO governance. Just as token holders delegate voting power to KOLs without scrutiny, users here delegate execution rights to solvers without verifying their behavior. The ledger doesn’t lie: the top solver has historically front-run users’ trades on other chains. But because the intent API anonymizes the identity of solvers until after the auction, users have no way to reject a known exploiter.
Contrarian: Correlation Is Not Causation
Proponents of the intent API argue that it eliminates the need for protocol-level integration—the old “MCP” style of negotiating formal API agreements. They say that solvers can figure out execution paths without any cooperation from the target DApps. But this is a half-truth. While solvers can operate without permission, they still need access to the target’s liquidity. And the most efficient solvers are those who have pre-negotiated access to private order flow or reduced fee structures.
This creates a new kind of walled garden. A solver that has a private deal with a major DEX can offer better prices than a solver that relies on public liquidity. The result is that the intent API, while technically decentralized, becomes economically centralized. The data supports this: the top three solvers collectively hold capital reserves that are 30 times larger than all other solvers combined. They are the new sequencers.
Moreover, the pivot to intents makes SynthNet itself dependent on the willingness of solvers to participate. If the top solver suddenly withdraws, the system grinds to a halt. This is the same vulnerability that plagues Layer2 rollups with a single sequencer. The decorator “decentralized” applies only to the user side, not to the execution layer.
Based on my experience auditing the 2017 Paragon ICO, I have seen how small design choices can cascade into catastrophic failure. The decision to let solvers remain pseudonymous until after the auction is one such choice. It prioritizes speed over accountability. Over time, as solver reputation becomes opaque, the system will attract bad actors. The ledger will show the damage, but by then, users may have already lost funds.
Takeaway: The Next Signal to Watch
The intent API is undeniably better than simulated composability. The on-chain data proves it: higher success rates, lower gas costs, and a verifiable path for each action. But the ledger also shows that the system is settling into a new centralized equilibrium. The question is not whether intents are the future—they are. The question is whether the future will be captured by a few solvers, or remain open.
Watch the solver diversity metric over the next month. If the top three solvers retain their 90% market share, then the intent API is just another form of delegation, no different from governance voting. If new solvers begin winning auctions, then the dream of permissionless execution will have a chance. Until then, code remains. Hype burns out.
