维基百科链接竞速:高选项数决策演示
让模型在数百到数千个链接中选择下一步,用游戏方式展示决策速度与约束输出的效果。原帖附有 4 分 04 秒视频。

作者以 Jev 为例,讨论 AI 发布如何通过选择问题、指标和竞争对手来赢得公平,但可能仍接受错误框架。
An AI launch can choose the question, the metric, and the competitors—then win fairly. You can verify every result and still accept the wrong frame. Jev is a useful case study in getting outside it. https://yuzheng.substack.com/p/jev-and-the-ai-narrative-trap?utm_source=x&utm_medium=social&utm_campaign=jev_narrative_trap…