"How to Reduce Bugs Generated by AI in Code?" AI code often corrects one bug but introduces more. This is likely the norm in AI programming. Here are some methods to reduce AI-generated bugs: 1️⃣ Have the AI write high-determinism tests for the specific bug first. 2️⃣ Establish ironclad rules for bug fixing. Refer to: 【Ironclad Rules - Immediate Termination for Violation】 1. Never refactor code that has no problems. 2. Do not add any defensive checks (null checks, optional) unless the test explicitly fails there. 3. Do not modify the input parameters, output parameters, or exception types of existing functions. 4. Do not introduce any new dependencies. 5. Do not change naming conventions, folder structure, or import order. 6. When fixing bugs, prioritize the most localized and minimal changes (ideally < 15 lines). 7. Solve only one explicitly stated problem at a time. 3️⃣ Small iterations: Fix only one specific point at a time, like a surgeon making precise incisions. After using this combination, the AI became much more obedient and did not "creatively" introduce new problems, saving time and tokens!