Verdict
For swapping between assets that should trade near parity, nothing beats Curve's maths. The interface is from another era, the governance economy is baroque, and the protocol has one serious exploit in its history that was not its own code's fault.
- Best for
- Large stablecoin and pegged-asset swaps
- Cost
- 0.01–0.04% on stable pools
What works
- Stableswap invariant gives dramatically lower slippage on pegged assets
- Fees on stable pools are as low as 0.01%, an order below standard AMMs
- Deep liquidity for stablecoins and liquid staking derivatives
- veCRV vote-escrow aligns emissions with pools that actually get used
What does not
- Interface is dense and unwelcoming to anyone new
- July 2023 Vyper compiler exploit drained roughly $70m from several pools
- The vote-escrow and incentive economy is genuinely hard to evaluate
- Founder's leveraged CRV position became a systemic event in its own right
Why the maths matters
A standard constant-product AMM prices every pair the same way, which is correct for ETH against a stablecoin and badly wrong for USDC against USDT. Two assets that should be worth the same amount do not need a pricing curve that assumes they might diverge tenfold.
Curve's stableswap invariant flattens the curve near parity and only steepens at the edges. The practical result is that a $2 million USDC-to-USDT swap executes at a price a standard AMM cannot approach, and the pool fee is 0.01% to 0.04% rather than 0.30%.
For stablecoins, liquid staking tokens and wrapped assets, this is not a marginal improvement. It is the difference between a viable trade and an expensive one, and it is why Curve remains structurally important despite everything else in this review.
The governance economy
Lock CRV for up to four years and you receive veCRV, which gives voting power over which pools receive CRV emissions. Protocols that want liquidity for their own token then compete to influence those votes, usually by paying veCRV holders directly.
It is an elegant solution to a real problem — how to direct emissions to pools that matter rather than pools that are already profitable — and it created an entire secondary market in vote incentives, with its own intermediaries and its own derivatives.
It also means the yield on a Curve pool is partly the output of a political process. Evaluating a pool requires understanding that process, and if you cannot explain where a pool's yield comes from in one sentence, you should not deposit into it.
The exploit
In July 2023 a reentrancy bug in specific versions of the Vyper compiler — not in Curve's own logic — allowed roughly $70 million to be drained from several pools. A substantial portion was returned by white-hat actors and the affected pools were made largely whole.
The distinction matters for how you read it. Curve's contract code was not the flaw; the language toolchain beneath it was. The lesson is that an audit covers the code, not the compiler, and that dependency risk in DeFi goes further down the stack than most users imagine — including into tools that are themselves open source and widely trusted.
You can audit a contract and still be exploited by the compiler that produced it. In DeFi, the trusted computing base is larger than the part anyone reads.
The founder position
A large leveraged borrow against CRV held by the protocol's founder became, for a period, a systemic risk to the protocol itself: a liquidation would have dumped enough CRV to break the very pools that backed the loan. It was unwound through a series of OTC sales and repayments.
No user funds were lost. But it remains a permanent example of a governance-token concentration risk that no audit would have caught and no smart contract could have prevented.
crvUSD
Curve's own stablecoin uses a soft-liquidation mechanism that converts collateral gradually as the price falls rather than triggering a single liquidation event. It is a genuinely novel design and it has behaved well through several drawdowns.
It is also complex enough that most borrowers do not understand what their position does in a fast move, or that recovering from a partial soft-liquidation leaves them with a different asset mix than they started with.
Verdict
Where Curve sits in the stack
Curve is infrastructure other protocols depend on. Lending markets price collateral partly off Curve pool depth. Liquid staking tokens rely on Curve liquidity to hold their peg to the underlying. Stablecoin issuers seed Curve pools before they seed anywhere else.
That makes Curve's health a systemic question rather than a product question. A severe liquidity withdrawal here propagates into lending liquidations and peg dislocations elsewhere, which is exactly what made the founder's leveraged position a market-wide concern rather than a personal one.
Depositing, practically
Prefer pools whose assets you would be happy holding in any proportion — in a depeg, you end up holding all of the broken one.
Read the pool's amplification parameter. A high value means tighter pricing near parity and a sharper move once the peg breaks.
Separate the base yield from the CRV emissions and the vote incentives before comparing anything.
Check whether the pool holds a wrapped or bridged asset, which adds a bridge's risk to the pool's risk without adding it to the name.
Score: 7.9. The best execution in DeFi for the specific job it does, with an interface that activelydiscourages casual use and a history that includes one large toolchain-level exploit. For stablecoin routing at size, use it directly. For everything else, an aggregator will find it for you anyway.