The DAO decides: who votes on the token price

A proposal lands on the network’s board one evening. A group of members who run racks in a cold building with cheap electricity want the price per token of compute raised for the class of work they mostly serve. Their argument is sound: newer hardware, better uptime, and a rate set when the network was smaller.

Across the same network, a member who runs a small design studio reads the proposal on her phone. She does not own a rack. She rents compute every week for rendering jobs and pays in the tokens her shelf node earns overnight. For her the proposal is a cost increase with no benefit she can see, written by people she will never meet.

Both of them have a vote. The question is what that vote is worth, and what it should be allowed to decide.

The real cause: a user-owned cloud has two kinds of owner

An ordinary cloud has one party setting prices and everyone else paying them. A cloud owned by the people who use it has suppliers and users inside the same room, and they want different numbers. Suppliers want the price per token high enough to cover hardware, energy and the risk of idle hours. Users want it low enough that renting beats buying. The network needs both to stay, because without suppliers there is nothing to rent, and without users there is nothing to earn. Governance exists to hold that tension without letting either side win outright. This is not an offer, and no token has been issued; what follows is design.

Where governance of a shared cloud goes wrong

Everything is put to a vote. When every rate, admission and dispute goes to the whole membership, people stop reading proposals. Turnout falls, and the few who still vote are the ones with the most to gain. A DAO that votes on everything is governed by whoever has the time.

Votes are weighed by tokens alone. If one token is one vote, then the members who have earned the most, which is to say the largest suppliers, decide the price suppliers are paid. The design studio’s vote rounds to zero. The people the price is charged to end up with no say in it.

Votes are bought. A block of voting power can be borrowed for an afternoon, used, and returned. Or a proposal that benefits one group is dressed as a technical fix and passed before anyone outside that group notices. Any system where a vote is a transferable asset has a market in votes, whether it admits it or not.

The rules live in a document, not in code. A governance page can promise anything. If the settlement software can still be changed by whoever runs the most nodes, the vote is advisory and the real decision is made elsewhere.

Nobody can leave. A vote means little if losing it traps you. A member who disagrees with a price and cannot take her node, her tokens and her work out of the network has not been governed; she has been captured. The homepage promises no exit fee and no lock-in, and that promise is part of the governance, not separate from it.

What should be decided by vote and what should be fixed

Fixed in code, changed only by a slow process. How capacity is measured. How settlement is recorded. That the ledger is readable by every member. That a node lends capacity only when its owner switches that on. These are the things no majority should be able to flip in a week, because they are what makes the network worth joining in the first place. The way a node’s work is checked at all is the subject of how a node proves it computed honestly.

Decided by vote, within bounds set in code. Pricing rules per class of work, admission of new members, how disputes are handled, and what the shared income pays for. The homepage lists these four, and the honest way to run them is as rules with limits: a rate can move, but only by so much per period, so that a single vote cannot double the cost of a studio’s renders overnight.

Two chambers, not one weight. The cleanest answer to the supplier-and-user problem is to count both. A price change should need support from members measured by what they contribute and from members measured by what they rent, so that neither side can set the other’s price alone. The exact mechanism is one of the things still being designed, and it will be published before anyone is asked to rely on it.

Votes that cannot be rented. Weight that depends on time in the network and on capacity actually used, rather than on tokens held at the moment of the vote, is harder to borrow for an afternoon. It is not impossible to game, only expensive, and that cost is a parameter to be chosen openly.

Proposals and tallies in public. Every proposal, every vote and every result visible to every member under pseudonyms, the same way the ledger is. Governance that cannot be audited is an operator with extra steps.

The honest limits of a DAO

A DAO does not make a network fair. It moves the question of who decides from a company to a set of rules, and rules can be badly written. It does not remove the need for people: someone still writes proposals, maintains the software and handles the dispute nobody foresaw. It does not settle the legal question of who is responsible when something goes wrong, and that is one of the reasons tokens and the DAO are still being analysed under EU law before anything goes live. It can be slow. We would rather say all of that than present a vote as a guarantee.

A test you can run today

Pick one service you pay for by usage: a cloud, a utility, a subscription that scales. Find out how its price is set, and write down who gets to change it and what has to happen before the new price applies to you.

Then write down what you would do if it doubled. If the honest answer is “pay it”, you have no vote and no exit, only a bill. If the answer is “leave”, you have an exit but still no vote. The KyberFusion homepage describes a network that is trying to give its members both, and this text is the part of that design we can describe today.