Lauren Tan on the limits of formal methods and layered verification

Lauren Tan says formal methods cannot guarantee bug-free code.

In a post on X, she argues they only show that code matches a specification under certain assumptions, and that specifications come from human intent rather than from mathematics. She says she built the /statechart skill in pstack to make implicit state machines visible so reviewers can check that intent. She also says TLA+ with the TLC model checker has helped her find real bugs in her game. Her recommendation is to layer types, tests, and formal verification, applying formal verification to the parts of a system that need it.