βeta
2.2K posts

βeta
@levbeta
rust lover /n/ hft enthusiastic /n/ @ bulktrade



If you are forking percolator this is my LLM flow to minimize bugs. 1) use max thinking mode to generate a spec for feature X Loop: 1) is the spec succinct, unambiguous, formally verifiable in kani? Write the audit to audit.md 2) read the audit, is it correct? Update the spec with any issues. Then basically loop the same thing for the implementation. 1) implement the spec and pass all the tests and proofs and is the proof and test coverage complete? Loop: 1) is the implementation succinct? Are all the tests valid? Are all the proofs non vacuous? Is the coverage complete? Assume dev is adversarial. Write the audit to audit.md 2) fix any implementation issues in audit.md I sometimes use two different LLMs, codex vs Claude




GTE is the world's fastest blockchain and decentralized exchange. Live on private testnet now. Team is from Jump, Citadel, Palantir, etc & early infra engineers from Facebook, Amazon, Uber, etc who've scaled apps to billions of users. And you're not bullish?

Hey Solana community, we've seen our low IBRL score which is due to Firedancer running the revenue scheduler, leading to late packing. The revenue scheduler will be deprecated in the next release, but we're switching to the balanced scheduler immediately. Committed to IBRL! 🚀 ibrl.wtf Cheers to @jito_sol for the awesome tool

Jito made a whole website about us. If you're wondering whether we're a threat to their monopoly, I think they just answered that for you.
















