toly 🇺🇸

toly 🇺🇸

667,058 Twitter followers

Follow

Posts

Is percolator better?

HLP made ~$50M from the Oct 10 crash, and ~$20M again on Jan 31. Most debate focused on ADL. That's not where the money came from. With @ciamac (@Columbia_Biz @paradigm), @Qiaoqiao2001(@Columbia_Biz), and @xxxin001 (@StanfordMSE), we traced the full pipeline. Here's how it

Pretty good baseline for standard smart contracts. Formal verification still expensive but AI is pretty good at writing FV tests - so the hard part is thinking what invariants are worth enforcing. If the smart contract manages risk / depends on market volatility, the bar is much higher. * Oracle design is a huge rabbit hole * Failovers on cranks / keepers, oracles resilience * Risk parameters calibration * Invariant checks --- Previously a team had to have funding + good judgement + good devs to do the above. With code becoming cheap / free, the good judgement is more and more of the differentiating factor.

DeFi checklist that should be standard across solana: ✅ source available ✅ formal verification ✅ clear program upgrade or immutable policy ✅ bug bounty ✅ emergency halt policy

Scored 517. My journey: 1) Getting to ~450 was easy, just charge more when volatility spikes. 2) Up to ~510, I asked Claude to reverse-engineer leaderboard strats. The breakthrough was directional fee asymmetry: infer price direction, then charge more on the side arbs would trade and less on the retail side. 3) From 510 to 517, it is pure LLM alchemy. The biggest improvement came from realizing that not all trades carry equal information. Arb trades tell you where the real price is. 4) Beyond 517, I hit a wall. +50 ideas tested via Claude and Codex, but marginal improvement at best. Huge thanks for putting together such a fun and thoughtful challenge, @bqbrady and @danrobinson 🙏 twitter.com/0xrafal/status/202...

Loading..