For builders testing whether an addition contract behaves the same when its inputs are swapped, a clean fuzzing run is useful—but it is not proof. A September 1, 2026 comparison used a simple Soroban contract to examine that distinction.
The experiment tested five contract variants with the fuzzing tools cargo-fuzz and Komet, and the formal verification tool Certora Sunbeam. Three variants contained bugs that broke commutativity—the property that swapping the inputs should not change the result—only for specific inputs.
Fuzzing quickly found obvious or easier-to-trigger bugs, but missed defects that required precise inputs. Formal verification could instead produce a counterexample to the rule or prove that the rule held for all inputs. It offered a stronger guarantee than a fuzzing run that found nothing, though verification could also time out.
The practical distinction is what a passing result means: fuzzing reports that it did not find a bug in its run; it does not establish that the contract is correct.
