Skip to content

Intent-Based Execution

An intent can simplify what a user expresses while moving complexity into authorization, competition, resource control, execution, and settlement boundaries.

The question

What must an intent system prove between a user's desired outcome and an observed settlement?

Intent interfaces can reduce the choices exposed to a user, but the system still needs an executable mandate, controlled resources, a method for discovering or selecting execution, and evidence that the outcome satisfied the mandate.

A boundary model

A useful review follows six boundaries:

  1. Expression — what outcome, limits, expiry, and permitted flexibility did the user authorize?
  2. Resource control — which assets or permissions are available, for how long, and to whom?
  3. Discovery — how can solvers or fillers propose an execution without gaining excess authority?
  4. Validation — which oracle, simulation, or onchain conditions determine acceptability?
  5. Execution — what exact payload can move resources, and what prevents replay or substitution?
  6. Settlement — what evidence establishes the user outcome, fees, residual value, and failure state?

The label “intent” does not answer these questions. It changes where they are implemented.

Prototype findings

The UHI5 and Unite DeFi experiments expose complementary parts of the model:

  • a delayed Uniswap v4 hook tests how execution can be deferred while another actor seeks a route;
  • limit-order interactions test how a signed order can permit bounded pre- and post-execution behavior;
  • a Tycho-backed executor tests one path from solver-selected route material to multi-DEX execution;
  • oracle-conditioned orders test the difference between price evidence and execution authority; and
  • ERC-6909-oriented resource locking tests how reusable permissions can be separated from immediate custody transfer.

The public prototype records also show the limits of early integration work: intent creation, listener, solver, partner, cross-chain, and production paths can remain planned even when one hook or contract interaction is executable.

Engineering conclusions

An intent system should:

  • make the mandate and permitted solver discretion explicit;
  • bind resource authority to an identity, scope, expiry, and replay rule;
  • distinguish quote, proposal, selection, simulation, submission, and settlement;
  • validate solver output without granting arbitrary-call authority;
  • define who receives residual value and who bears execution costs;
  • preserve a recoverable state when execution is delayed, rejected, reverted, or ambiguous; and
  • record which claims are prototype observations versus protocol goals.

Atomic execution can reduce principal exposure for a bounded flow, but it does not remove infrastructure, RPC, operational, or gas expenditure. Failed transactions may still incur unreimbursed costs.

Alternatives and counterevidence

Direct user-authored transactions can be simpler when the execution path is known and user choice is acceptable. Traditional order books or request-for-quote systems can make price discovery and counterparty selection more explicit. A general solver network adds value only when its search and competition benefits justify additional authority, availability, privacy, and settlement complexity.

Resource locking can reduce repeated custody movement, but it expands the importance of revocation, expiry, replay, and downstream permission review.

Limitations and freshness

  • The implementation evidence is historical prototype evidence, not a production solver network or audited protocol.
  • The analysis does not assert current behavior for Uniswap, 1inch, Tycho, The Compact, ERC-7683, or other third-party systems beyond the cited prototype context.
  • Seamless material is excluded from the active evidence set because contribution and disclosure boundaries remain unresolved.
  • Private solver strategy, partner plans, opportunity details, and operational configuration are excluded.
  • Revisit the analysis when a prototype supplies a complete intent-to-settlement trace and independently reviewed security boundary.

Go deeper