Launch forensics

A token's first hour is where most of the lying happens. This page explains what a staged launch looks like on-chain, why the most-quoted number about it — "% bundled" — is usually measured wrong, and which questions actually separate a legitimate presale from a fleet of sock-puppet wallets.

"Bundled" is two different numbers, and only one of them is real

When a checker says a token is "38% bundled", ask which of these it measured:

Supply currently held by a coordinated group — computed from balances. This number is bounded by 100% and means what you think it means.

Supply ever bought by that group — computed by summing purchases. This number is unbounded. On one token we audited, all buyers together had "bought" 493% of the supply, because the average token changed hands five times. A cohort can have "bought 366%" and hold 4%.

We compute every share from reconstructed balances, never from summed buys. If a number about a token can exceed 100%, it is measuring turnover, not ownership.

Infrastructure fakes coordination

On a modern DEX a single ordinary swap touches several addresses — the pool, a hook, a router. Counted naively, that makes almost every transaction look like a coordinated multi-wallet buy. On the same audited token, "multiple recipients in one transaction" flagged 64% of all activity as bundling; after classifying each recipient and removing infrastructure, the number of transactions that bought for two or more real wallets was zero.

Any bundle claim that has not classified its recipients is counting routers.

Presale or fleet: the two questions that decide it

A launchpad presale and a sock-puppet fleet produce the same concentration chart: a large share of supply spread across many wallets before trading opens. The difference is not in the distribution — it is in who distributed and who received.

Who distributed? If the addresses that spread the supply are contracts, you are looking at launch infrastructure doing its job. If it is a personal wallet fanning out to dozens of receivers, that is a hand on the strings.

Who received? Real participants have history: wallets with thousands of prior transactions and something in them. A manufactured fleet is born yesterday, numbered in sequence, and empty — wallets that exist only to hold this one token.

We ran both questions on a token that looked, by the numbers, heavily staged — 75% of supply moved in the 43 minutes before its pool existed. Every distributing address was a contract; the receivers had transaction counts in the thousands and real balances. It was a presale, and the token traded on. The same questions, asked of a different token, point the other way.

A wallet that looks like a contract

Some personal wallets now carry a small marker of code (account abstraction). Naive checks read that as "a contract holds this" — which sounds safer when the thing being held is the liquidity position. It is not safer. It is a person, able to withdraw at will, wearing a contract's coat.

Our liquidity checks now see through this: a delegated personal wallet is treated as what it is. When you read "LP held by contract — lock unproven" on a card, that phrasing has earned its caution; when it says a personal wallet holds the liquidity, believe it.

What a launch's numbers cannot tell you

• Concentration is not intent. On some chains the median launch allocates a third of supply before the pool opens — that is how the launchpads work. A number is only alarming against the base rate of its chain, and our flags are calibrated against measured base rates, not intuition.
• Early selling is not fraud. Presale holders taking profit in the first hour is a risk — one that has cost buyers dearly — but it is a structural risk, not a scam verdict. The distinction matters because the responses differ: one is timing, the other is avoidance.
• A clean first hour proves nothing on its own. These checks encode staging patterns we have seen. A genuinely novel structure gets one free run before anyone models it.

➡️ See also: The safety stack · Holder concentration