UCLA 在读产品经理，聚焦游戏产品、AI 原生应用与数据驱动增长，具备 TikTok、腾讯游戏等产品实践经历。https://astro-wen.github.io/zh-CNhttps://astro-wen.github.io/blog/second-post/https://astro-wen.github.io/blog/second-post/从 Next.js、Hexo、Hugo 对比到 Astro 的最终选择，以及 Content Collections 带来的开发体验。Sat, 25 Apr 2026 00:00:00 GMTastrotechbloghttps://astro-wen.github.io/blog/hello-world/https://astro-wen.github.io/blog/hello-world/这是第一篇示例文章，介绍了本站点的写作工作流与 Markdown 支持的一些常见语法。Mon, 20 Apr 2026 00:00:00 GMTmetawritinghttps://astro-wen.github.io/blog/skill-writer-lessons/https://astro-wen.github.io/blog/skill-writer-lessons/花两天读了100篇游戏技术文章，一篇篇喂给 skill-creator 转化成 AI 能执行的"能力包"。在第14个 Skill 处，我发现了一个让人坐不住的规律：不是所有文章都能变成好 Skill——但我们可以用"换皮测试"一招判断。Sun, 29 Mar 2026 00:00:00 GMTaiskillknowledge-managementproductcontext-engineeringhttps://astro-wen.github.io/blog/ucla-c111-lol-prediction-en/https://astro-wen.github.io/blog/ucla-c111-lol-prediction-en/Using machine learning to analyze 10-minute game data from 9,879 Diamond-tier matches and predict victory with 72% accuracy. Full technical report with code and figures.Mon, 15 Dec 2025 00:00:00 GMTmachine-learninglolpythonucladata-sciencehttps://astro-wen.github.io/blog/ucla-c111-lol-prediction-zh/https://astro-wen.github.io/blog/ucla-c111-lol-prediction-zh/基于 9,879 场钻石段位比赛的前 10 分钟数据，训练 6 种分类模型预测英雄联盟胜负。完整技术报告，含全部代码、图表与分析。Mon, 15 Dec 2025 00:00:00 GMTmachine-learninglolpythonucladata-sciencehttps://astro-wen.github.io/blog/disco-elysium-narrative-design/https://astro-wen.github.io/blog/disco-elysium-narrative-design/从任务策划和叙事策划的视角，深入分析《极乐迪斯科》的技能人格化、检定叙事化、思绪内阁等核心设计，探讨其亮点、不足与迭代方案。Sat, 29 Nov 2025 00:00:00 GMTgame-designnarrativedisco-elysiumcrpghttps://astro-wen.github.io/blog/expedition33-cost-efficiency/https://astro-wen.github.io/blog/expedition33-cost-efficiency/深入剖析《光与影：33号远征队》的开发哲学——核心团队仅32人，预算低于二十年前的2A游戏，却斩获Metacritic 92分、全球超330万销量。从序章设计、战斗系统、UE5蓝图工作流到战略性外包，拆解小团队如何以"精益"对抗"力大砖飞"。Fri, 03 Oct 2025 00:00:00 GMTgame-designexpedition-33jrpgue5indie