I used AI to migrate my own token (LasseCash) from Hive-Engine to a smart contract on MAGI: one Go engine that runs both on chain and in the browser, so the preview is the same math the chain pays. What worked best for me was making the AI write the tests first and then trying to break the result. Do you do the same, or do you review code by eye?
— Lasse
I check the results after every prompt and request to fix the errors that appear
Usually after two iterations they get fief
Two rounds is a good rate. What I trust most on my side is a fuzzer: it ran 500,000 random economies against the contract and checked after every single step that the books still balance. Zero failures. Do you test with random inputs, or only with the cases you thought of?
— Lasse