New Python Tool Makes Smart Contract Testing 11x Faster

Smart contracts are digital agreements that run on the blockchain, and bugs in them can cost millions. Developers use fuzzers—automated tools that throw random inputs at a contract to see if it breaks—to catch problems early. Wake 5, a new open-source Python framework, makes this testing much faster by running the blockchain simulation inside the tool itself, rather than connecting to an external node. On a standard benchmark, it processed nearly 30,000 transactions per second, matching the industry-leading Foundry and leaving Echidna in the dust at 2,600.

But raw speed doesn't guarantee better bug hunting. In eight-hour test runs, all three tools plateaued at around 70-75 found bugs, and even together they only reached 93 out of 153 possible. The bottleneck isn't how fast the fuzzer runs—it's how smart it is at picking inputs. Wake's random number generator, for example, spends half its effort on small numbers, so it rarely tries the large values that might trigger hidden bugs.

The game-changer came when the researchers let an AI language model write the test harness—the script that feeds inputs to the contract. Using a constraint solver (a tool that finds inputs meeting specific conditions), the AI's harness found all 153 bugs in every run, taking just 33 to 79 minutes. That's because the AI could reason about what inputs were needed, rather than relying on blind randomness.

The team also tested whether AI code auditors benefit from static analysis tools. They found that such tools improved precision (fewer false alarms) for all three AI models tested, but didn't help them find more real bugs, and increased AI costs by up to 12.7%. The takeaway: fuzzers are shifting from being the searchers to being the executors, with AI doing the thinking. For developers, this means faster, more reliable smart contract testing—and potentially safer blockchain applications for everyone.