buys back without announcing an order the market can trade against.
- sets a budget, a price range and a window
- locks the budget in the transaction that creates the buyback
- receives every token bought and the unspent budget, settled by contract
sealed buybacks
the problem a market buyback announces itself. a scheduled buy is the most predictable order on the chain, so bots trade ahead of it and into it, and a holder who wants to sell cannot name a price at all. they take whatever the pool pays at that moment.
what a sealed buyback does the treasury locks a budget onchain. holders lock the tokens they offer and seal the lowest price they would accept. at the close every offer opens at once, the cheapest fill first, and every seller is paid the same price.
the front-runner is not slower. there is no order to read.
¹ TenderClearing.findClearing ·
² offers are encrypted in the seller's browser with batched threshold encryption ·
³ the testnet committee's keys are published on purpose, see honest limits below
why it matters
protocols now route a large share of their revenue into buying their own token back. almost all of it executes the same way: market buys on a schedule, from a known address, into a public pool.
a modified dutch auction tender offer lets shareholders name the price at which they would sell, inside a range the company sets. the company buys at the lowest single price that fills its target, and pays it to everyone it buys from. motorola solutions ran one for up to $2 billion in 2015, marathon petroleum for up to $4 billion in 2021, and wix bought back $1.6 billion of stock at one price of $92. a sealed buyback is that mechanism, onchain, with the offers sealed until the close so nobody can tender against anyone else's number.
figures as reported by the linked sources, 2025 to 2026. they show the size of the flow, not that any of it was front-run; we found no study that measures that.
the problem
a buyback that runs every hour, or every day, from a known address, is an order everyone can see coming. anyone can buy just before it and sell into it, and the treasury pays the difference.
a holder who wants out sells into the pool at whatever it pays right now. there is no way to say "i would sell at 1.25 and not below", so the patient seller and the desperate one get the same deal.
every market buy pushes the price up against the next one. the treasury competes with itself, and the budget buys less with every block it spends in.
the mechanism
the treasury creates the buyback and funds it in the same transaction. no seller ever offers against money that is not already there, and once anyone has offered, the treasury cannot walk away.
each holder encrypts their quantity, the lowest price they would accept and a random salt to the committee, in their own browser, with batched threshold encryption. onchain sits a commitment, the ciphertext hash, and the tokens they offered, locked so the offer is binding.
at the close the operators each publish a decryption share and a threshold of them opens the whole batch together. the committee signs one root over every offer, and the contract checks each opened offer against the commitment its seller posted. an offer that goes missing halts the buyback into refunds instead of moving the price.
offers are sorted by minimum price and the price climbs from the lowest, as long as the budget can still pay every cheaper seller. where the budget runs out, the offers at that price share what is left. every seller filled is paid the same price, never less than they asked, and payouts never exceed the budget.
simulator
pick a book or write your own, set the budget and the ladder, then play the close. every number below is computed by clearTender, the rule the contract runs, in the same integer arithmetic.
six holders spread across the ladder. the budget runs out partway up.
| seller | offer | sold PEALD | paid DUSD | back |
|---|---|---|---|---|
| alicefilled | 200 ≥ 1.00 | 200 | 260.00 | 0 |
| bobfilled | 300 ≥ 1.10 | 300 | 390.00 | 0 |
| carolfilled | 250 ≥ 1.20 | 250 | 325.00 | 0 |
| danpartial | 400 ≥ 1.30 | 19 | 24.70 | 381 |
| erinkept | 150 ≥ 1.50 | 0 | 0.00 | 150 |
| frankkept | 500 ≥ 1.80 | 0 | 0.00 | 500 |
| treasury | budget 1,000.00 | +769 | -999.70 | 0.30 DUSD |
constant-product pool with the 0.3% fee, 20,000 DUSD deep at 1.35; the sandwich front-runs with half the spend. a sealed buyback buys only from holders who chose to sell, at a price they named, and never touches the pool. if holders ask more than the pool, the pool is cheaper, and this shows that too.
happy with the numbers? start this buyback for real on Tempo testnet, with free test tokens.
run this on testnetcompared
| sealed buyback | market buy | scheduled twap | otc desk | |
|---|---|---|---|---|
| the order before it executes | sealed until the close | in the mempool | schedule is public | seen by the desk |
| holders name their own price | every holder | no | no | large holders, privately |
| treasury moves its own price | no, clears off the pool | yes, into the pool | yes, a little each time | pays the desk spread |
| one price for every seller | yes | no | no | negotiated deal by deal |
| allocation anyone can check | recomputed from the opened book | transaction history | transaction history | trust the desk |
| budget committed up front | locked onchain | no | no | by contract |
a private rpc hides a market buy from the public mempool, but not the schedule, the address or the pool it lands in, and a holder still cannot name a price.
who it serves
buys back without announcing an order the market can trade against.
sells at a price they chose, or keeps every token.
approves a budget and a cap, not a trading desk.
use cases
protocols route fees into scheduled market buys. the schedule is public, so the order flow is the most predictable trade on the chain, and anyone can position ahead of it.
the treasury posts a budget and a window. holders who want to sell name their price privately, and the whole book clears once, off the pool, at one price.
a dao votes in public to buy its token back. the vote is the announcement, and the market reprices before the first purchase.
the vote approves a budget and a price cap. what each holder would accept stays sealed until the close, so the cap is the only number the market sees.
large holders sell over weeks through market makers, each sale visible on chain, while the project watches the price absorb it.
the project offers to buy a fixed amount at the close. holders who want out sell into the tender; nobody who wants to stay is affected, and nobody reads anyone else’s offer.
the cryptography
honest limits
a testnet demo with tokens that have no value. PEALD and DUSD come from a faucet. the testnet committee's keys are derivable on purpose, so treat every seal here as a demonstration of the mechanism.
the treasury cannot sell to itself from its own address, and that is all. it could offer through another address, as any seller could. nothing here detects that.
prices sit on a ladder. a minimum is one of at most 256 steps, so the hidden value is one of a known set. sealing stops anyone reading it, not guessing at it.
the reveal root is signed, not verified onchain. the contract trusts a threshold of committee signatures over the opened book rather than checking the decryption shares itself. what it does check is every opened offer against its seller's commitment.
questions
offers are sorted by minimum price. the price climbs from the lowest while the budget can still pay every cheaper seller at it. where the budget runs out, the offers at that price share what is left pro rata. nobody is paid below their minimum and the total paid never exceeds the budget.
no. each offer is encrypted in the seller's browser with batched threshold encryption and opens only when a threshold of operators publish decryption shares at the close. on testnet the committee keys are published, so treat that as a demonstration.
everyone who offered within the cap fills, the price stops at the dearest offer that filled, and the rest of the budget goes back to the treasury in the same settlement.
every offer is backed by the tokens it offers, locked when it is made. an offer that is not fully backed is excluded at the reveal and refunded, so it can never move the price.
it cannot change an offer: each opened offer is checked against the commitment its seller posted, and a seller whose offer was swapped can prove it and halt the buyback into refunds. if an offer goes missing, settlement refuses to run, and after the deadline everyone is refunded.
not fully. the tokens you lock are a visible transfer and bound your quantity from above. you can lock more than you offer to blur it and get the surplus back. your minimum price is what stays sealed.
a private rpc keeps one transaction out of the public mempool. it does not hide a schedule, stop the treasury moving its own pool, or give a holder any way to name a price.
start a buyback with test funds, or sell into one that is open.