Three pump.fun coins, one playbook: a wallet pays for the whole launch, then takes the SOL back

Update, 7 October 2026, 5:37 PM ET: the $SNDWITCH setup had a second payout. At 4:59:16 PM ET the coin's creator wallet, 7MaW…axev, collected 413.7 SOL in creator fees, the cut pump.fun pays a coin's creator on its trading, and 24 seconds later sent 414.0 SOL to CMvD…DMjN, the wallet that funded the launch. In the same minute, other wallets, including launch and seller wallets from this report, swept smaller amounts to it. At 5:18 PM ET, after a 1 SOL test, CMvD…DMjN sent 421.9 SOL to a new wallet, 5Pof…Jc8H, and was left empty. Within the same 92 seconds, three other wallets sent 5Pof…Jc8H SOL the same way, a 1 SOL test and then the rest; it held 791.4 SOL at 5:37 PM ET. We haven't checked whether those three are tied to these launches. Counting the fees, at least 770 SOL from $SNDWITCH reached CMvD…DMjN, against about 89 SOL it paid out. The $swordinu and $A1 creator wallets had collected less than 0.01 SOL each in creator fees by 5:35 PM ET.

Most pump.fun coins start on a bonding curve: a launch pool that sells about 793 million of the 1 billion tokens at a rising price, and moves the coin to a PumpSwap pool once it is sold out. That usually takes minutes to days. On 6 October, three coins sold out the curve in the same second they were created, so they reached PumpSwap with almost all of the tradable supply in a handful of wallets. We read each launch on-chain to see whose wallets those were.

$swordinu ("inu wif sword") was created at 2:33:57 PM ET. In that second, one wallet, 8CUb…WTHM, sent 21.31 SOL to 9o7w…P5o, 65.75 SOL to Rtq1…L2P and 1.11 SOL to the creator. The creator bought 34.2 million tokens as it created the coin, and the other two signed one buy together for 758.9 million; with the creator's tokens, that was 79.3% of supply. All three sold everything by 4:02 PM ET for 2,152 SOL and sent at least 2,165 SOL back to 8CUb…WTHM. None of them touched the coin again.

$A1 (AIOPAD) went the same way 14 minutes later. At 2:47:55 PM ET, 53RF…wfw6 sent 91.0 SOL to a wallet, BPe3…YB29, and 0.11 SOL to the coin's creator. One second later the creator made the coin and BPe3…YB29 bought 792.7 million tokens, the rest of the curve, for 86.06 SOL. Its first sell came 21 seconds after that. Over 454 sells it sold the lot by 4:43:58 PM ET for 2,091.8 SOL, and by 4:43:59 PM ET had sent 2,096.8 SOL to 53RF…wfw6, the wallet that funded it. (The next morning the same two wallets briefly traded A1 again, a 1.15 million-token round trip; it isn't counted here.)

$SNDWITCH (Sand Witch Kitten) used more wallets. At 3:37:05 PM ET, CMvD…DMjN sent SOL to seven brand-new wallets, 0.30 SOL to the future creator and 1.1 to 45.6 SOL to six others, 88.8 SOL in all. Thirteen seconds later, in one block, the creator made the coin, the six wallets bought the curve in two transactions and the creator moved it to PumpSwap. Five of the six buyers and the creator sold 757 million tokens for 307.2 SOL, most of it within seven minutes, and sent 309.8 SOL to CMvD…DMjN. The sixth wallet didn't sell. In transactions CMvD…DMjN co-signed, it handed its 33.8 million tokens out to 18 other wallets, 1.1 to 2.6 million each. Each of those got 0.01 SOL from CMvD…DMjN for fees, sold its tokens within 25 minutes and sent the SOL back to CMvD…DMjN, at least 47.2 SOL in all. One of them, 5LKB…89H, received its tokens at 3:42:25 PM ET, sold them a minute later for 4.94 SOL and sent the SOL to CMvD…DMjN three seconds after that. Thirteen of the 18 were new wallets; the other five had run the same receive, sell and send-back cycle on other coins earlier that day. By 4:03 PM ET, 26 minutes after launch, CMvD…DMjN had at least 356 SOL back on about 89 SOL paid out.

Why do it this way? We can't see intent on-chain, but the trades suggest what it buys. Whoever buys the whole curve in the launch second pays the curve's lowest prices, before anyone else can buy, and the coin lands on PumpSwap already graduated, where it shows up on lists of new pairs. Funding fresh wallets seconds earlier keeps the buyers from looking like one party at a glance. On SNDWITCH, splitting the sale across 18 small wallets would make it look like many small sellers rather than one; that is a possible reason, not something the chain shows.

The buyers who came later did not buy from these wallets. On swordinu and A1, the funded wallets had sold out by 4:02 and 4:43 PM ET on 6 October. swordinu's hourly high at that point was about $1.1 million; it was about $20.3 million at 3:27 PM ET on 7 October, per DexScreener. A1's hourly highs while its buyer sold were about $313,000 to $576,000; it later traded far higher, with an hourly high of about $1.86 million in the 7 AM ET hour on 7 October, per GeckoTerminal, and about $1.38 million at 3:27 PM ET per DexScreener. SNDWITCH's later high was about $1.26 million in the 5 AM ET hour on 7 October.

Not every coin that graduates in its launch second fits this. On $EVERYTHING, launched at 3:02 AM ET on 7 October, eight wallets bought the curve in its launch second, but they were funded by different wallets, one of them an exchange-style wallet that pays many users. That is a coordinated launch too, but not a single funding wallet getting its SOL back. None of $DARWIN, $TWEETCRAFT or $RARI, which we also checked, sold out its curve in the launch second.

How to check a coin yourself: on Jupiter, compare the coin's created and graduated times. If they are the same second, look at the first block's buyers on Solscan, then each buyer's first transaction. If they were all funded by one wallet seconds before launch, the 79% at launch was one party's, whatever the holder count says later.