DEV Community

liao-zhizhi profile picture

liao-zhizhi

Independent researcher. Author of harness-audit-sidecar (MCP audit proxy) and agent-metabolism-os. Building reliable and verifiable AI agent runtimes.

Joined Joined on  github website
AI Agents Keep Lying About Finishing Tasks: Why LLM-as-a-Judge Fails and How to Fix It with Physical Witnesses
liaozhizhi profile

AI Agents Keep Lying About Finishing Tasks: Why LLM-as-a-Judge Fails and How to Fix It with Physical Witnesses

#ai #opensource #programming #llm
Comments 1
3 min read

Want to connect with liao-zhizhi?

Create an account to connect with liao-zhizhi. You can also sign in below to proceed if you already have an account.

Create Account
Already have an account? Sign in
From Code to Motor: Why Software Audit Sidecars Fail in Embodied AI
liaozhizhi profile

From Code to Motor: Why Software Audit Sidecars Fail in Embodied AI

#ai #robotics #systems #opensource
Comments
2 min read
Why Your LLM Agent Keeps Burning Tokens (and How to Cut It at the HTTP Layer)
liaozhizhi profile

Why Your LLM Agent Keeps Burning Tokens (and How to Cut It at the HTTP Layer)

#ai #llm #programming #opensource
Comments
3 min read
Why Your AI Evaluator Is Structurally Blind (and how to mathematically prove it)
liaozhizhi profile

Why Your AI Evaluator Is Structurally Blind (and how to mathematically prove it)

#ai #opensource #machinelearning #lean4
Comments
3 min read
The Geometry of Progress: How Bio-Quantum Interfaces Break the Laws of Scale
liaozhizhi profile

The Geometry of Progress: How Bio-Quantum Interfaces Break the Laws of Scale

#ai #science #causality #complexity
Comments
2 min read
Trust, but Verify: How We Audited OpenAI's Lean 4 Math Proofs with a Zero-Dependency Sidecar
liaozhizhi profile

Trust, but Verify: How We Audited OpenAI's Lean 4 Math Proofs with a Zero-Dependency Sidecar

#ai #opensource #math #lean4
Comments
3 min read
loading...