Weighted draws over a transfer-log holder book, funded by bonding-curve commission, settled in native ETH.
We describe a protocol that pays a single token holder every second, indefinitely, without staking, claiming or registration. Each second a winner is drawn from a holder book reconstructed from the token's ERC-20 transfer log, with probability equal to the holder's share of eligible supply. Payouts are funded exclusively by the commission the bonding curve already charges on trades; no tokens are minted and no treasury is inflated. We give the cost arithmetic that makes a per-second cadence tractable, describe the settlement path that makes double payment impossible under crash and network failure, and specify the adaptive cadence that keeps the protocol solvent when fee income falls below the cost of its own gas.
Most token distribution schemes pay rarely and in bulk: a weekly claim, a monthly airdrop, a leaderboard settled at some epoch boundary. Each of those requires the holder to do something, and each concentrates the payout into a moment that can be gamed by arriving just before it and leaving just after.
Paying continuously removes both problems. If a winner is drawn every second, there is no boundary to arrive before, no snapshot to be present for, and nothing for the holder to sign. The only reason this is not done is arithmetic: on most chains, 86,400 transactions a day costs more than the fees that would fund them.
Let B be the set of eligible holders at time t
and b(a) the balance of address a. Each tick,
one address is drawn with probability
The draw is independent each second. There is no cooldown, no exclusion of previous winners, and no memory. The same address may win twice consecutively; an address that has never won is owed nothing.
This is deliberately not a leaderboard. Over n draws, the
expected share of wins for a holder converges on their share of supply
with standard error proportional to 1/√n. At 86,400
draws a day, a holder with one per cent of supply expects roughly 864
wins a day with a standard deviation near 29, a spread of about
three per cent. The scheme is therefore fair in the only sense that
matters operationally: it does not require trusting anyone's judgement
about who deserves what.
The holder book is not a snapshot and is not maintained by an indexer.
It is reconstructed from the token's ERC-20 Transfer log
from block zero, applying every event in order. A wallet becomes
eligible in the block it receives tokens and stops being eligible in the
block it sends its last token away.
A snapshot taken every n seconds necessarily misses everything between snapshots, and the size of what it misses is invisible. The log has no such gap: it is the definitive record of every balance change, each one stamped with the block it occurred in. Replaying it is deterministic, so a restarted engine reconstructs a byte-identical book with no state to reconcile.
Addresses that are machinery rather than holders are removed from the book: the bonding curve, the router, the wrapped-ether contract, the launch factory and locker, the position manager, the protocol fee recipient, the deployer, and the payout wallet itself. Without these exclusions the curve alone, which holds the entire unsold supply, would win almost every draw.
Payouts are funded by the creator's share of the trading fee. On the
Pons V2 bonding curve the creator share accrues inside the curve
contract and is claimed by calling sweepFees, which pays out
in native ETH and cannot touch the trading reserve. Claimable value at
any moment is
With the protocol share at 3,000 basis points, the creator retains seventy per cent of the base fee. Nothing is minted, and the pot is exactly what the market paid in.
A native transfer costs 21,000 gas. Robinhood Chain produces a block every 100 milliseconds and, measured over ten samples at the time of writing, prices gas at 0.0436 gwei. One payout therefore costs 0.00000092 ETH, and a full day of per-second payouts costs 0.0790 ETH.
Two properties matter beyond the headline number. First, block capacity is not a constraint: blocks presently carry around seven transactions against a gas limit that would admit tens of billions of transfers, so a cadence of one payout per second occupies a tenth of one block. Second, 100 millisecond blocks mean a payout confirms an order of magnitude faster than the interval it belongs to, so each second's payout settles well inside that second rather than bleeding into the next.
Break-even follows directly. At a one per cent pool fee with a seventy per cent creator share, gas at the measured rate is covered by roughly 11.3 ETH of daily trading volume.
The failure that matters is not a missed payment but a duplicated one. A payment is duplicated when the engine believes a transaction failed that in fact landed, which is exactly what a lost HTTP response looks like.
Every payout is signed locally first. Signing yields the transaction hash before the network has heard of it, so the hash is written to a durable journal before the broadcast. A crash at any point afterwards leaves a record that can be checked against the chain.
A journaled payment is resolved only by a definitive answer from the chain. A receipt showing success applies the payment; a receipt showing revert counts a strike against the recipient. Anything else (a timeout, a throttled RPC, a receipt not yet found) leaves the payment pending and rebroadcasts the identical signed bytes, which is idempotent because the hash is unchanged. A payment is declared dead only when the account nonce has advanced past it with no receipt.
A contract without a payable fallback can never accept a plain transfer. After three consecutive reverts such an address is written off: it is excluded from subsequent draws rather than absorbing a share of every pot forever.
Gas is paid from the same pot as the payouts. When fee income falls below the cost of a full per-second schedule, the interval stretches rather than the pot going negative. The rate published on the site is the measured current rate, not the target.
This makes the protocol solvent by construction. A quiet day produces a slower cadence and an honest number on the page; it does not produce a deficit, a pause, or a silent stop.
Pons V2 factory
0x7ed598bcEf8bd9edd8c97a195c6d13F40801Ec7E
RPC
https://rpc.mainnet.chain.robinhood.com
Explorer
https://robinhoodchain.blockscout.com
Claim
sweepFees(uint256 minBuybackTokensOut) on the token's curve