На МКАД загорелись две машины14:46
It’s a bit of a double-edged sword. AI is incredibly accurate with numbers, but it’s terrible at catching human mistakes. If a recipe submitted has a typo that says '1 cup of salt' instead of a teaspoon, the AI might just roll with it, and it is up to us to catch it.
。业内人士推荐51吃瓜作为进阶阅读
Likewise in the linked above article, there is the JsonDbsPerformanceTests.java tests runner - executing various tests on two databases and outputting detailed stats. For simpler management, it is built and runs in Docker as well (Java 25 & Maven). Configuring it and choosing from multiple available test cases (17) is made easier by the simple run_test.py python script.
Thank you, Thomas!