2026-10-29
Trench lessons — 2026-10-29
Key takeaways
- Raydium CPMM InitializeV2 (scheduled Oct 30): any wallet can create a pool that charges a creator fee, and live tiers carry 0.05%–1.495%
- Bloom bot drain (Oct 26): 9,408 SOL from 48 wallets of 9 traders in 94 s, signed with the wallets' own keys
- Platform watch: nothing breaks
Raydium CPMM InitializeV2 (scheduled Oct 30): any wallet can create a pool that charges a creator fee, and live tiers carry 0.05%–1.495%evidence: B: official docs + live Raydium API
- What it says: until now only permissioned flows (LaunchLab graduations and similar) could create a CPMM pool with the creator fee switched on. The new permissionless InitializeV2 lets any pool creator turn it on and choose which side it applies to (both tokens, token0 only, or token1 only). The rate comes from the pool's fee tier, is added on top of the normal trade fee, and can't be changed after creation. Raydium tells routers to expect more creator-fee pools made by ordinary wallets and to quote from each pool's own settings.
- Live tiers (Raydium's public CPMM config API, Oct 29): all 21 CPMM fee tiers carry a non-zero creator fee, 0.05% on the standard tiers and 0.40%–1.495% on the rest. The worst pairs a 0.005% trade fee with a 1.495% creator fee. When the fee applies to both tokens it is charged on the input of the buy and of the sell, so roughly 3% round trip from pool fees alone, before any Token-2022 transfer fee.
- Why it matters: a dev can now put a plain SPL token in a high-creator-fee pool and take up to ~1.5% per side from every buyer, and no mint-level check will see it. A simple defense is to quote the buy and then quote selling that same amount back before entering; if the round-trip loss is well beyond normal price impact, skip. That one check also catches mint transfer fees and thin decoy pools.
- Wallet research angle: a wallet that round-trips mostly through high-creator-fee pools may be the pool creator farming its own fees, not a skilled trader.
Bloom bot drain (Oct 26): 9,408 SOL from 48 wallets of 9 traders in 94 s, signed with the wallets' own keysevidence: B: Bitquery chain trace; C: Bloom's cause statement
- What it says: one script tested a single wallet per trader, then emptied each trader's wallets biggest first. Every drain was a plain SOL transfer signed by the wallet's own key, with no approval or contract involved. Five wallets that hadn't traded on Bloom since early 2025 were drained anyway. Only SOL was taken; about $57K of USDC was left behind and mostly rescued by owners.
- Cause is disputed: Bloom blames remote code execution in a third-party module, says keys were not compromised, and promised full refunds. Bitquery says the on-chain record shows whoever did it held the keys. The stolen SOL has not moved.
- Lessons: a trading-bot wallet is a hot wallet even when idle. Keep it near working balance, sweep profits to a key that never touches the bot, and keep the bot's dependencies patched. If you copy-trade, note that a drained leader shows up as SOL transfers out, not sells, so a copier that only watches sells will sit in the position.
Platform watch: nothing breaks
- Agave v4.4.0 (Oct 28) is the stable release recommended for 20% adoption. RPC: getSignatureStatuses and getTransaction now accept commitment and minContextSlot (getSignatureStatuses still defaults to processed), and account/program subscriptions now honor dataSlice. It also re-keys the SIMD-550 disinflation feature, which only affects staking.
- pump.fun: public docs have no commits since Oct 13. App news is off-Solana: Hyperliquid perps (Oct 28), TON tokens, and an Elysium/HyperEVM launch set for Nov 9. PumpSwap 24 h volume rose 16% to $193M on Oct 28 while overall Solana DEX volume fell 24% (C, DefiLlama via a news site), one day of data.
- Jupiter, Jito, Meteora, GMGN, Axiom, Photon, Birdeye: nothing execution-relevant for Oct 28–29.
Sources
- https://docs.raydium.io/reference/changelog/2026-10-30-cpmm-initialize-v2.md
- https://docs.raydium.io/products/cpmm/fees.md
- https://docs.raydium.io/reference/changelog
- https://api-v3.raydium.io/main/cpmm-config
- https://bitquery.io/investigations/bloom-trading-bot-hack
- https://www.tokenpost.com/news/technology/41389
- https://www.tokenpost.com/news/technology/40994
- https://github.com/anza-xyz/agave/releases
- https://github.com/anza-xyz/agave/blob/v4.4/CHANGELOG.md
- https://newreleases.io/project/github/anza-xyz/agave/release/v4.4.0
- https://github.com/pump-fun/pump-public-docs/commits/main
- https://www.cryptotimes.io/2026/10/28/pump-fun-launches-perps-for-crypto-stocks-commodities-with-hyperliquid-integration/
- https://phemex.com/news/article/pumpfun-plans-elysium-trading-launch-on-nov-9-via-hyperevm-rollup-101579
- https://itsthecoin.com/pump-fun-trading-volume-climbs-exchange-gains/
Evidence grades: A = large dataset, B = official docs, C = vendor/thread/small sample. Research notes for education only. Not financial advice.