Aura Finance was Convex for Balancer: a protocol whose only product was other people's voting power. Depositors handed it BAL, it locked that BAL as veBAL and never unlocked it, and the resulting bloc steered Balancer's gauge weights on behalf of whoever held the second token. The lineage is not an analogy but a fork – Aura's own GitHub org still carries convex-platform, forked from Convex's repository.

It worked. By Balancer's own account Aura was responsible for "at least 60% (usually >75%) of the votes of every Snapshot proposal". Then Balancer decided to retire the vote-escrow system Aura was built on, and Aura – a protocol with no product of its own beneath the one it was layered over – voted to dissolve itself.

What makes Aura worth a page is not the wind-down but its shape. Four months after its holders voted to shut it down, the position it accumulated is still the single largest bloc in Balancer's electorate, and it cannot be unwound until 2027. This is the part of dissolution that governance design usually forgets: a lock outlives the organisation that took it out.

The bloc that outlived the protocol

Every figure below is a direct eth_call against Ethereum mainnet at block 25,758,271 (15 August 2026, 05:11 UTC), not a dashboard reading.

80/20 BAL/WETH BPT locked by the Aura VoterProxy	3,684,354.766
Its lock expiry (locked__end on veBAL)	1808956800 → 29 April 2027
Its veBAL balance	2,592,008.185
veBAL total supply	3,573,920.103
Aura's share of all veBAL	72.53%
AURA total supply	86,413,430.86
vlAURA outstanding (vote-locked AURA)	332,317.70 – 0.38% of AURA

Read the last two rows against the fifth. The electorate that nominally directs this bloc has almost entirely gone home: less than four tenths of one percent of AURA is still vote-locked, and the proposal evicting Aura notes that the locker itself has been shut. The stake it accumulated, meanwhile, is unchanged and immovable. A vote-escrow lock is a one-way commitment by construction, which is exactly what made it credible on the way in and exactly what makes it a liability on the way out.

auraBAL, still backed to within one BPT

Aura's deposit token is the mechanism worth understanding, because it is what a liquid locker actually sells. Deposit BAL – strictly, the 80/20 BAL/WETH pool token that veBAL takes – and Aura locks it permanently and hands back auraBAL, a transferable claim on a position that will never be individually withdrawable. The depositor gets liquidity; the protocol gets a vote that never decays because it is never allowed to.

The chain shows the accounting held to the end. auraBAL total supply at block 25,758,271 is 3,684,353.766; the BPT locked by the VoterProxy is 3,684,354.766. The difference is exactly 1.000 BPT – the seed unit – against a supply of 3.68 million. Whatever else the wind-down costs holders, it is not a shortfall between the token and the thing behind it.

What the depositor cannot do is leave early. AIP-78 is explicit that auraBAL "was expected to be permanently locked and now has almost no counter-trade liquidity", which is why the redemption is staged over a year rather than paid at once. The comparison with Convex's cvxCRV is instructive: same design, same permanence, and the same absence of any redemption path that does not route through the secondary market until the underlying protocol chooses to unwind.

AIP-78: eight addresses vote to dissolve

AIP-78, "Orderly Protocol Wind-Down and Treasury Redemption for Aura", ran 9–12 April 2026 and passed 17,171,688.83 vlAURA for, 0 against, 0 abstain, on eight voting addresses. It is the last proposal the aurafinance.eth space has carried, out of 905.

The reasoning is unusually plain for a governance document. Once Balancer moved to sunset veBAL, "the foundational mechanism that Aura was built on top of is potentially being retired … Without veBAL or with a watered down version, Aura's core value proposition ceases to function." The proposal adds the operator's arithmetic behind it: projected vlAURA yield had fallen below the cost of running the protocol, after BAL emission cuts, a price decline in both tokens, and the TVL withdrawal that followed the November 2025 Balancer exploit.

The terms it set:

• Stage 1, four weeks after passage. Withdrawal-only mode for LPs, AURA emissions off, the veBAL position begins to unwind, and 100% of treasury assets are redeemable pro-rata by AURA holders – who also receive a 1:1 auraRedeem token for the second stage. Treasury-held, ecosystem and unvested AURA are excluded from the calculation.
• Stage 2, fifty-six weeks after passage. The veBAL lock expires, the underlying BPT is withdrawn, and 90% of it goes to auraBAL redeemers with 10% to auraRedeem holders, alongside any unclaimed stage-1 assets.
• Estimated redemption: approximately $0.0226 per AURA, on the proposal's own figures.

Aura also claimed 67.2% of Balancer's BIP-920 veBAL compensation airdrop – 336,074.19 of 500,000 USDC to the same VoterProxy – which AIP-78 routes to AURA holders in full. A wind-down vote by eight addresses redistributing a seven-figure position is a fair test of whether quorum design is doing any work at the end of a protocol's life.

BIP-924: Balancer evicts the aggregator and halves its own quorum

Aura's exit created a problem in the other direction, and Balancer is fixing it in public. BIP-924, "Exclude Aura from Governance and Reduce Quorum", opened 14 August 2026 and closes on 18 August. At the time of writing it carries 19,050,782 BAL for, 0 against, from two voting addresses, against a 10,000,000 BAL quorum.

It does two things. It overwrites the voting power of the Aura VoterProxy 0xaF52695E1bB01A16D33D7194C28C42b10e0Dbec2 to zero in the balancer.eth aggregator, so the underlying BAL becomes eligible again only as individual users redeem it after the 2027 unlock. And it halves the quorum from 10m to 5m BAL, on the stated ground that Aura's votes were capped in power but still counted toward quorum.

That second clause is the interesting one, because Balancer had already been failing its own bar. BIP-922 and BIP-923 both closed unanimously on 4 August 2026 with 6,473,256 and 6,454,723 BAL – on eleven and ten voting addresses – well short of 10m. A quorum calibrated around a participant that has left stops measuring participation and starts blocking it. Balancer's BIP-921 had already moved the space from veBAL weighting to a 1-BAL-1-vote formula in May 2026; BIP-924 is the second half of unwinding the same experiment.

For the wider pattern this belongs to – a governance parameter tuned to conditions that no longer hold – see quorum and threshold design and how DAOs fail.

The documentation died before the protocol did

Anyone researching this wind-down should know where the record actually is, because two of the three obvious places are gone. Checked 15 August 2026:

• forum.aura.finance resolves to nothing. No A record. It is the discussion link on AIP-78 itself – the deliberation behind the wind-down vote no longer answers at its own address. The Internet Archive's last capture of the forum root is 18 May 2026, and the AIP-78 topic survives there at 8 May 2026.
• docs.aura.finance resolves to nothing either, though the GitBook source repository is still public.
• The code and the app are both still up. aura-contracts is MIT-licensed, unarchived, last pushed 13 May 2026, and app.aura.finance still serves the redemption interface.

The ordering is the lesson, and it is not unique to Aura: the prose dies first, the code and the ledger outlive it. Cite the Snapshot space, the contracts and the chain for anything load-bearing, and treat a project's own forum as the first thing to disappear. The same failure sequence is catalogued at DAO tooling discontinuity.

How Caper approaches this

Aura's members were never able to leave. Their capital was locked in a position only a collective vote could unwind, so when the reason for the lock disappeared the only available exit was a proposal to dissolve the whole protocol – and then a year of waiting for a timestamp. That is the structural cost of buying influence with a commitment device: the commitment is real in both directions.

A caper puts the redemption in the member's own hands. Exit is a unilateral call against the treasury, not a proposal – no counterparty, no quorum, no wind-down vote, no unlock date. The condition is participation rather than duration: exit() requires a non-empty bucket of the caper's soulbound vote tokens alongside the governance tokens and aborts without them, so the members who took part in deciding what the treasury was for are the ones who can claim against it. A holder who never voted has no exit to take, not a zero-valued one.

That does not make Caper better at what Aura did – a caper has no gauge weights to sell and no vote market to run. It makes the failure mode different. Nothing in a caper can arrive at the position Aura is in, where the organisation has agreed to dissolve, the electorate has dispersed, and the treasury is still hostage to a lock until 2027. Compare rage-quit and exit rights for the wider design space.

References

• AIP-78 – Orderly Protocol Wind-Down and Treasury Redemption for Aura (primary; Snapshot, aurafinance.eth, closed 12 April 2026).
• BIP-924 – Exclude Aura from Governance and Reduce Quorum and BIP-921 – 1-BAL-1-Vote Reconfiguration (primary; Snapshot, balancer.eth).
• forum.aura.finance – the AIP-78 discussion thread, via the Internet Archive (the live host no longer resolves).
• On-chain, Ethereum mainnet at block 25,758,271: veBAL, the Aura VoterProxy, AURA, auraBAL and the vlAURA locker.
• aurafinance/aura-contracts and aurafinance/convex-platform (the Convex fork Aura was built from).