Why Intent-Based Execution Needs a Scoring Model

"Swap 1,000 USDC for ETH" is an intent, not an execution plan. Turning it into a plan means choosing among candidate routes — different solvers, different liquidity paths, different tradeoffs — and that choice is where most of the interesting protocol design work actually lives.

Candidates aren't ranked by one number

A naive implementation picks whichever quote returns the best output amount and calls it done. In practice, at least three independent dimensions matter, and they don't collapse into each other:

• Cost — the actual amount received net of fees and slippage.
• MEV exposure — how vulnerable the chosen path is to being sandwiched, reordered, or front-run before it settles.
• Latency — how long the user waits for the route to actually confirm.

A route that wins on cost but loses badly on MEV exposure isn't obviously better than one that's slightly more expensive but far harder to attack.

Why this has to be a policy, not a hardcoded rule

Different callers legitimately want different tradeoffs — a market maker optimizing for latency looks nothing like a retail swap optimizing for worst-case cost. Encoding "the" correct weighting as a hardcoded if-statement bakes in someone's specific risk tolerance as if it were a protocol invariant. Making the weights an explicit, adjustable policy instead means the tradeoff is visible and owned, not accidental.

Determinism matters more than sophistication here

Given the same candidates and the same policy, the selection has to be reproducible — the same inputs should always score the same way. A scoring model that can't be re-run and re-verified after the fact isn't trustworthy enough to explain why a particular route was chosen, which matters as much as the choice itself once MEV is one of the inputs being weighed.

---

Draft — this is seed content for the Research section's initial layout, not a finished piece. Expect this to be rewritten.