Fuzzing vs. Formal Verification: What a Soroban Contract Test Can Show

A comparison published on September 1, 2026, used a simple Soroban addition contract to test how cargo-fuzz, Komet and Certora Sunbeam handle bugs and correctness claims. The experiment included five contract variants. Three contained non-commutativity bugs that appeared only with particular inputs.

The fuzzing tools, cargo-fuzz and Komet, found obvious or easier-to-trigger bugs quickly but missed defects that required precise inputs. A clean fuzzing run therefore meant no bug was found, not that the contract was correct.

Certora Sunbeam, the formal verification tool in the comparison, could instead return a counterexample or prove that a specified rule held for all inputs. That stronger result has a practical limit: verification can time out.

Further reading