Checking agent code against your intent Microsoft Research on 18 November described a neuro-formal verification (NFV) code reviewer that derives precise formal properties from a pull request's stated intent, has an agent attempt to prove the code satisfies each property, and uses a small trusted program to check the proof; on 38 recent pull requests the tool checked 217 properties. The post argues such proofs show only that code has the property written down under its listed assumptions, not that the property was the right one or that it was proved about the code that actually runs, so developers should still read agent-written code but aim to make agents write less and better code. The author cites Dan Luu's August runs, in which asking agents to use Rust's built-in test framework roughly doubled the number of tests written "but this didn't result in better correctness," and notes Harmonic's Aristotle, Midspiral's LemmaScript and Don Syme's Lean Squad work in a similar way. 1Introduction Coding agents now write more code than I can read. One answer is to stop reading. Robert C. Martin, better known as Uncle Bob, says his current strategy “is to not read any of the code written by my agents”. Instead the code has to “run the gauntlet” of his tests, quality metrics and mutation testing