--- name: listing-frontrun description: Study of wallets that buy Robinhood chain tokens before Robinhood "added to assets" and Binance Alpha events; read when Ben asks who front runs listings or wants a listing alert sources: [cowork] aliases: [front run study, robinhood listings, binance alpha listings] --- [stated] Ben asked (11 Sep 2026) to find wallets that front run Robinhood listings (JUGGERNAUT and FRONG were added to Robinhood assets on 10 Sep and did ~10x) and then the same for Binance. [stated] Ben calls the Robinhood event the "shortlist"; @NewListingsFeed calls it "Added to Robinhood assets" (ticker appears in the app; not the same as live on spot). Method used: fr.py on the droplet (/root/fr/fr.py) streams ERC20 Transfer logs from the RH RPC for 12h before and 12h after the event, nets per wallet, keeps EOA and 7702 wallets that accumulated at least 0.02% of supply before the event; results in /root/fr/.json, intersection in multi.json, nonces in fresh.json. Events run: Robinhood adds JUGGERNAUT+FRONG 10 Sep 12:29:34 UTC, PONS 2 Sep 13:29:41, CASHCAT 6 Aug 12:46:47; Binance Alpha featured PONS 3 Sep 01:14, CASHCAT+UP+AI 4 Sep 10:54, CNPY 6 Sep 06:32, MEME live 9 Sep 10:22 (times are @NewListingsFeed post times). Finding: no wallet appears in 3 or more events; 15 wallets in 2, nearly all same token or same joint announcement. Insider shaped buys are one off burner wallets (fresh, one buy of 0.3 to 1% of supply hours before, full dump after), so there is no repeat front runner to follow. Pipeline lead: Robinhood add often precedes Binance Alpha featuring (PONS 12h later, CASHCAT 4 weeks later). Gotchas: RH RPC returns None for eth_getTransactionCount at historical blocks (use latest nonce as freshness proxy; 7702 wallets keep low nonce via bundler so nonce means nothing for them); Blockscout moved to robinhoodchain.blockscout.com, its API works only from Chrome on the Mac and rate limits after ~6 calls; RH RPC 429s right after a heavy log pull, wait 20s; linear SPB drifts for events more than a couple of weeks back so "hours before" for old events is approximate. [stated] Ben asked whether the burner wallets share a funder (same person seeding burners) and to widen the window to 3 days before. Funding trace result (11 Sep): three different funders, no shared origin. JUGGERNAUT wallet 0x1e3da2a35b83dddf26c18bdcf1a3c1e2dffa78da is a side wallet of swing trader 0x99b0242dd5c8d8da7a3fc962119ecdb552c76a98 (trading JUGG daily since 6 Sep, 7 wallet cluster incl. 0xe90b574b4e31e29504804b0a4cfbde41159ffd56); tokens were moved in from his own wallets, not bought, then sold 3 minutes after the add. Trap repeated: fr.py counts wallet to wallet transfers as buys; always check the counterparty before calling a wallet fresh money. CASHCAT wallet 0x19d5a002cdad67ebb699ed3cce43459567b32d4c was funded 1 ETH by 0x6de92985925e90df0d7bc1e1e7b60db81c718d7c 9 minutes before the feed post, bought 1% of supply, dumped; feed post time may lag the real app add, so inconclusive. 0x0d0707963952f2fba59dd06f2b425ace40b492fe is the Gate.io hot wallet (same address on every EVM chain, confirmed by X posts); on Robinhood Chain it sweeps and pays out PONS, CASHCAT, AI, STONKBROKER, PIPEDOG, USDG, and dust funds withdrawal wallets with ~0.008 ETH. PONS wallet 0xc882b111a75c0c657fc507c04fbfcd2cc984f071 (0.51% of supply, still holding) is a Gate withdrawal wallet, not an insider. A wallet gas dusted by 0x0d0707 means "Gate customer", never "same person". 3 day window (72h before) on JUGGERNAUT: bigger accumulators appear and dump, e.g. 0xc5b3795628598054a607721dbf30cdbd6524269f (1.24% over 70h, sold 1.53% after) and 0x0bf95824106415de918ba6e996454406a22e47a1 (0.80%, 61h, sold all); FRONG: 0x395be6c099977a0b6dc52967b93579d32328db0a (0.88%, 35h, holding). Still no wallet crossing tokens. PONS_3D and CASHCAT_3D were still running at /root/fr when this was written. Blockscout v2 API (/api/v2/addresses/{a}/transactions?filter=from, /token-transfers, 50 per page with next_page_params) works from Chrome when the v1 ?module=account endpoint is rate limited; run long pagers as a background window job and poll, the javascript tool times out at 45s and its output truncates near 2KB.