One wallet launched 79 tokens on BNB Chain in 28 days and traded $6 billion against itself. Real outsiders put in about $15,000

01 · The tokenA token called Anthropic, up and gone in 13 hours

On Monday 26 October, Anthropic filed to go public. Within a day a token called Anthropic, ticker ANT, was climbing the BNB Chain trending lists. By Wednesday it had traded over a hundred million dollars. More than five thousand wallets had touched it. It was the second ANT in two days from the same source. An ANTHROPIC and an oANTHROPIC had come before those.

The pool was on PancakeSwap, the largest DEX on BNB Chain. The volume was steady. Too steady. The pool traded almost exactly the same amount every hour for 13 hours, then at 03:01 UTC on 28 October a single sell took nearly every dollar out of it.

The sell was not a liquidity pull. The pool's LP tokens were locked at PinkSale, a service that holds them so the owner cannot withdraw. Rug checkers would have shown the lock. What emptied the pool was a transfer of tokens that should not have existed.

The token had a supply of half a million at launch. Seventy-one billion came out of nowhere in one block. The contract let its owner mint, and the owner used it once, at the end.

02 · The walletSeventy-nine of them

The wallet that sold is 0x34f09a622e7d5959843be498064fdf13ac6ac385. We call it the hub. It had funded ANT's deployer, it held ANT's launch supply, and it took the dump proceeds. Its own trade history, read with Bitquery's BNB Chain data, is a list of sells, each worth six or seven figures, each in a different token.

Every token address ends in 7777. A matching pattern like that comes from grinding contract addresses before deployment, the same way vanity wallet addresses are made. Each token was deployed by a different fresh wallet, and each of those wallets got its first and only funding from the hub seconds before it deployed. Every deployment we checked used exactly the same gas, which means the same bytecode.

The names follow the news. Four tokens carried Anthropic's name in some form around its filing. There was a TSLAB and an RTX5090. A pPOLY and a PREDICT, after the prediction market apps. AXIS launched three times in two days while a real token called AXIS was trending on Base. DOPE, DNA, BIBI and GANG all had namesakes on the trending lists the same week. Five tokens were called 龙虾, lobster. Several were about Binance: CZ, SUPERCZ, and two that translate as Binance emperor and Binance princess.

The whole run is in the audit file linked at the end. The gap between first trade and dump is 24 minutes at the shortest and 30 hours at the longest. The median is eight hours.

03 · The tradersFive and a half thousand wallets that only trade these

Trackers showed each token with about 5,500 traders. That number was the same on token after token, which is the first thing wrong with it. We pulled every trade in the series from Bitquery's DEX trade data and counted, for each wallet, how many of the 79 tokens it had touched.

That is the whole market. One group of 5,502 wallets traded 51 or more of the tokens each. Six of them, picked at random, had traded between 61 and 66 of the series and nothing else at all in the last month. Everyone else, 411 wallets, traded about fifteen thousand dollars between them. The buyers who looked like they were piling into ANT in its first five minutes were the same wallets that had bought the previous token, and the one before that.

The cluster has a shape. The wallets were created in the first week of July and each received a dusting of spam tokens then, which is how mass-made wallets usually look. They do not sign their own trades. Every one of their trades on the tokens we sampled was sent by a single contract, 0xf9ffcdc55376c811ee88c18d1e2d425b4db985dd, deployed on 12 August. The contract holds no USDT. The hub pays for every buy straight into the pool, and the tokens land in the bot wallet. On the way back out, the bot's tokens go into the pool and the USDT returns to the hub.

So the volume is a loop. The hub's USDT goes into a pool, a bot wallet receives the token, the bot sells it back, the hub gets its USDT. Over a month the hub has sent more than eleven billion dollars of USDT into these pools and received the same amount back, within a rounding error. What it has kept is under eight hundred thousand dollars. The exact figures are in the hub table below.

04 · The lockLiquidity locked, supply unlocked

Every pool in the series did the thing rug checkers look for. Within a minute or two of creation, the hub sent the pool's LP tokens to PinkSale's PinkLock contract. We counted 75 such transfers, one per pool. A locked LP means the owner cannot pull the paired USDT out by removing liquidity. It says nothing about whether the owner can print more of the token.

The dump is the same transaction every time. The hub calls transferFrom on the token contract for a number of tokens tens of thousands of times larger than the supply, the pool accepts them, and the pool's USDT comes out. Seventy-six dumps so far have taken $74.3 million. Almost all of that is the hub's own wash money coming home, which is why the hub's net gain is under a million dollars while its dumps add up to seventy-four.

05 · The moneyThree hubs since September, handed on like a baton

The hub did not start with its own money. At the end of September it received just under two million dollars of USDT from a pass-through wallet. That wallet had received almost the same amount minutes earlier from the previous hub, 0xe3b3fcd8c3126eed4615b99d2c1884811035baf9, which ran the same loop for the second half of September. Before it, 0x180f7f35dfeb7a0da941ad44bf602541350b0689 ran the first half. Its seed arrived from a PancakeSwap pool, which means it was the dump of a token we have not traced.

Each handover goes through a fresh wallet that peels off about a thousand dollars to the WBNB pool for gas and sends the rest on. The 5,500 bot wallets never touch USDT. The engine contract never touches USDT. Everything is paid from the hub and returns to the hub.

Two more details place the pieces together. The wallet that deployed the engine contract has had more than half a million incoming transfers in a year and was itself first funded through the PancakeSwap router. And each token sends 200 million units to a wallet at 0x93d4d4cfa8b386b87dfec0382b06b08b2601d1b5 right after launch, which has been active since April and distributes more than it receives. Those tokens are what the bot wallets that only sell, never buy, were selling.

06 · The pointWhat is it for

A rug that takes fifteen thousand dollars from outsiders over 79 tokens is a bad rug. The crew spends gas on about 17,000 trades a day and ties up nearly two million dollars of USDT to make under eight hundred thousand in a month. The chain does not say why. It does show what the loop produces.

Each token exists for a few hours with the three things token trackers rank on: volume in the tens of millions, thousands of distinct traders, and a locked LP. PREDICT, the live one, sat inside the top twenty of our own trending list for two-day volume on BNB Chain when we checked, with the same five and a half thousand traders as every token before it. Any screener that sorts by those columns will surface these tokens. Any newcomer who buys one is buying from the loop, and the loop ends in a mint.

Three readings fit. The series is bait, and the catch has been small so far. The series is a demonstration or a rental of a volume service, run on the crew's own tokens between paying jobs. Or the series farms something off-chain that rewards volume, trader counts or holder counts on BNB Chain. We cannot pick between them from the ledger, and we do not try.

07 · How we checkedMethod and limits

We started from the trending list and ANT's hourly volume, then followed the wallet that sold. Its trade history gave the token list. Each token's deployer, pool, first trades, LP lock, and dump came from Bitquery's BNB Chain transfers and DEX trade tables. The trader cluster is a SQL count over every trade in the 79 tokens since 29 September, grouping wallets by how many of the tokens they traded. The hub's USDT totals are its complete transfer history for that token. Address roles come from the address labels API, which identified PinkLock, the PancakeSwap router and pools, and an MEXC hot wallet behind one holder.

08 · The addressesWhere to look

FAQ

What is the 7777 token series on BNB Chain?

A run of seventy-nine tokens launched over four weeks in October by one BNB Chain wallet, every contract address ending in 7777. Each gets a PancakeSwap pool, a PinkSale liquidity lock, about a day of trading from a cluster of about five and a half thousand bot wallets, and then a mint-and-dump by the wallet that launched it.

Was the Anthropic token on BNB Chain real?

No. Three tokens with Anthropic in the name came from this series around Anthropic's public filing. None had any link to the company. The second ANT traded over a hundred million dollars, then lost everything in a single transaction in the early hours of the next day.

If the liquidity was locked, how did the pool get drained?

The lock holds the LP tokens, which stops the owner from withdrawing the paired USDT by removing liquidity. It does not stop the owner from minting more of the token. The hub minted about 71 billion ANT against a launch supply of 500,000 and sold them into the pool, which paid out its USDT.

How much did outsiders lose?

Wallets outside the bot cluster traded about fifteen thousand dollars across the whole series. Nearly all of the six billion dollars of volume was the same cluster of wallets buying from and selling to each other with the hub's money.

Is PREDICT the same scheme?

Yes. PREDICT, at 0x79ad05d307b465da8d0b4f1f1f6674dc7c097777, launched on 28 October from a deployer the hub funded, locked its LP at PinkLock, and was trading with the same cluster of wallets when we checked. It had not been dumped yet.

How can I spot a token like this?

Look at distinct traders over time rather than volume. A trader count that sits at the same figure on token after token is a cluster. Check whether the biggest early buyers have traded other tokens from the same deployer family. And treat a liquidity lock as one check among several, never as proof the supply is fixed.