2026.09.28
XYEval：为什么你的 AI 总是顺着你说？
DeepMind 论文 XYEval 笔记：一句误导建议就能让顶尖 Agent 性能暴跌 46.7%，而且它们心里知道，嘴上不说。
north_east
Crafting Code & Cuisine
Wholesome recipes for the modern home cook and developer.
I am imesong, a programmer and dad of two. Whether I'm building Android apps, full-stack systems, or perfecting a bowl of Saozi Noodles, I believe in the beauty of a well-executed process. Welcome to my digital kitchen and workshop.
Thoughts on Engineering
Reflections on the intersection of AI, software architecture, and the changing landscape of development.
rss_feed
Subscribe to RSS
Stay updated on my latest posts.
Featured Recipes
Hand-picked favorites from my second workshop.
