10 Using Strong AI to Write Evaluation Scripts to Score Other AIs

10 Using Strong AI to Write Evaluation Scripts to Score Other AIs. Business models emphasize high-frequency beating low-frequency; AI emphasizes high-intelligence beating low-intelligence—those beaten have no power to fight back. On the last day of the Spring Festival holiday, I used gpt5.2 xhigh to add an automated evaluation script to the previously set up AI Programming 2.0 evaluation task; this time it's a formal benchmark. The script's correctness passed on the first try, the test cases were adjusted once after not meeting expectations, and then the functionality for automated leaderboard statistics and auto-pushing was added—all passing on the first try. Writing code with strong AI now feels like this: regardless of the task size, you basically just need to state the requirements and expectations, and a simple acceptance of the results is enough—no need to look at the code at all. Next, I'll detail exactly how I did it. As usual, look directly at the chat records. Before starting, first /model to switch to the strongest gpt5.2 xhigh. Then directly...