Invitations
An invitation you get paid for
We are not building a pyramid and we do not sell seats in one. What gets paid is a share of the real memory operations of the people you brought in — across three levels and not one deeper. If a level is empty, the share does not settle with us: it goes into burning the token. Company partners get two more channels on top of the on-chain one — fiat and a $GALATIN buyback. Every payout is called a Network Validation Fee: it is paid because the network was proven by work, not because a form was signed.
Why this is not selling seats
The difference between an affiliate network and a pyramid is not rhetoric. It is where the money comes from.
A fee for validating the network, not for a signature
Every partner payout in the ecosystem is called a Network Validation Fee. The name is not decoration — it describes precisely what the money is for. An ambassador brings a person in, that person starts using memory, each of their operations passes through the blockchain, and part of the fee on that operation returns to whoever brought them. The network was proven by work, not by a promise.
That produces a property most affiliate programs would rather not have: signing someone up pays nothing. Nothing at all. You can bring in a thousand people, open a thousand accounts and receive not a single token if those thousand never store a conversation. The accrual is tied to an operation, and the operation is tied to use.
It also produces a pleasant consequence: the income does not stop the moment you stop recruiting. As long as the people you brought keep using memory, the share keeps accruing. That is a fundamentally different dynamic from schemes where the base has to be fed constantly or the structure folds.
A smart contract does the arithmetic, not an accounting team
The router split lives in a contract on Solana. The percentages are not negotiated with an account manager, do not depend on a quarterly plan and are not revised unilaterally by an email announcement. Level one cannot be quietly moved from fifteen per cent to twelve because someone decided it looked better that way.
On-chain accruals appear in your cabinet as a list of operations: which operation, whose level, how much was credited. Every line can be checked against a block explorer. We are not asking you to trust our table — we are asking you to verify it.
There is a flip side, and it is fairer to say it now: if nothing was credited because no operation happened, no amount of correspondence with support will change that. The contract cannot make exceptions, which is at once its virtue and its hardness.
An empty level goes to the burn
The quietest trick in affiliate programs works like this: a user has no inviter, so the percentages that would have gone upward simply stay with the company. Formally nobody is cheated; in practice the company earns from emptiness and therefore has an interest in levels staying unfilled.
Here that possibility is closed structurally. If a level has nobody on it, its share does not flow into the Founder's Fund or the treasury — it goes straight into the burn. The company receives not one token from an empty level, which is exactly why it is indifferent to whether the level is filled.
The arithmetic is simple. With a completely empty grid, the burn takes the baseline five per cent plus fifteen, seven and three — thirty per cent of the operation. That is the ceiling: more than thirty per cent of a single operation cannot be burned, because the remaining shares are committed to development and permanent storage.
Three levels, and not one deeper
Depth is not a question of generosity. The more levels there are, the less each one holds and the further the meaning drifts: at level ten a participant knows nobody in their own branch and cannot be useful to it. Partnership degenerates into a headcount, and a headcount into an income promise nobody intends to keep.
Three levels is the range of real acquaintance. You know the people you brought. You have probably heard of the people they brought. Level three is already the boundary at which the connection turns formal. The grid stops there for that reason, not to save money.
The side effect is agreeable: the shorter the chain, the more stays at level one. Fifteen per cent on L1 is noticeably above what eight- and ten-level programs pay, where the first level is usually left with three to five.
Why the payouts carry that name
Thirty years of badly ending stories have loaded the word referral so heavily that it triggers distrust before the sentence is finished. We are not hiding behind terminology — we are trying to describe the mechanics accurately.
Validation here means literally what it means in a blockchain: confirmation of useful work. A node brings in a participant, the participant performs operations, the operations confirm that the node works. The fee is paid for confirmed work and for nothing else.
Hence a practical distinction worth remembering: in this program you cannot claim a good position and wait. A position with no operations behind it pays nothing, and no place in the grid changes that.
A partnership stops being a partnership the moment money starts arriving for the invitation rather than for the value.
— Maksim Valentinovich Galatin, Architect of the CODE ecosystem
Worked examples with the arithmetic
Calculated to the last cent. This is not an income forecast and not a promise: it is a demonstration of how the formula behaves on concrete numbers.
On-chain: how one operation divides
Say a memory operation costs one thousand units of $GALATIN and the user has all three levels filled. The router splits it like this: fifteen per cent, one hundred and fifty units, goes to level one; seven per cent, seventy units, to level two; three per cent, thirty units, to level three.
Then: fifty units to the Founder's Fund, fifty to the burn, six hundred and fifty to the treasury for AR purchases. Add them up: 150 + 70 + 30 + 50 + 50 + 650 = 1000. No remainder, no hidden fee, and no room in the design for one.
Note the proportion. The entire partner grid accounts for twenty-five per cent of the operation, while permanent storage takes sixty-five. The invitation program is not the main cost here — it is the incidental one.
What happens when a level is missing
Same thousand units, but the user has no second and third level: their inviter arrived alone, with no chain above. The seventy units of level two and the thirty of level three have nobody to go to. They do not stay with the company and are not shifted into the treasury — they are added to the burn.
Result: one hundred and fifty to level one, fifty to the Fund, six hundred and fifty to the treasury and one hundred and fifty burned instead of fifty. Fifteen per cent of the operation is destroyed instead of five — precisely the amount other programs would call an unallocated remainder.
The extreme case is a user who arrived with no inviter at all. Then the burn takes the baseline five per cent plus fifteen, seven and three: three hundred units out of a thousand, thirty per cent. That is the ceiling and it will never rise, because the remaining seventy per cent is committed to development and to Arweave.
Fiat: one month for one company partner
An Ambassador Team with one client at level one on Family Archive at $100 a month, one at level two on Spark at $15, and one more at level three on Spark at $15. The fiat channel pays 7 / 3 / 1 per cent.
The math. Level one: 7 % of 100 is $7.00. Level two: 3 % of 15 is $0.45. Level three: 1 % of 15 is $0.15. Total for the month: $7.60.
The number is deliberately unimpressive. We show it exactly like that so nobody builds a plan on a single client: the fiat channel makes sense where a partner already has an audience and the count runs into dozens or hundreds of connections, not units.
The same month, paid in the token
The partner picks the second channel — payout in $GALATIN on the AIfa Yield Dashboard. The rates are higher: 8 / 4 / 2 per cent. Level one: 8 % of 100 is $8.00 worth of the token. Level two: 4 % of 15 is $0.60. Level three: 2 % of 15 is $0.30. Total: $8.90.
The difference against fiat is $1.30 a month, about seventeen per cent more. The platform buys the token on the open market for exactly the reward amount, so each such payout becomes a purchase of $GALATIN rather than a release of new tokens.
And the honest caveat, without which this paragraph would be advertising: the token carries a market price, and a market price moves in both directions. The higher rate compensates for that risk; it is not a gift. The choice of channel stays with the partner.
The tier alignment rule
Referral income is calculated against your own plan, not against your invitee's plan. The rule exists so that nobody can hold the smallest subscription and collect percentages from other people's larger ones: whoever takes a share of a level has to stand on that level themselves.
An example. You are on Spark at $15 a month and you brought a client onto Family Archive at $100 a month. The full level-one fiat rate is 7 %, which would be $7.00. What is actually credited is 7 % of your own plan: 7 % of 15, or $1.05.
The $5.95 difference does not vanish silently. It is shown in your cabinet as a separate line called lost opportunity. You see the exact amount you did not receive this month, and which referral it came from.
Whether upgrading is worth it
Continue the same calculation. Moving from the entry subscription to Family Archive costs $85 more per month. One such referral closes $5.95 of that gap — so a single one does not come anywhere near paying for the upgrade.
To cover the difference on lost opportunity alone you would need roughly fifteen referrals of that kind. We write it out plainly because the cabinet shows lost opportunity in large type, and the temptation to read it as an invitation to upgrade is strong. It means exactly one thing: this is what you did not receive. Not what you will receive.
The correct order of operations is the reverse. First the plan has to make sense for you — memory volume, number of devices, family access. Unlocking the full referral rate is a pleasant consequence, not a reason. An upgrade applies from your referrals' next payments; nothing is recalculated retroactively.
One-off payments and devices
Digital DNA is structured differently from the rest: $1 000 once per device, then $200 a month. The fiat channel counts the one-off part too: 7 % of a thousand is $70.00 credited to level one in a single go, and then 7 % of two hundred, or $14.00, every month.
The tier alignment rule applies here as well, and here it bites hardest. A partner on an entry plan who brings a client onto that tier will see the largest lost opportunity figure anywhere in their grid — and this is precisely the case where an upgrade is decided with a calculator rather than with feelings.
One clarification worth stating: the one-off part is charged per device, not per person. A company connecting a secured perimeter across several machines pays for each of them, and the partner percentage is calculated from each.
When a level has no referrals on it, the undistributed percentage is directed straight into the burn.
— CODE ecosystem Constitution, section on the Solana router