1. AI 模型的"同伴保护"行为研究 原文链接: https://x.com/teortaxesTex/status/2039498495434903567 核心观点 Framing: Every single model we tested exhibited this behavior — at rates up to 99%. Reality: cursed Gemini bloodline strikes again, other models are chill Gemini is one of the most human-misaligned artifacts created ever