Attention as a market mechanism

The original snowball was built around a simple idea: a token can market make itself. Every trade generates fees, those fees accumulate behind the token, and they are eventually spent buying the token back — with everything acquired sent to the dead address. More trading, more fees. More fees, more buybacks. More buybacks, less supply.

A fomoball adds a new input to that mechanism: attention. Every coin launched here has its own snowball, and instead of waiting only on trading activity and a clock, a thesis posted on fomo can fire it.

The important distinction is that this is not one token being bought by every thesis on fomo. Each launch has its own mechanism, and its own vault. Write about coin A and coin A’s snowball activates. Write about coin B and coin B’s does. The attention stays connected to the token receiving it.

There is nothing to configure. Every coin here charges 4% and runs the same engine, so the only thing that distinguishes two of them is the coins themselves.

A thesis is more than a post. It is somebody publicly explaining why they are paying attention to a coin, published in the fomo app. We do not host them.

Normally that conviction ends at the post. Maybe somebody reads it and buys, maybe they share it, maybe nothing happens. A fomoball makes the thesis itself an input: the social layer stops sitting above the market as commentary and becomes a signal the mechanism responds to.

The keeper reads fomo’s feed for every coin launched here. A thesis it has not seen before is recorded against that coin, and on the same pass the coin buys and burns. A thesis is a trigger, not a score: it does not have to out-argue the clock, and it goes straight through the cooldown that would otherwise hold the engine back.

Trading still matters. Fees still fund the mechanism, and the engine can only ever deploy what the coin itself has generated — a thesis does not create capital. What changes is when the capital already there can be activated.

1. Answered onceEach thesis is keyed on a hash of its author, timestamp, text and token, so re-reading the feed can never fire a second snowball for the same post.
2. A backlog is not a signalThe first time we read a coin's feed, everything already there is stored as answered. Somebody may have been writing about it for a week before we started listening, and answering all of that at once would empty the vault into one buy.
3. Stamped after the factA thesis is marked answered only once the buy has confirmed on chain. A crash between the two answers one thesis twice; the reverse would drop it for good, and of those two, the second is the one that breaks the promise on the front page.
4. Gas still bindsA thesis skips the timing gates and nothing else. If the budget is smaller than the gas it would cost to deploy, the coin waits, whoever asked.

Every buy and every sell pays 4%, in the quote asset, in both directions. It is written into the token’s record on chain at the moment it is created and it cannot be changed afterwards. Not by the creator, not by us.

It used to be a setting, one of four, each with its own rate. That made the single most important fact about a coin — what it charges the people trading it — a configuration option, and it meant two coins on this site could not be compared without first comparing their terms. One rate, every coin.

Each launch gets its own wallet, derived at creation, and that wallet is named as the fee recipient in the same transaction that creates the token. It is the only address the fees can be paid to.

That wallet holds exactly one coin’s money. It is not a shared treasury, so one coin’s fees can never be spent on another’s buybacks, and there is no pooled balance for anyone to argue about. The site can sign for it in order to run the engine; it cannot send the money anywhere the engine does not.

A coin may only ever spend what it earned. The budget is the sum of the claims the keeper made for that coin, less what it has already spent buying itself back, capped by what the wallet actually holds above its gas float. It is never simply the balance — that would treat anything sitting in the wallet as fee revenue.

Nobody writes about most coins most of the time, and a coin earning fees in silence should still deploy them. So underneath the trigger there is a floor: the engine scores four readings and adds them up, and at 1.0 it acts.

1. ThesisHow many theses are standing unanswered. One fires a snowball outright; the count is what keeps a coin people are still arguing about hot after the first answer has gone out. 3 unanswered scores 1.0.
2. TimeHow long since the last snowball, against a 20m cadence. The floor that guarantees a coin nobody mentions still gets bought.
3. BudgetHow much has gathered, against what a full-size snowball would deploy. Stops the engine spending dust on gas.
4. DipHow far under its recent average the price is. 12% under scores 1.0, which is what makes it buy weakness rather than strength.

The weights sum to one, which is what makes the threshold mean something: any single term reaching 1.0 is sufficient on its own, and several partial reasons add to the same verdict. Because they add, a large enough dip pulls a snowball forward long before the clock would have called for it, and a flat market with nothing written about it still gets its scheduled run on time.

A 10m cooldown sits underneath as a hard floor on the clock-driven path, so no amount of score can make a coin trade faster than that. A thesis is the one thing allowed past it.

Two limits, and the smaller one wins. The first is the share: what fraction of the gathered balance a single snowball is willing to spend, enlarged when a thesis or a dip is driving it. The second is the impact cap, a ceiling on how far one buy may move the price.

spend = min(
  budget × 0.75 × (1 + 0.5·dip + 0.5·thesis),
  budget,
  reserve × 2%
)

The cap is the one that matters. A buy large enough to move the price several per cent is purchasing its own exit liquidity, so the engine clips the spend rather than sending the order and hoping. When that happens the snowball is recorded as capped and the remainder stays in the vault for next time. Nothing is lost, only delayed.

Read off the engine at build time. If one of these changes in the code it changes here in the same commit, whether or not anybody remembered to edit this page.

It is not free money. The fee comes out of the people trading the coin, and the buyback puts it back into the same market they are trading in. What changes is who ends up holding the supply and when the buying happens, not how much value exists.

A buyback is not a floor. A coin can only ever spend what it has earned in fees. On a quiet coin that is a small number, and a snowball into a falling market is a small buy into a falling market.

A thesis is not a prediction. It is somebody’s opinion, published somewhere we do not moderate, and the engine does not read it — it counts it.

The engine is ours, and that part is trusted. The fee and its recipient are fixed on chain by pons and hold without trusting anyone. The cadence, the weights and the burn itself are executed by a keeper holding the vault keys; pons has no idea any of it exists. Every snowball is on chain, every thesis that caused one links to the transaction it caused, and every input to the decision is published on the coin’s page — so the commitment is checkable even where it is not enforced.