AI 随想第十篇说代码学会了自我进化。那是观察。这篇是动手。

In Part 10 of Agentic AI Thoughts, I wrote that code had learned to evolve itself. That was the observation. This is the implementation.

两个账号都是企鹅，第二个还降级了。于是我用 Claude 逆向了它自己的宠物系统，0.2 秒出金。

Claude Code shipped a virtual pet Easter egg. Two accounts, two penguins — the second one downgraded from Rare to Common. So I used Claude to brute-force a Shiny Legendary Dragon in 0.2 seconds.

一行代码就能量文本高度——iOS 能，Android 能，Flutter 能，Web 不能。Pretext 用 15KB 补上了 CSS 欠了三十年的债。

A 15KB JavaScript library just solved one of the oldest architectural debts in web development — knowing how tall text will be without asking the browser to reflow the entire page.

The most important file in the Claude Code snapshot isn't a prompt file or a tool definition. It's a state machine. Once you see that, a lot of agent design decisions start making sense.

The most novel thing in the Claude Code snapshot isn't a tool or a prompt. It's how aggressively the runtime is built around preserving prompt-cache identity across forks, compaction, and long-running sessions.

After auditing a publicly posted Claude Code source snapshot and reading OpenAI's harness essays, I came away with one conclusion: the model is only the brain. The product is the runtime around it.

从发现 pretext 到在两个网站上线 7 个交互特效——爆炸物理、弹跳子弹、还有一只叫 Clawd 的像素螃蟹。一个 session 搞定。

From discovering pretext to shipping 7 interactive effects across two sites in one session — explosion physics, bouncing bullets, and an animated crab named Clawd.

跟 Claude Code 配合半小时写了个脚本，Apple Watch 按一下就能自动转文字、双语总结、同步到飞书和笔记。零成本，零依赖，比市面上所有 AI 录音产品都好用。

I wanted a voice-to-notes pipeline. The internet said build a watchOS app. A friend who tried said don't. Thirty minutes later — with Claude Code doing the heavy lifting — I had a zero-cost, zero-dependency solution using nothing but Voice Memos, iCloud, and one Gemini API call.

从第一性原理看 AI：所有趋势拆到底，都是算力够不够多、够不够好

First principles on AI: every trend, traced to its root, is about whether there's enough compute — and whether it's good enough

在 macOS 上搭了一套微信聊天记录自动化管道，踩了六个坑，终于让 Claude 能读我所有聊天记录还能帮我回消息

How I built a hybrid read+write WeChat automation pipeline on macOS — the tools, the architecture, and the 6 pitfalls that'll save you hours

三条需求曲线、一条成本下降线，以及为什么算力是新石油

Three demand curves, one deflation trend, and why compute is the new oil

听了一位前谷歌 TPU 工程师的深度访谈，终于理解了 TPU 和 GPU 不是同一道题。一个是通用性之王，一个是定制化之刃。Apple、Anthropic、Meta 都在用 TPU 了——这不是替代，是生态在裂变。