“According to Surge CEO Edwin Chen, all frontier AI labs view rigorous human evaluation—including fact-checking and assessing writing quality—as the gold standard for model assessment.”