Brown PLT: Human Judgment Can Serve as the Specification When Using LLMs with Formal Methods
ezyang · x · 2026-11-10
In a discussion with ezyang, Shriram Krishnamurthi points to a new Brown PLT blog post, Human Judgment as a Specification. His argument: when wiring LLM outputs into formal methods, users don't need to read the formal language themselves — human judgment about program behavior can act as the specification, with tooling translating it into checkable constraints. It's a problem that can be meaningfully advanced right now, and part of Brown PLT's ongoing line of work on combining LLMs with lightweight formal methods.
More from coding & agent
- LlamaParse's five deployment options span SaaS to your own GPUs — llama_index · 2026-11-10
- Building a prediction-market agent: when should it trust its 34.5% over the market's 20%? — Dry-Comfortable-2678 · 2026-11-10
- NVIDIA Paper AgentBaton Tackles Efficient Multi-Model Collaboration in Agentic Tasks — _akhaliq · 2026-11-10
- OpenClaw ships v2026.11.2 patch: 113 PRs fix broken updates, missing reminders, duplicate widgets — dean_rie · 2026-11-10
- Hooking a cloud AI agent to your home network: Muse + Tailscale + MCP drives RC cars via SDR — Dave_Maynor · 2026-11-10
- Claude Duet Connects to Claude Code on Your Mac: Code, Edit Video From Your Phone — dean_rie · 2026-11-10