▶ THE PROOF — WALLETS THAT COME BACK
Same wallets, different coin.
one coin scan sees a cluster. karma remembers every wallet it has seen bundle a launch or work a call, so when the same wallets turn up again, it notices. everything below cleared an evidence bar, and every wallet, block and transaction opens on solscan. check it yourself. that's the point.
1 · CREWS THAT TRAVEL
a crew is a set of wallets that bought the creation block of at least 3 launches together, took 3%+ of supply on average, and was fully out within seconds on at least 50% of those buys. one wallet in many launches can be a public sniper bot. the same wallets arriving together, launch after launch, and cashing out before anyone else can, is not luck.
2 · CALLERS' OTHER HANDS
a side wallet buys a coin before its caller posts the call, then sells into the people the call brings in. evidence comes in layers. pattern: the same wallet front-ran at least 10 of the caller's calls, spread over two weeks or more (a trader holding by luck gets there about once in a million). proven: a pattern plus a hard link, the caller funded it or the two signed the same transaction. crew: side wallets working the same calls together, named after the caller they work for. copy-trade bots that follow the caller's own buy by seconds, and bots that do this to every caller, are filtered out.
nothing has cleared the bar yet. the hunt runs over the graded callers; cases land here as they're proven.
CREWS ACROSS COINS
wallets that keep turning up together across coins, as top holders, launch-block buyers, identical-amount buyers or the wallets behind a sharp move, whatever the coins did. the evidence builds: together on 3+ coins is a suspect (kept, never named); 5+ coins, 10× more often than chance, is a pattern; a shared first funder on top is proven. read so far: 607 coins, 5,801 wallets.
0 proven · 5 pattern · 36 suspects building up
crew #1 · 3 wallets together on 7 coins · every pair on 7+ coins, 86.7× chance
as launch-block buyers: DrYe…, 49Pp…, DrXN…, 5Tsw…, 43X1…, 9Cfb…, Cze4…
crew #2 · 2 wallets together on 6 coins · every pair on 6+ coins, 101.2× chance
as launch-block buyers: $Dog, $Buddy, 92nd…, 3wCB…, 6zNq…, Bj6G…
crew #3 · 2 wallets together on 6 coins · every pair on 6+ coins, 20.9× chance
as launch-block buyers: $GTAfolio, $ALTAIDOPE, $KnightDog, 535A…, oziy…, NDPV…
crew #4 · 2 wallets together on 5 coins · every pair on 5+ coins, 60.7× chance
crew #5 · 2 wallets together on 5 coins · every pair on 5+ coins, 121.4× chance
THE PLATFORMS BEHIND THE FRONT-RUNS
the wallets that buy before a call and sell into it are mostly bots, and bots run on trading platforms. each leaves a fingerprint in its trades: a fee wallet taking ~1% of every buy, a gas relayer, a router program. of 800 sampled front-run buys (from 15,522 around graded callers' calls), 71% trace to one. a platform is named only when a public, open-source record names that exact address: DefiLlama's fee adapters, Jupiter's platform list, Dune's Spellbook, or the program's own on-chain metadata. the rest go by fingerprint, since anyone can mint a vanity address.
- pump.fun app25%
- Axiom22%
- Fomo16%
- GMGN7%
- Jupiter swap fee4%
- Padre3%
- Phantom swap1%
- Bloom1%
- own-built bots21%
share of the front-running traders whose platform we could trace.
| PLATFORM | TRADERS | CALLERS | FEE | TREND |
|---|---|---|---|---|
| pump.fun app relayer FHpc…7wCT | 167 | 36 | gas only | ██ |
| Axiom fee wallet 3Pvq…ZaTa | 142 | 36 | 1.02% | ██ |
| Fomo relayer AgmL…zN51 | 107 | 33 | gas only | ██ |
| GMGN fee wallet ByRR…BxRy | 48 | 30 | 1.01% | ██ |
| Jupiter swap fee token fee 7iWn…YBxE | 26 | 18 | 0.1% | ▇█ |
| Padre fee wallet term…ZzN3 | 20 | 18 | 1.02% | ██ |
| Phantom swap token fee 9yj3…djXf | 5 | 5 | 0.86% | ▆▆ |
| Bloom fee wallet 7HeD…1vKM | 4 | 4 | 1.01% | ▄▄ |
| own-built bots call pump.fun or an AMM directly, no platform fee | 141 | – | none | ██ |
| untraced: no fingerprint we can tie to a platform yet | 10% | – | – |
own-built bots, grouped by the shape of their transactions: via Jupiter · no tip · CU price ~1e5 µ-lamports · CU limit 180k (7) · PumpSwap · Jito tip · CU price ~1e6 µ-lamports · CU limit 200k (6) · pump.fun · Helius Sender tip + Jito tip · CU price ~1e5 µ-lamports · CU limit 120k (4) · via Jupiter · no tip · no priority fee · default CU limit (4) · PumpSwap · no tip · no priority fee · CU limit 150k (3)
LIVEevery trade Karma reads is classified as it's read: 7,611 trades from 801 wallets so far (6931 graded buys, 26 top-holder sells, 639 launch-block buys, 15 front-runs).
wallets by the platform they mostly use: own-built bot 68% · Fomo 15% · GMGN 4% · Axiom 4% · Padre 3% · Jupiter swap fee 2%
tracked over time: the study re-runs with the daily pipeline on the latest calls, 2 runs so far. last run 2026-10-15.
callers whose front-runners lean on one platform:
- @cap100x 53% of 19 front-runners via Axiom
- @CookerFlips 50% of 20 front-runners via pump.fun app
- @jackduval 47% of 19 front-runners via pump.fun app
- @Flippingprofits 47% of 19 front-runners via Axiom
- @hexiecs 45% of 20 front-runners via Fomo
- @TheMisterTurtle 45% of 20 front-runners via pump.fun app
- @sambati_ 45% of 20 front-runners via pump.fun app
- @Schoen_xyz 40% of 20 front-runners via Axiom
a signal, not a verdict: a front-runner here is anyone who bought in the 30 minutes before a call and sold in the 15 after, and ~20 per caller is a small sample. partly it reflects which platforms a caller's crowd uses.
signals from the chain, not accusations: a wallet here did what the links show, and you can read every one. memory grows every time someone opens a coin or a launch, so this page fills in with use. rebuilt 2026-10-18. the receipts test the grades.