ForestSharing insights on learning, data-driven innovation, AI thinking and my life
Karpathy posted a 5000-star gist this month: stop feeding Claude raw docs every time, compile them into a wiki first. I've been re-reading the same 80 notes for months. Here's why he's probably right — and where the pattern breaks.
Your teammates use Claude Code, Copilot, and Codex. The enterprise plans don't fix the problem. Here's a four-layer model for governing AI coding at the project and task level — and why the repo, not the tool, has to be the source of truth.
My OpenClaw agent had been silently dropping half of its Telegram responses. No error, no warning. I found the bug, and nobody was fixing it — so I did.
Get new articles delivered to your inbox.
整理博客时翻起 Google Photos，七万张照片里自己的身影寥寥无几。却在丽江和大理的老照片里，找到了一段关于小火车、一条古街、一首歌、和一个在街上哭着叫妈妈的小朋友的故事。
Anthropic launched Managed Agents yesterday. I spent the day reading, thinking, and realizing the game has shifted — from who can build the best agent infrastructure to who can design the best agent strategy. Some thoughts on speed, consulting, and the uncertain road ahead.
QQ音乐上偶然听到程璧版的《恋恋风尘》，一下子就回到了大四那间深夜的实验室。白大褂、变性淀粉、闪烁的走廊灯——和一些后来再也没见过的人。
一开始我对 Codex 的印象并不好。用惯了 Claude Code 之后，更觉得它差得远。但在 Claude Code 的优惠结束、session 限制变紧之后，我抱着备胎心态重新试了试 Codex，结果 GPT-5.4 反而刷新了我的判断，也直接推动我把它接进了自己的 Web Claude Code Pilot。
I built an OpenClaw agent to send my son daily knowledge bites on WhatsApp. Then a version update broke echo detection, and the bot spent hours in a self-conversation loop. Here's what went wrong, how we fixed it, and what it says about where tools like OpenClaw actually stand.
上篇文章说要找风云之天下会来玩。找到了，但在 Windows 11 上跑不起来。网上下载的修复包全是病毒。于是我决定跟 AI 一起，从零开始修复这个 1998 年的游戏。PE 分析、DLL 代理、DirectDraw 兼容层——这些我以前完全不懂的东西，现在居然都能搞定了。
I spent weeks building a web-based Claude Code interface for remote access from my phone. Then Anthropic shipped a native remote control feature. My version is still better — but the clock is ticking.
AI coding assistants forget context between sessions, reintroduce fixed bugs, and make baffling 'improvements.' Tests are the guardrail that survives the amnesia. Here's what I learned building a real project with Claude Code — and what Kent Beck, Martin Fowler, and the data say about testing in the age of AI.
'The AI PM’s Playbook' is an article about how to become an excellent AI PM. It introduces the responsibilities, skill requirements, workflows, and how to collaborate with teams.
