How the whole thing is meant to work.
One document covering the token, the node network, how contribution is proved, how rewards are released, and how liquidity would be funded. Written to be argued with: every figure comes from a published model and the assumptions are stated rather than buried.
What exists today
- The $ARCU token contract, deployed and verified
- 2,000,000,000 minted at genesis, held by the treasury
- A cap of 20,000,000,000 that cannot be raised
- No address holds the minting role, so the rest cannot be created yet
What does not
- Emission — the contract that would release the remaining supply
- Node licences, and any sale of them
- The node network, rewards and reward claiming
- Any market, exchange listing or liquidity pool
Twenty billion, and no more.
The cap is set once in the constructor of the deployed contract. There is no function anywhere that can raise it, no pause, and no upgrade path — which you can check in the verified source rather than take from this page.
Emission is proposed to taper by roughly a quarter each year and never reach zero, so the total approaches the cap without ever arriving. The pre-mint costs node operators nothing: 18 billion at two thirds is the same 12 billion lifetime node pool as 20 billion at 60%. It only converts the treasury's slow drip into something available on day one.
| Year | Emission | To nodes | Cumulative | % of cap |
|---|---|---|---|---|
| 1 | 4,500,000,000 | 3,000,000,000 | 6,500,000,000 | 32.50% |
| 2 | 3,600,000,000 | 2,400,000,000 | 10,100,000,000 | 50.50% |
| 3 | 2,700,000,000 | 1,800,000,000 | 12,800,000,000 | 64.00% |
| 5 | 1,350,000,000 | 900,000,000 | 15,950,000,000 | 79.75% |
| 10 | 320,361,328 | 213,574,219 | 19,038,916,016 | 95.19% |
| 15 | 76,023,245 | 50,682,163 | 19,771,930,265 | 98.86% |
Proposed schedule, illustrative. Emission is not deployed.
A fixed number, priced in steps.
A Founder Node licence is the right to one reward-eligible node slot. It is not a share in anything, not a loan, not a dividend and not an entitlement to revenue. A maximum of 12,000 is proposed: 11,500 sellable and 500 reserved for prizes and giveaways, with the two pools unable to borrow from each other. Licences cannot be transferred between wallets until the sale completes — the price rises in steps, and a secondary market running alongside the sale would let early buyers undercut it. The lock lifts automatically on the final sale mint, and can be released early only if the sale stalls short of 11,500.
| Sold | Price at that point | Cumulative raise |
|---|---|---|
| 100 | $60 | $6,000 |
| 1,000 | $150 | $105,000 |
| 2,000 | $650 | $530,000 |
| 5,000 | $2,150 | $4,805,000 |
| 10,000 | $7,150 | $28,305,000 |
Price is a function of how many have been sold, not of who is buying — so the ladder cannot be quietly restarted. It is enforced by reading the licence contract's own issued count rather than a number the sale contract keeps for itself. Illustrative structure, not an offer.
Nodes distribute the games.
A node keeps a share of the network's game content — builds, updates, assets — on a machine the operator already owns, and serves it to players while that machine is on. Every byte it delivers is a byte that did not have to come from ArcadeUnion's own servers.
Why an hour a day is useful
Content is spread with many copies of each piece. With sixteen copies, a network of machines online one hour a day still answers roughly half of all requests without touching origin; three hours answers closer to nine in ten. Copies are affordable precisely because these are ordinary machines rather than rented servers — spare disk costs nothing.
Why there is no minimum
An uptime bar exists to make availability predictable. This job does not need predictability, because a request that misses falls back to ArcadeUnion's own copy. So there is no bar, and nothing to fake: an hour a day earns roughly a twelfth of what twelve hours earns, because it delivers roughly a twelfth as much.
Neither half can be faked by the node.
That the data is held
A node is asked at random to return a hash of a whole piece, salted with a number issued at the moment of asking. It cannot be precomputed and answering needs every byte to hand. Which piece a node holds is not chosen by anyone — it follows from published data, so the assignment can be recalculated and checked by anyone who wants to.
That the bytes arrived
A node claiming to have delivered ten gigabytes is making a claim, and paying on claims means paying for the most convincing liar. So whoever received the bytes signs for them, and the node is paid on a signature it cannot produce. Receipts are cumulative, so a whole session is one signature and an old one re-sent achieves nothing.
Scoring happens off the chain; the contract settles a cryptographic summary of each period. The full data behind every period is published, so anyone can rebuild the summary and check it — a total nobody can reproduce is not transparency.
Where the pool would come from.
This is the part most projects leave vague, so it is set out here in full — including the constraint that actually binds.
A pool has two sides, and only one of them is easy
The 2,000,000,000 minted at genesis is the token side of a liquidity pool. That side is not the problem. The other side is dollars, and the treasury's only dollars are licence revenue.
Liquidity depth is therefore limited by dollars, not by tokens. Holding two billion tokens does not create a deep market; pairing them against the dollars available does. This is why the opening price is not a valuation — it is an arithmetic choice about how many tokens to pair against the money in hand.
What that choice looks like
At 2,000 licences sold — about $510,000 of cumulative revenue — the options would be roughly:
| Opening price | Tokens in the pool | Effect of a month of released rewards, a tenth of it sold |
|---|---|---|
| $0.0005 | 1,020,000,000 | −3.1% |
| $0.0010 | 510,000,000 | −6.1% |
| $0.0050 | 102,000,000 | −25.8% |
| $0.0100 | 51,000,000 | −42.8% |
A higher opening price means fewer tokens in the pool for the same dollars, which means a thinner market and a larger move from the same selling. The figures above are what a model produces from stated assumptions, not a plan, a target or a prediction. No opening price has been chosen and no listing date has been set.
Why rewards are held back at first
Node rewards are proposed to accrue as a claimable allowance from the start but not to become claimable for three months, and then to release evenly over six months rather than all at once. The reason is this pool. A cliff releasing everything on one day puts a wall of tokens in front of a market that cannot absorb it; the same amount spread across six months is a flow it can.
Modelled, the daily release is around 180 times smaller than the hard-cliff alternative — roughly 0.7% of the accrued balance per day instead of all of it on one.
Rewarding play without paying by the hour.
Participation should be rewarded: a network of games with nothing in it for the people playing them is not much of a community. The question is what to reward, because the obvious answer does not survive the arithmetic.
Why hours cannot be the unit
A pool divided by activity settles where one more unattended account earns about what it costs to run. Work that through and a real player’s share turns out to depend only on what an account costs its operator and how much more activity it can produce unattended — not on how large the pool is. Modelling year-one emission, a twentyfold swing in pool size and token price left a player’s share unchanged. Paying by the hour funds automation rather than players.
What is proposed instead
Rewards attach to outcomes and milestones — seasons, tournaments, ranked placement, first completion — rather than to time logged. Anything funded from emission is proposed to carry a per-account ceiling, daily and lifetime, so there is a limit on what one identity can draw and nothing that improves by being left running overnight.
Competitions that mostly fund themselves
Where there is a prize pool, it is proposed to come mainly from entry, with emission added as a visible top-up rather than as the pool itself. Entry money returned to entrants pays nobody for merely turning up: an automated entrant has to genuinely outplay the field, and pays for the attempt. The top-up is the only part that can be farmed, and it is a figure that gets set rather than discovered. A season can also ask entrants for a history on the network instead of a fee — a gate paid in time rather than money. How a board is run and checked →
Identity has to cost something
An account that costs nothing to create can be held in unlimited numbers, and no measure downstream of that recovers. Anything emission-funded is proposed to sit behind something a player was paying anyway — the game itself, or a season pass — so that holding ten thousand accounts costs ten thousand times as much as holding one.
None of this proves a person is playing, and nothing can. What it does is make pretending cost more than it pays — the same approach taken on the node side, where storage a node must genuinely hold is what stops one machine presenting itself as twenty. Player rewards are also proposed to accrue as a claimable allowance rather than a nightly transfer, which leaves a window in which an allowance can be withdrawn before it is claimed, under the same published appeals route that covers operators.
What can be checked, and what cannot.
Checkable by anyone
- Every token and dollar movement, on a public chain
- Game purchases, which are proposed to be paid in USDC or $ARCU rather than by card, so revenue arrives on-chain
- The supply cap and the absence of a pause function, in the verified source
- Which node should hold which piece of content, recalculable from published data
- Every reward period's full allocation data, so the published summary can be rebuilt
Stated, not proved
- What off-chain spending was actually used for — a payment leaving an operating wallet is visible, the invoice behind it is not
- Anything ArcadeUnion reports about its own off-chain costs
- Whichever party runs the coordinator — a real centralisation, disclosed rather than hidden
Where a figure is a statement rather than a proof, it is proposed to be labelled as one. A wrongly excluded operator would have a published appeals route with deadlines binding both ways, and corrections are made upward in the following period — which is also why the reward engine is designed to under-allocate when it is uncertain, since an under-payment can be fixed and an over-payment cannot.
What is unsolved.
Listed because a whitepaper that only describes what works is a brochure.
The game client has to sign for what it receives
Nothing is paid for delivery until the receiving software signs. This is the piece the reward model rests on and it is the piece that still has to be built into the games.
Nothing yet proves a receiver is a real player
What limits abuse today is arithmetic rather than detection: each receiving session is capped, so a forged one is worth a bounded amount. Tying sessions to a real player account is the actual fix.
One party, one key
Assignment is recalculable and every period's data is published, which limits how far the coordinator has to be trusted. None of it removes the need.
The network is distribution, not storage of record
ArcadeUnion keeps the authoritative copy of everything. A node network can be a good way to distribute files long before it is a trustworthy place to keep the only copy.