Somebody paid
to put that token
in front of you.

Ranking by market cap gives you the wrong three. In the last seven days 83 pump.fun tokens passed $300,000. Sort them by market cap and the top of the list is a token two hours old showing $1.5 billion — with 100% of its top holders on one exact balance and 0.1% liquidity. Below it: another at $1.46B, 98% identical. Then $936M, 100% identical. Ten of the twelve largest launches of the week are the same pattern.

So we rank after the checks, not before. No snipers and no bundlers among the top holders, under 20% on a shared position size, liquidity at least 4% of market cap, top holders under 35%, over 500 holders, organic score medium or high. Of 83 launches, 4 survive that. The top three of those are the week's real winners — and they are two orders of magnitude smaller than the ones the leaderboard shows.

Then we follow those three wallets. Every token they create, and every fresh token where their address turns up among the largest holders — which is what a buy looks like on-chain. Each of those is held to the same bar before it appears here, and removed the moment it breaks one.

The three are recomputed every half hour, their activity checked every minute. A wallet's past is a fact, not a forecast.

pump.fun has no leaderboard of investors, so we built one from outcomes. We take the tokens that are alive above $1,000,000 today, pull each one's largest holders, and look for the wallets that appear in more than one. A wallet counts as smart money when it holds a position in 2 to 12 of them and at least $1,000,000 across them. Above twelve it is not an investor — it is a pool, an exchange or a burn address, sitting in nearly everything. Then we scan fresh launches and check whether any of those wallets is in the book.

The winners had to be verified first, and most were not. Ranking wallets by what they hold, the top twenty all held exactly $8,466,699 of the same token — to the cent. Those are not investors, they are distribution addresses. So we compared the exact balances of the largest holders of all 212 live pump.fun tokens above $1M: 22% have a manufactured holder base, where dozens of top holders sit on balances identical to the last digit because the tokens were handed out, not bought. AROS 100%, VSOF 100%, WSOS 100%, TNT 100%, USDF 98%. By age the split is stark — older than 30 days, 98% are clean (131 of 134); younger than 30 days, only 45% (35 of 78). Only the clean ones are used to find smart money.

And the fresh token gets the same treatment. No snipers and no bundlers among its top holders, under 20% of them on a shared position size, market cap over $30,000, more than 150 holders, liquidity at least 4% of market cap, top holders under 35%, mint and freeze revoked, and no fall of 50% since we listed it. One candidate had 3,006 holders, 7 smart-money wallets and clean sniper flags — and 39 of its top 50 held exactly 4,900,000 each. It did not make the list.

A wallet's past positions are a fact, not a forecast. Empty most of the time is the expected state.

Who gets watched. We take every pump.fun token trading above $1,000,000 today and look up the account that deployed it. pump.fun publishes each account's follower count and linked X handle, so a person with an audience is distinguishable from a wallet made this morning. An account joins the watch list with 1,000+ followers, or a linked X handle and 300+. Then we watch those wallets and wait for their next launch.

An audience alone is not enough, and this is the part that matters. Of the accounts that cleared the follower bar, several were thrown out on their record: one has 17,812 followers and a token that reached $201M — and sent 9 of its 9 past-$300k tokens to zero. Another has 12,632 followers, a linked X handle and a $168M token, with 8 of 11 to zero. A third, with a linked handle and 1,917 followers, rugged 2 of 2. Followers are not a record. A wallet is only watched if no more than a third of its past $300k tokens ever collapsed.

And the launch itself is re-checked live, every time this page loads. A watched deployer with 3,462 followers and a token that once hit $90.8M also mints launches nobody buys — one holder, two holders. Those are not signals. A launch is shown only with market cap over $30,000, more than 150 holders, liquidity at least 4% of market cap, top holders under 35%, mint and freeze revoked, and no fall of 50% since we listed it. Anything that breaks one of these is removed from the list and from our database.

This list is empty most of the time. These accounts launch a few times a month, not a few times an hour — that is what makes a launch from one of them worth knowing about. A clean record is a fact about the past, not a promise about this token.

The one honest way to be early. For a token under six hours old there is no flow data worth trusting — we measured the organic share at zero across every young launch we checked, including one with 3,090 traders and a single dollar of organic volume. Holder counts, charts and volume at that age tell you nothing. But the deployer's record exists from the second the token is created, and it cannot be faked or bought.

So the record is the only filter here, applied at its strictest. A launch appears only if the wallet behind it has an earlier token — not this one — that passed $300,000, and not one of its past $300k tokens ever fell to zero. One rug in its history, ever, and it never appears here.

How rare that is. In a 24-hour window we graded 1,050 launches against this rule. Four passed. This is not a feed and it will never scroll — it is a short list that changes a few times a day. If it is empty, nobody qualified, and we would rather show you nothing than fill the space.

A clean record is a fact about the past, not a promise about this token. Most launches fall, including these.

What this list is. Every pump.fun token trading above $300,000 right now — around 525 of them — with the wallet that deployed it graded behind it. A token appears only when its deployer has at least one earlier token that passed $300,000 and not a single one of those fell below 5% of its peak. Zero tolerance, not a low rate.

Every token here has had its holder base checked, and one in five does not survive it. We pull the largest holders of each token and compare their exact balances. On a real token no two match — people bought different amounts at different moments, down to the last digit. On a manufactured one dozens are byte-identical, because the holders were handed their tokens rather than buying them. We measured all 212 live pump.fun tokens above $1M: 22% have a manufactured holder base. AROS, VSOF, WSOS and TNT each have 100% of their top holders on one exact balance; USDF 98%. Split by age: older than 30 days 98% clean, younger than 30 days 45% clean. pump.fun's own sniper and bundler flags miss all of it. A token appears here only if under 20% of its top holders share an exact balance.

This measurement was wrong until 23 September 2026, and the correction is why this list exists. The earlier version rounded balances to the nearest 100,000 units before comparing them, which invented matches that were never there. Fourteen large, entirely clean tokens — PAID, JEANPHIL, fone, Jimothy, TOAD, Stamp, NASDUCK among them — scored 22–30% "identical" and were thrown off the list; on exact balances every one of them scores 2%, a single coincidental pair out of fifty. The old method also produced the "61% manufactured" figure published here before. It is 22%. We would rather correct a number in public than keep a comfortable one.

Why this list and not a feed of new launches. We measured it: for tokens under six hours old, the organic-flow share is zero across the board, so there is no way to tell a real launch from a bot swarm. Anything that fills up every minute is filling up with those. This list is the opposite trade — it gives up being first in exchange for being checkable.

Each row shows how far the token sits below its own peak, how old it is, and the deployer's record: how many of their tokens are live above $300k, how many ever passed it, and how many went to zero. That last number is always zero here — that is the entry condition.

Rescanned continuously. A deployer's record is a fact, not a forecast. Most tokens fall, including these.

Why peaks are the wrong measure. We graded every pump.fun deployer active in a 24-hour window — 932 wallets. 202 had launched a token that once passed $1,000,000. That sounds like a track record until you look at what happened next: one wallet had 23 launches, 23 of which passed $1M, and 22 of those went to near zero. Another: 15 of 15. Reaching $1M is not a sign of quality. It is the signature of a professional pump and dump, because that is exactly what a well-run dump looks like on the way up.

Followers are not a record either. These accounts all cleared the audience bar and were still thrown out: 17,812 followers, a $201M token, 9 of 9 past-$300k tokens to zero. 12,632 followers with a linked X handle and a $168M token, 8 of 11 to zero. A linked handle and 1,917 followers, 2 of 2. Every one of them would top any influencer list. None of them is on ours.

What it takes to be watched. A live token above $1,000,000 that this account deployed, an audience — 1,000+ followers, or a linked X handle and 300+ — and no more than a third of its past $300k tokens ever collapsed. Nothing about hype, nothing we can be talked into.

Each row opens the deployer's pump.fun profile so you can check every claim yourself. Rescanned every ten minutes. A deployer's record is a fact, not a forecast.

What this list is. Three gates, all checked live on-chain and on pump.fun. A launch is shown only if it clears every one of them.

1 — pump.fun only. No other launchpad, no unknown pool. Mint authority revoked, freeze authority revoked, top holders under 25%, dev wallet under 5%, market cap between $50,000 and $300,000 — big enough to be real, small enough to still be early — liquidity over $15,000 and at least 5% of market cap, over 150 holders, first pool under 48 hours old.

Who sits at the top. Counting holders is not enough — what matters is who they are. We pull each token's largest holders and read pump.fun's own per-wallet flags. A launch is rejected outright if any of them is a sniper or a bundler, and if 20% or more sit on exactly identical balances, which means the tokens were handed out rather than bought. One launch sitting on this list had 9 snipers and 16 bundlers among its top 50 — half the top of the book — and every other check had passed it.

2 — no bot flow. Jupiter separates organic trading from bot trading. We require at least 4% of the last hour's traders to be organic buyers, at least 3% of volume to be organic, a positive net-buyer count, and under 15 trades per wallet — the wash-trading ceiling. Volume bots fail here.

3 — known deployer. The wallet must clear one of two bars on its live, unbanned pump.fun profile: an earlier token that peaked above $1,000,000, or an audience of 500+ followers. Throwaway wallets — the ones behind almost every rug — carry five to ten followers and no history, and fail both. Promotion can be bought; a past $1M cannot.

The bar does not move, so the list fills with time, not with compromises: a launch that passes stays for 7 days, and every row shows when we listed it and what it has done since. Where we have it, each row also shows the narrative and how many callouts we recorded.

These are chain facts, not a recommendation. Shillproof does not rate tokens as good or bad. Passing all three gates means the obvious ways to be rugged are closed — it does not mean the token will rise.