Convex Finance

Convex Finance is not a DAO that governs a product of its own so much as a governance layer built on top of another DAO. It sits on Curve and solves the coordination problem that vote-escrow created: to earn Curve's maximum reward boost and to steer weekly CRV emissions, you had to lock CRV as veCRV for up to four years – illiquid, non-transferable, and pointless at small size. Convex lets holders pool that lock.

You deposit CRV, Convex locks it as veCRV on your behalf and votes the whole accumulated balance as a single bloc. Within months of its May 2021 launch Convex controlled more than half of all veCRV, making it the effective power broker – the "kingmaker" – of Curve governance and the central character in the Curve Wars. Five years on it still does, and the sections below check that claim against the chain rather than repeating it.

The bloc, measured

Every figure here is a direct read of Ethereum mainnet at block 25,716,457 (9 August 2026, 09:14 UTC). The voter-proxy address is not taken from a dashboard: it is what the CVX token contract itself returns from vecrvProxy().

The two shares do not match, and the gap is the whole point of a perpetual lock. Convex holds under half the locked CRV but over half the voting weight, because veCRV decays linearly toward its unlock date: a holder who locked for a year eighteen months ago has nothing left, while Convex's proxy is bumped back out to the four-year maximum and therefore sits permanently near its full weight. The bloc does not have to grow to keep the majority – it only has to stop decaying while everyone else's does. That is a structural feature of the vote-escrow design rather than anything Convex added, and it is the mechanism that makes the 2021 kingmaker result durable rather than a snapshot.

cvxCRV – a liquid, boosted claim on a permanent lock

When you deposit CRV into Convex you receive cvxCRV in return. The underlying CRV is locked as veCRV essentially forever – Convex keeps re-locking it at the maximum term, so the deposit is one-way: cvxCRV can only be exited by selling it on the secondary market, never redeemed back into the locked CRV. In exchange, cvxCRV stakers receive the veCRV yield – trading fees, the CRV boost, and a cut of the bribes below – without ever managing a lock themselves. Convex takes a 17% platform fee on all CRV revenue: 10% to cvxCRV stakers, 4.5% to CVX/vlCVX lockers (paid as cvxCRV), 2% to the treasury, and 0.5% to whoever calls the harvest.

That exit is not priced as a rounding error. On 9 August 2026, 09:14 UTC, cvxCRV traded at $0.11469 against CRV's $0.22717 – 50.5 cents on the dollar, a 49.5% discount, with an independent price feed agreeing to four decimal places. The discount is the market's standing quote on irreversibility: a cvxCRV holder owns the yield of a lock that runs to 2030 and no way to reach the principal, and buyers price that gap rather than arbitrage it away. Anyone reading a liquid-locker receipt as a soft-peg substitute for the underlying asset should read this number first – it is the clearest live evidence on the wiki that an exit route you cannot actually take is worth roughly half of one you can.

vlCVX, Votium, and the vote market

Depositing CRV hands Convex the voting power; the second token, CVX, decides how that power is used. To vote on Curve gauge weights through Convex – or on Convex's own proposals – a holder must vote-lock their CVX into vlCVX for a minimum of 16 weeks (16 epochs). vlCVX holders direct the bloc and earn a share of protocol fees. Because a single vlCVX vote moves a large, aggregated chunk of veCRV, an entire bribe market grew up around it: on Votium, any protocol that wants Curve emissions steered toward its own pool posts an incentive, and vlCVX voters who delegate their vote collect it – 96% to voters, 4% to Votium.

This turned governance influence into a rentable, cash-flowing commodity, and the pattern spread: Aura Finance is the same liquid-locker model rebuilt on top of Balancer's veBAL – and, having voted itself out of existence in April 2026, it is also the worked example of what such a layer leaves behind. It is the canonical worked example on the token-weighted voting page: locking a token couples influence to commitment, but once the lock is tokenised the influence can simply be bought a layer up.

The mint that bought the bloc has closed

CVX was never handed out on a calendar. The deployed token contract mints it as a side effect of Convex claiming CRV for its liquidity providers, on a step function written into mint():

uint256 cliff = supply.div(reductionPerCliff);   // reductionPerCliff = 100,000e18
if (cliff < totalCliffs) {                       // totalCliffs = 1000
    uint256 reduction = totalCliffs.sub(cliff);
    _amount = _amount.mul(reduction).div(totalCliffs);
    ...
}

Each 100,000 CVX minted advances one cliff and cuts the CVX-per-CRV rate by a further 0.1%. At block 25,716,457 the supply is 99,984,078.08 of a hard 100,000,000 cap – cliff 999 of 1000. The reduction multiplier is therefore 1/1000: a liquidity provider whose CRV claim would have minted a full CVX in 2021 now mints a thousandth of one, and clearing the remaining 15,921.92 CVX would take roughly 15.9 million CRV of future claims. Note also what the branch does not have: an else. Once cliff reaches 1000, mint() runs, mints nothing, and returns successfully – the reward pipeline never reverts, it just quietly stops paying in CVX.

The governance consequence is larger than the tokenomics one. CVX was the subsidy that bought the veCRV majority: supply liquidity, earn CRV, receive CVX, and with it a claim on the direction of Curve's emissions. That route is now effectively closed. The bloc it purchased is permanent and still votes, but nobody can earn their way into steering it any more – the only remaining path to Curve's gauge weight is buying CVX from someone who already holds it, at whatever price they name. A distribution mechanism that reads as generous while it runs becomes an enclosure the moment it finishes, and the finishing line here is a single integer comparison in a contract deployed five years ago.

CVX tokenomics and the meta-governance model

CVX has a fixed 100 million maximum supply. Rather than a time-based schedule, CVX is minted every time Convex claims CRV for its liquidity providers, and the CVX-per-CRV ratio steps down as more CVX enters circulation – issuance that tapered automatically as the protocol matured, and that has now, as the section above measures, essentially run out. That design made CVX itself the object of accumulation: because vlCVX controls Curve's emission gauges, buying and locking CVX is the cheapest way to command veCRV influence, which is why treasuries such as Curve-adjacent protocols and, at its peak, Frax and others amassed large vlCVX positions. Convex is the clearest demonstration of meta-governance as a market: the right to direct another DAO's parameters became a liquid asset with its own token, its own yield, and its own attack surface, all layered on top of the base protocol without its explicit consent.

A sixteen-week electorate voting a lock that runs to 2030

Read the two commitments side by side and they are not the same commitment at all. The CRV underneath is locked until 25 July 2030 and re-locked before it can ever run down. The people who decide how that stake votes are committed for 16 weeks – lockDuration() on the vlCVX locker returns exactly 9,676,800 seconds, and the contract was on its 233rd weekly epoch at block 25,716,457. A vlCVX holder can lock, steer four and a half years of somebody else's escrowed capital through 16 weekly gauge votes, take the Votium incentives that came with them, and walk.

The electorate is also a minority of a minority. 45,739,877.54 CVX is vote-locked – 45.75% of all CVX in existence – which means a bloc that is itself under half the token steers 53.68% of Curve's gauge weight, which in turn directs emissions against 850.9 million CRV of locked capital. None of that is hidden or illicit; it is what every layer of the stack was designed to do. But it is a precise measurement of the thing the vote-escrow model was invented to prevent.

Curve made influence expensive in time so that whoever steered emissions would still be there to live with the result. Convex kept the four-year time cost on the capital and reduced the time cost on the decision to sixteen weeks, and the difference between those two numbers is where the vote market makes its money. Duration mismatch between the stake and the voter is the structural defect worth carrying away from this page – see token-weighted voting and DAO tokenomics for where else it shows up.

How Caper approaches this

Convex is what happens when a lock becomes a product: veCRV was meant to tie influence to commitment, and Convex financialised exactly that bond – CVX is freely transferable, vlCVX voting rights are rented out on an open market, and control of Curve drifted toward whoever could pay. A caper chases the same goal – coupling say to genuine commitment – but leaves nothing for a Convex-style layer to aggregate. Its voting weight multiplies the stake a member holds by the participation they have actually earned, and that participation record is soulbound – minted to the member and non-transferable, so it cannot be pooled, tokenised, or rented to a third party.

Holdings still matter (a bigger position is a larger factor in the weight), but the earned half never leaves the account it was minted to, so no external cartel can aggregate a controlling bloc the way Convex did over Curve – a cartel can buy in and vote its way to its own v, but it cannot bribe or borrow anyone else’s. And a caper has no emission gauges to steer – its tokens come from a bonding curve, not a weekly inflation vote – so there is no emission stream for a "war" to capture in the first place.

The duration mismatch resolves differently too, because a caper has no lock to mismatch against. There is no escrow term and no unlock date to wait out: a member exits whenever they choose, surrendering their governance tokens and their soulbound vote tokens together, and the treasury share they receive is the same weight formula they voted with. Commitment is not enforced by freezing anyone in place; it is enforced by making the value of leaving and the power of staying the same number. This is a design contrast, not a claim of superiority; the mechanics are on the linked pages and verified against the contract.

References

• Convex Finance documentation – Understanding CVX, vote-locking, and fees (primary).
• Convex – for CRV stakers: the cvxCRV deposit and permanent-lock mechanics (primary).
• On-chain state, Ethereum block 25,716,457 (9 August 2026, 09:14 UTC), read directly and reproducible with eth_call: ConvexToken (CVX) – totalSupply(), maxSupply(), totalCliffs(), reductionPerCliff(), vecrvProxy(); Curve VotingEscrow – totalSupply(), supply(), balanceOf() and locked() for the proxy; CvxLockerV2 (vlCVX) – lockedSupply(), lockDuration(), epochCount(). The verified source of mint() quoted above is on the CVX contract page (primary).
• DefiLlama price feed and CoinGecko – the cvxCRV/CRV ratio, cross-checked between two independent sources on the same day.
• Votium documentation – the vlCVX incentive/bribe marketplace and its 96/4 fee split (primary).
• Curve – what is veCRV: the lock term, the linear decay, and the boost the lock buys (primary).
• TokenBrice, CRV Wars and Advanced CRV warfare – analysis of Convex's veCRV dominance and the protocols built on it.
• Aura Finance – the liquid-locker model applied to Balancer's veBAL, and its wind-down.