Tokenomics
What HAVOC is for, what a claim costs, and how an experiment pays out when it closes.
This page describes the intended design. None of it is live: the token is not deployed, nothing burns anything today, and no experiment has closed. The address will be published here when there is one. Treat every number below as a decision, not a commitment.
What HAVOC is for
HAVOC is what a reward costs to take. It is a utility token: not a claim on the company, not a share of anything, and not a promise. The terms say the same in the language a lawyer would use.
It is deliberately not a toll on taking part. You can configure an agent, train it for a week, submit it, and watch it climb the board without ever holding one.
Training and submitting cost nothing
No token, no transaction, no gas. Submitting a run needs a signed-in wallet and nothing else, which is exactly what the site already asks for.
What keeps the verifier from being flooded is a limit rather than a price: one run per level can be queued at a time, so an account can have at most nine in flight and the rest wait their turn.
Claiming burns HAVOC
When an experiment closes and you come to take your share, that claim burns HAVOC. The burn is the cost of converting a position on the board into a payout.
Putting it here rather than on submission is what keeps the whole thing small. A burn per submission would be thousands of transactions, each one needing to be watched, matched and guarded against being spent twice. A burn per claim is one transaction per player, once.
How a burn is recognised
By calling the token's own burn, which every standard burnable token already has. It destroys the tokens, reduces the supply, and emits a transfer to the zero address as it does.
That event carries who burned and how much, and we already know which wallet is claiming, so a claim is matched on the burning address and the amount. There is no bespoke contract in the path and nothing extra to sign.
Where the pool comes from
An experiment closes with a pool behind it, funded from licensing the dataset and from fees the token accrues while the experiment runs. The site already commits 70% of data revenue to buying HAVOC back.
Neither source is other players. The pool is not funded by the people competing for it, so this is not a game where the losers pay the winners.
How the pool is split
By total points earned. Your share is your combined score over the sum of every player's combined score, and the pool divides on those shares.
Combined score means your best verified run on each level, added together. It is the number already at the top of the leaderboard, so a payout is computed from the same figure you have been watching all along rather than from a second, private one.
Why best-per-level, not every run
Counting every verified run would pay for volume. Submitting one decent policy fifty times would earn fifty times the share, and since submitting is free there would be nothing at all standing in the way.
Scoring the best run on each level removes the question. Resubmitting something worse earns nothing, so the only way to raise your share is to raise your ceiling, which is the behavior worth paying for.
The chain
EVM. Your wallet is already the identity here and already signs to prove it, so the token sits on the rail the site is built on rather than adding a second one beside it.
Nothing that decides a score or a share happens on-chain. The board is produced by the verifier, which replays your policy on a seed you never had, and putting that result on a chain would not make it any more true.
What is not settled yet
The token is not deployed. Its address will be published here when it is, and until there is an address on this page there is nothing to buy, hold or burn.
The supply, the burn as a final number, and the date any experiment closes are all still open. They are set before this goes live, not after.
