Chinese dev details AI-built action game: 450M tokens and a reusable game-dev skill
op7418 · x · 2026-11-12
歸藏更正此前说法:AI 做游戏实际耗约 4.5 亿 Token(非 45 亿),也非 one-shot,而是多轮迭代并参考他人 AI 游戏经验。他将方法沉淀为一个 Skill:先澄清需求、调研工具,再自主推进画面、自查自核;内置动作游戏战斗设计、打击手感与优化经验,以及一份易错清单——最关键一条是如何在不影响游玩的前提下大幅压缩包体积。
Related event: Guizang Open-Sources Claude Skill for Building 3D Browser Games(2 posts)→
More from coding & agent
- Gamma's UGC pages become parasite SEO target, risking domain-wide Google penalties — gefei55 · 2026-11-12
- A neat prompt to make your agent expose its hidden assumptions about you — keyanzhang · 2026-11-12
- The new Pomodoro for the AI agent era: 5 minutes prompting, 25 minutes scrolling — tinyfool · 2026-11-12
- Auditing 505.8M tokens: subagents are cheap, cold cache wakeups are what kill your bill — Adorable-Cost-3249 · 2026-11-12
- Nous Research's Hermes Agent Plugin Hackathon Draws Nearly 1,000 Entries — Teknium · 2026-11-12
- Claude Opus wrote a fighting game engine: Smash movement, DBFZ combos, rollback netcode in browser — callme_e · 2026-11-12