The Problem No Single Transaction Can Solve
Imagine signing one transaction in your wallet that moves native BTC on the Bitcoin network and arrives as native ETH on Ethereum within seconds. No wrapped tokens. No waiting for a 7-day fraud window. Just funds on the destination, almost instantly.
That experience is real today — but achieving it quietly involves solving one of the hardest problems in distributed systems. A single blockchain transaction is atomic: it either fully executes or fully reverts inside one block. There is no mechanism that guarantees a burn on chain A and a mint on chain B simultaneously. The two chains run independent finality clocks, and no transaction can straddle both.
Traditional lock-and-mint bridges work around this by locking native tokens in a custodied vault on the source chain and issuing synthetic wrapped representations on the destination. That vault is a permanently high-value target — and bridge infrastructure remains the most-exploited component in crypto, with cumulative losses exceeding $2.8 billion since 2022 and major incidents still occurring in 2026.
Intent-based cross-chain protocols take the opposite approach. Rather than maintaining a custodied float, a relayer fronts destination-chain liquidity out of its own inventory the moment your deposit is detected, then claims reimbursement from a source-chain escrow after a proof or oracle confirms the original deposit. The user receives near-instant finality; the relayer carries the cross-chain timing risk. This is the same declarative "outcome over instruction" model introduced in Part 1 of this series and applied to single-chain swaps by UniswapX and CoW Swap in Part 2.
Three protocols have emerged as leading implementations of this model, each choosing a different trust stack to manage the timing gap between fill and reimbursement: Across (optimistic oracle), Chainflip (threshold-signature validator vaults), and Relay (dedicated settlement chain). ERC-7683, co-authored by Uniswap Labs and Across and introduced in 2024, formalises the cross-chain intent interface so solvers and fillers across these protocols can share a common order format.
Across: Optimistic Oracle as Arbiter
Across is an intents-based bridge where relayers front destination-chain funds in 2-30 seconds; 85% of fills are L2-to-L2 and complete in roughly 1 second.
Settlement flows back via UMA's Optimistic Oracle. A designated actor called a dataworker bundles fulfilled intents and posts them on-chain. If no challenger disputes the bundle within a 2-hour challenge window, the HubPool contract on Ethereum mainnet reimburses the relayer from pooled LP capital. Disputed bundles are adjudicated by UMA's dispute mechanism. Dispute rates sit below 0.01% of all transactions.
The full reimbursement cycle is roughly 3 hours — the 2-hour challenge window plus approximately 1 hour for repayments to settle. Relayers must be capitalised enough to float their fills for that period. If you are an LP, your capital sits in a single Ethereum HubPool rather than fragmented per-chain pools, which maximises capital efficiency but concentrates counterparty risk in one contract.
Across co-authored ERC-7683 with Uniswap Labs. After its solver network migrated to the standard in Q3 2025, ERC-7683 orders represented 88% of total Across volume, with 70+ protocols supporting the standard. Total volume exceeded $19 billion by early 2026, with monthly volumes reaching $1 billion. The protocol launched V4 in July 2025, introducing a ZK-proof architecture that allows new chains to be added within hours. Median transaction fee is approximately $0.04.
Chainflip: Native Assets via Threshold-Signature Vaults
Chainflip's thesis is eliminating wrapped tokens entirely. A user can swap native BTC directly for native ETH or native SOL — the swap never touches a synthetic representation of any asset.
This works through a network of 150 proof-of-stake validators that collectively control multi-chain vaults using the FROST threshold signature scheme (a TSS/MPC design). Signing any vault transaction requires a two-thirds supermajority: 100 of 150 validators. No single validator, and no minority coalition, can unilaterally move vault funds.
Supported chains as of early 2026 include Bitcoin, Ethereum, Solana, Arbitrum, Polkadot/AssetHub, and BNB Chain (in rollout). USDC is available across several of these.
Chainflip prices swaps using a JIT AMM (Just-In-Time AMM). Rather than relying on passive pool liquidity, market makers observe incoming deposits in real time during the roughly 60-second confirmation window, then compete by updating range and limit orders just before the validator network executes the swap. This produces tighter spreads, particularly for large trades: a $10M TVL pool can absorb $3.5-4M directional trades with minimal slippage.
Because source-chain confirmation is required before the validator network acts, fills take approximately 60 seconds rather than the 1-2 seconds achievable on purely EVM bridges. The protocol has processed over $6.39 billion in total volume across 2 million swaps as of February 2026; 30-day volume was approximately $336 million across 34,000+ swaps at an average size near $9,800. In early 2026 Chainflip launched native BTC lending — a permissionless cross-chain lending market for native Bitcoin, ETH, SOL, and stablecoins without wrapped tokens.
Relay: The Speed-First Execution Layer
Relay positions itself as the fastest and broadest cross-chain execution layer. As of 2026 it covers 85+ networks, has served 5M+ users, processed 50M+ transactions, and settled $20B+ in volume. It is integrated into wallets and exchanges including Phantom, MetaMask, OpenSea, and Coinbase Wallet.
The model is familiar: the user signs an intent, a relayer fronts destination-chain funds immediately (sub-2-second fill in most cases), then proves fulfillment to unlock reimbursement from the origin-chain escrow.
What distinguishes Relay is its settlement layer. Rather than batching reimbursements through a long challenge window, Relay raised a $17M Series B to launch Relay Chain — a dedicated settlement chain where each individual order settles point-to-point at roughly $0.005 per order. A Relay Chain oracle attests fills and updates solver balances in real time, compressing the reimbursement cycle from hours to near-instant.
Relay also supports not just token bridges and swaps but arbitrary contract calls in a single intent across its 85+ networks, making it a preferred infrastructure layer for wallets and exchanges that need one API across many chains. Solvers stake capital and face slashing for misbehaviour, establishing economic security without a fixed validator set.
Compared to Across: Relay has broader chain coverage and sub-2-second fills, but its settlement model is newer and the Relay Chain introduces a distinct trust surface with a shorter track record. Compared to Chainflip: Relay routes through relayer inventory rather than native-asset vaults, so users typically receive a canonical bridged asset rather than the native token.
Settlement Models Side by Side
| Across | Chainflip | Relay | |
|---|---|---|---|
| Fill speed (user) | 1-30 seconds | ~60 seconds | Sub-2 seconds |
| Relayer reimbursement | ~3 hours (2-hr challenge window) | Same cycle as fill | Near-instant via Relay Chain |
| Trust model | UMA Optimistic Oracle + HubPool | 100-of-150 validator supermajority (FROST) | Relay Chain oracle + staked relayers |
| Asset coverage | EVM chains, major ERC-20s | Native BTC, ETH, SOL — no wrapping | 85+ networks, canonical bridged assets |
| Median fee | ~$0.04 | JIT AMM spread | ~$0.005/order |
| ERC-7683 support | Yes (co-author) | No | Partially |
Speed to the user is comparable across all three: funds arrive in seconds. What differs is how fast solvers recover their capital, and which entity you are trusting to govern that recovery.
For moving value between Ethereum rollups, Across is the dominant intent bridge today, with deep liquidity and a long dispute-resolution track record. Chainflip is the only option when you need native BTC or native SOL on the other side without wrapping. Relay wins on chain coverage breadth and developer integration surface.
Risks to Understand Before You Participate
Intent bridges repackage bridge risk — they do not eliminate it.
Source-chain escrow as honeypot. The liquidity pool or escrow contract that reimburses relayers is still a high-value, concentrated target. Anyone who can compromise the settlement mechanism — the oracle, the validator set, or the settlement chain — can redirect reimbursements and drain pooled capital.
The scale of what is at stake. Bridge infrastructure remains the most-exploited component in crypto. Cumulative bridge losses have exceeded $2.8 billion since 2022, roughly 40% of all value hacked in Web3. 2026 alone saw major bridge-related incidents, including a ~$292M LayerZero-related exploit at a liquid restaking protocol and an $11M Verus-Ethereum bridge hack in May 2026. How interconnected protocols can fail together is not a theoretical concern.
Relayer solvency risk. Your fill is only as good as the relayer's balance sheet. If the relayer that fronted your funds becomes insolvent before reimbursement clears, escrowed funds may be delayed or contested. Professional relayers reduce this risk; they do not remove it.
Oracle risk for Across. The UMA Optimistic Oracle has a sub-0.01% dispute rate — evidence the system is working, not evidence it is invulnerable. A governance attack, a malicious dataworker posting unchallenged false bundles, or a contract bug in HubPool could allow fraudulent reimbursements.
Validator and MPC risk for Chainflip. The FROST key-generation ceremony and 100-of-150 threshold are robust, but not unconditional. A supply-chain attack on validator software, or a flaw in the FROST implementation, could threaten multi-billion-dollar vaults.
Settlement-chain immaturity for Relay. Relay Chain is newer infrastructure without a long record of adversarial stress. The staked-relayer slashing mechanism has not been battle-tested under sustained attack.
Practical hygiene: Do not bridge more than you can afford to lose through any protocol with a short audit history. Always confirm you are on the official UI — phishing front-ends are active. Verify any protocol-generated deposit address independently. Check route liquidity depth before large transfers. Confirm the destination-side fill on a block explorer before considering a transfer complete.
Key Takeaways
- Cross-chain atomicity is impossible — intent bridges solve this by having relayers front destination liquidity and recover from a source-chain escrow after settlement proof arrives.
- Across uses UMA's Optimistic Oracle with a 2-hour challenge window and a single HubPool; fast for users, slower reimbursement cycle for relayers.
- Chainflip eliminates wrapped tokens using a 100-of-150 FROST validator threshold; it is the only major protocol handling native BTC-to-ETH swaps without wrapping, but fills take ~60 seconds.
- Relay prioritises speed and breadth with sub-2-second fills across 85+ networks, backed by a dedicated settlement chain that compresses reimbursement to near-instant — at the cost of a newer trust surface.
- All three repackage, not remove, bridge risk: the escrow or liquidity pool remains a target; evaluate the trust model before routing large transfers through any of them.
