BonkDAO is the governance body that spends the treasury of BONK, the community memecoin launched on Solana on Christmas Day 2022. It is worth a directory entry for a reason that has nothing to do with memes: in July 2026 BonkDAO became the clearest worked example of a treasury whose market price was published in its own governance settings. A proposal passed when Yes votes reached 1% of token supply, and nothing at all was required of the other 99% – so anyone willing to buy 1% of BONK could pass a proposal alone. On 6 July 2026 someone did exactly that and moved roughly $20 million out of the treasury (CoinDesk). No contract was broken. The vote was valid, and the whole of it is still on-chain.

An airdrop that became an ecosystem

BONK launched on 25 December 2022 with an initial supply of 100 trillion tokens and no private sale, no venture allocation, and no team round. Half of that supply was airdropped across the Solana community – traders, NFT collectors, artists, developers and contributors – during a period when Solana activity had collapsed in the wake of the FTX failure, which is most of why the token attached itself to the chain's identity so quickly (bonkcoin.com). A share of the launch supply was set aside as a DAO-controlled treasury; that treasury is the thing BonkDAO votes on.

What grew on top of it is unusual for a memecoin. BONK is now the settlement token for a cluster of consumer products – a swap venue, spot and perpetual trading, a Telegram trading bot, and the LetsBonk token launchpad – several of which route a slice of fees into buybacks and burns. Supply has fallen from 100 trillion at launch to 87,994,588,756,876.71 BONK, read directly from the mint account on 14 August 2026, with a market capitalisation near $255 million (CoinGecko, 2 August 2026). Developer documentation for the ecosystem sits at build.bonkcoin.com.

The gap between that product surface and the DAO's governance surface is the story. BONK acquired an ecosystem, a burn mechanism and 18,476 registered token-owner records on its realm; its DAO kept the governance configuration of a 2022 community project.

How BonkDAO governs

Proposals are filed as Bonk Improvement Proposals and voted token-weighted on Realms, the standard Solana governance front end built on the SPL Governance program. Realms is a capable framework – it supports councils, multiple vote thresholds, and configurable execution delays – but those protections are per-realm settings, not defaults, and BonkDAO's realm was configured without them. The settings below are not reported figures: they are the fields of the community governance account Uq5BRkVfdBpMknZJHw6huS3dunEgJpUDv3M2DG3BfQg, readable by anyone against a Solana RPC node.

• Approval threshold: YesVotePercentage(1). This is the detail most accounts of the attack get slightly wrong by calling it a quorum. A quorum is a floor on participation; this is a floor on Yes weight alone, measured against the full token supply. No votes do not raise it, abstentions do not raise it, and a turnout of one wallet satisfies it as readily as a turnout of ten thousand. At the supply recorded on the proposal itself, the line stood at 879,947,133,976.03 BONK.
• Voting window: 432,000 seconds base plus an 86,400-second cool-off – five days of open voting and a sixth day in which positions can only be softened or withdrawn. The proposal is public for the whole six days. Visibility only helps if someone is watching.
• Execution hold-up: min_transaction_hold_up_time = 0. There is no timelock. A passed instruction becomes executable the instant the tally closes, so there is no window in which a community could cancel a transfer it had just noticed (Halborn).
• Cost of filing: 100,000,000 BONK (min_community_weight_to_create_proposal) – a few hundred dollars at the July 2026 price. The expensive part of the attack was never the ticket to file.

None of these are Realms defects. They are the choices a DAO makes when its governance is set up in a week and never revisited as the treasury grows – the general failure pattern catalogued under How DAOs fail. The part that generalises past this one DAO – a pass line denominated in tokens, guarding a treasury denominated in dollars, with an exchange rate between them that moves without any vote – is treated in general on Quorum and threshold design.

BIP #76: the treasury bought for a fifth of its value

On 30 June 2026 at 08:25:24 UTC an anonymous wallet filed BIP #76 – "Sowellian BonkDAO", presented as a governance-renewal plan that would install new members and a council and "rebuild from the ashes". Its own description, stored on-chain, names the operative instruction in plain text: "Send 4.426.104.450.305 Bonk to 9bxWkNf3BtJ6iehq9KbX9uCWMjem4TFiPZ19T2sYJHvQ", and it offers the reader a motive to vote for it – "all YES voters are eligible to receive tokens". Sign-off followed 15 seconds after drafting and voting opened immediately. Over 4–5 July the same actor bought just over 1% of BONK supply on Bybit and Binance, part-funded by borrowing against DeFi collateral, for roughly $4.4 million – an amount sized to clear the pass line and little more.

Voting closed on 6 July 2026 at 08:25:39 UTC, six days to the second. The proposal account records the whole tally, and seven vote records exist against it – seven wallets out of 18,476 registered token-owner records, or 0.038% of them:

• Yes: 882,383,387,283.79 BONK, or 1.0028% of the 87,994,713,397,603.04 BONK supply the proposal snapshotted – over the 879,947,133,976.03 line by 0.28%.
• No: 710,848,288.77 BONK across four wallets. Four holders read the proposal, understood it, and voted against it; their combined weight was 0.08% of the Yes side, which under an approval threshold makes no difference whatsoever.
• One wallet, CyEE7oHVDaFJ5xZLbXY3h2Z2uk1VwhTkdy72kPUEtypQ, cast 882,285,249,588.80 BONK – 99.91% of all weight cast, and on its own 0.27% above the pass line. The two other Yes votes, 97.9 million and 0.25 million BONK, were not needed.

Total weight cast came to 883,094,235,572.57 BONK: 1.004% of the token supply decided the vote. Execution began 38 seconds after the tally closed, at 08:26:17 UTC, and the proposal was closed out 11 seconds later – four transactions, all four executed. The buyer then relinquished their vote record, releasing the deposited tokens. BonkDAO confirmed the loss publicly, said it had identified the exchange wallets used to accumulate the stake, and began working with exchanges, bridges, the Solana Foundation and law enforcement; at least one exchange suspended BONK transfers while the funds moved (CryptoSlate).

The full technical account – the attack families it belongs to, the defences it defeated, and how it compares with Beanstalk, Steem and Compound Prop 289 – is in DAO security and governance attacks. The participation side of it is treated in Voter apathy and governance participation.

The same proposal had already been defeated once

The on-chain record holds a detail no account of the attack has reported, and it changes the moral of the story. BIP #76 was not a novel idea that caught a sleeping DAO by surprise. Its shape had been tried seven months earlier and rejected.

On 22 November 2025 a proposal titled "The Orange Project" (DZpjvxst93zDWUSPVhVBQdQCxpcXLtTzVtZmr1EqcwPe) was filed against the same governance, carrying two transactions and the same inducement in its description – "Active supporters of this proposal will be recognized and rewarded." It was defeated: 30,212,382,299 BONK in favour against 77,334,737,047 BONK opposed, reaching only 3.4% of the pass line. Twenty-two wallets voted, one of them casting a pointed 69,420,000,000 BONK against.

So the DAO that lost $20 million in July 2026 had mustered three times as many voters to kill a similar proposal in November 2025 – and, between BIP #75 on 7 February 2025 and BIP #76 almost seventeen months later, passed nothing at all. The exposure was never that BonkDAO's members could not recognise a treasury grab. It was that a threshold measured only in Yes weight lets an attacker choose an electorate of one, and that nothing in the configuration required the twenty-two who had shown up before to show up again.

What it changed, and what it didn't

The reform list circulating around BonkDAO and the wider Solana ecosystem is the expected one: mandatory execution timelocks on treasury instructions, a pass line scaled to the value at risk rather than to a fixed share of supply, multisig or council sign-off on large transfers, and anomaly review before execution (crypto.news). These remain proposals rather than shipped configuration: read on 14 August 2026, the community governance account still carries YesVotePercentage(1) and a hold-up time of zero. Recovery of the drained tokens is unresolved.

The transferable lesson is narrower than "memecoin DAOs are unserious". BonkDAO's 1% was a number chosen when the treasury was worth little and the token was a joke; it was never re-priced as the treasury grew into eight figures. Any DAO whose pass line is denominated in tokens while its treasury is denominated in dollars has the same exposure, and the exchange rate between the two moves without anyone voting on it. The general treatment of that problem sits under DAO treasury management; the mechanics of the vote itself under Token-weighted voting and Proposal lifecycle.

How Caper approaches this

Caper does not solve governance capture, but it prices it differently, and it does not settle the question at the ballot. Voting weight on a caper is not a function of tokens held: each ballot is credited (t·v)/(V·T) – the voter's token balance t multiplied by their soulbound vote tokens v, over the vote-token supply V times the circulating supply T, with T frozen when the proposal is raised. v accrues on one surface only: one per ballot cast, with buying, selling and transfers minting none (VOTE_MINT, contracts/logic/src/lib.rs). So capital moves only one factor of the product – a buyer who takes the whole float still has v = 0, and therefore zero weight, until they start casting ballots at one per proposal per account. That is not a defence, it is a different bill: the attacker has to keep showing up as a member rather than pay once. It is not the flat one-token-one-vote line BonkDAO's attacker paid roughly $4M to cross.

The second phase is the one BonkDAO had no analogue for. A ballot that carries is only legislation. A proposal offers two to five ranked options, always including a “do nothing”; the leading option passes only if it takes 1.5 / option_count of the weight actually cast – a tally with no weight behind it fails outright – and is not the “do nothing”. That earns the right to trigger, nothing more. Triggering locks the caper's trailing average token price as a baseline and opens the market window, and the action executes only if the token's time-weighted average price across that window is at or above the baseline. An attacker who buys a majority and passes a drain then has to hold the price up for the whole market window against everyone selling into it. See Voting and Proposals; the redemption path, which pays out at that same weight, in leaving a caper.