Auditing in the age of (good enough) AI
Summary
The Trail of Bits post discusses using AI-powered tooling to elevate security audits of a new zero-knowledge VM (Miden). It details building an LSP server, a decompiler, a static analysis engine, and a Lean model from scratch, enabling deep, machine-checked verification and bug finding, including a high-severity issue related to proving input validation. The piece argues that frontier AI enables more thorough, repeatable security reviews and outlines future directions such as formal modeling and agent-driven workflows.