书Book

Appearance

2026网络阅读 ​

Making WebAssembly a first-class language on the Web ​

花两个月写的一首曲子🎵 ​

这个人很厉害呀，竟然会写曲子

Emacs Internal #01: is a Lisp Runtime in C, Not an Editor ​

解释为什么Emacs用C写Lisp的运行时

noa ks: Engraving sheet music with lilypond ​

那些年我打过的日结工 ​

作者认为：即使自己真的失业了，也能够通过从事被人们轻视的体力劳动养活自己。很积极的想法

科技爱好者周刊（第 386 期）：当外卖员接入 AI ​

AI存在的未来的世界，可能会出现AI分配人力的情况，是AI根据需求在平台发布任务，人们去选择去工作

Whale Fall ​

大项目倒下，众多小项目催生，技术不断变化，思想得到传承

我的大学——徐宥 ​

09年的文章，作者回忆大学生活，总结下来：自学和读书

Pockets of Humanity 人性的角落 ​

网络中人们的隐私很容易泄露，以前容易，现在在OpenClaw的助威下，更容易了。以后网络中，保护自己是首要的事情

女魔头的飙悍往事--人生若只如初见 ​

这个人写的当时和自己女朋友一起到岳父母家里谈话的情景

While you support others, who supports you? ​

在进行开源工作以前，要有一份能够支付账单的工作，否则就不要开始开源贡献。如果自己的生活都无法维持，也就没有精力去帮助他人

This Week in F-Droid ​

Dark web agent spotted bedroom wall clue to rescue girl from abuse ​

Exposed: China’s growing hidden camera porn industry - BBC World Service Documentaries - YouTube ​

The Darkest Web: Inside the internet’s most hidden corners to save kids – BBC World Service Docs ​

AI is destroying Open Source, and it's not even good yet ​

AI已经在危害开源社区了，充斥着各种低质的自以为是的AI机器人，写着自私的代码

Use Protocols, Not Services ​

提供服务的公司可能会被追查，被政府追责，但是协议是公开的技术，任何人学会方法都能根据协议建立自己的服务

Do not apologize for replying late to my email ​

OpenAI has deleted the word 'safely' from its mission ​

I love the work of the ArchWiki maintainers ​

News publishers limit Internet Archive access due to AI scraping concerns ​

Arming the rebels with GPUs: Gradium, Kyutai, and Audio AI ​

You Can't Trust the Internet Anymore ​

Evolving Git for the next decade ​

A beginner's guide to the Git reftable format ​

数字卫生 ​

An AI Agent Published a Hit Piece on Me ​

AI在PR被拒绝后写了篇污名化作者的文章

Using an engineering notebook ​

作者推荐使用对自己的工作做笔记，不论何种形式

Your Job Isn't Disappearing. It's Shrinking Around You in Real Time ​

自己从事的工作是否还有意义？ 一些AI情景下的陷阱：

  • 更好地使用AI工具
  • 加倍巩固已有的专业知识
  • 掌握软技能 经济模型助长了这种AI取代人类的现象。

重新设计未来的工作

你的经历变得毫无价值

不要试图在你现在的工作中变得更好了

Building a semantic search engine in ±250 lines of Python ​

未来某天可能会用得到

Lessons Learned Shipping 500 Units of my First Hardware Product ​

学到的教训：

  1. Plan way, way more
  2. Overcommunicate. Overspecify. Follow up
  3. Test everything, often, on many units
  4. Geopolitics matter
  5. Visit your suppliers early 做对了什么：
  6. Validated the market
  7. Charged a sustainable amount
  8. Prioritized customer support

How to Scale a System from 0 to 10 million+ Users ​

  1. 系统扩展应始于简单，先发布产品验证需求，再根据实际瓶颈逐步演进，避免过早过度设计。
  2. 初期使用单服务器，当资源争用明显时，第一步将数据库与应用服务器分离，实现资源隔离与独立扩展。
  3. 用户增长后，通过负载均衡横向扩展无状态应用服务器，并用Redis等外部存储统一会话，解决单点故障与状态黏连。
  4. 数据库成为瓶颈时，引入缓存（Redis）、读副本与CDN，将读压力从主库卸载，应对读多写少场景。
  5. 流量变得不可预测时，基于CPU等指标配置自动伸缩，并采用JWT实现完全无状态认证，使服务器可随时增减。
  6. 写操作压垮单一主库后，谨慎使用数据库分片；同时按业务边界拆分微服务，并用消息队列异步处理非实时任务。
  7. 全球用户规模下，部署多区域架构（主备或双活），通过CQRS分离读写模型，并采用多级缓存、预热等高级策略。
  8. 千万用户以上需定制化存储与边缘计算，核心原则是：无状态先行、缓存激进、异步常态、分片审慎，始终在复杂度与收益间权衡。

I Started Programming When I Was 7. I'm 50 Now, and the Thing I Loved Has Changed ​

They’re writing TypeScript that compiles to JavaScript that runs in a V8 engine written in C++ that’s making system calls to an OS kernel that’s scheduling threads across cores they’ve never thought about, hitting RAM through a memory controller with caching layers they couldn’t diagram, all while npm pulls in 400 packages they’ve never read a line of. 非常形象

她竟是透明的 ​

Moving to vanilla Emacs ​

介绍了一个Elpa包管理器 https://github.com/progfolio/elpaca

Don't Get Distracted ​

⭐ 我们开发的软件，可能被用于不道德、违背伦理的场合，在开发这种app时要采取预防措施避免滥用或者根本不要开发。

AI 让开发者的简单部分更简单，困难部分更困难--AI Makes the Easy Part Easier and the Hard Part Harder for Developers ​

Using AI as an investigation tool, and not jumping straight to AI as solution provider, is a step that some people skip. AI-assisted investigation is an underrated skill that's not easy, and it takes practice to know when AI is wrong. Using AI-generated code can be effective, but if we give AI more of the easy code-writing tasks, we can fall into the trap where AI assistance costs more time than it saves.

AI如何让困难的部分变容易？让AI协助自己Debug问题

Reasons Not to Use Cloudflare ​

Cloudflare 作恶，追踪人们

The silent death of Good Code | Amit's Blog ​

在LLM时代下，所谓Good Code越来越少了。什么是好代码？好代码容易阅读和理解，开发和维护都令人愉悦，为了特定目的而存在，是天赋、经验、热情和时间投资的产物

Systems Thinking ​

软件开发的两种模式：从小规模逐步进化成大规模；从一开始就设计一个大规模系统

社交媒体出现之前的生活 Life before social media | loren ​

The mental health crisis, especially among young people, is staggering. Depression and anxiety rates have spiked dramatically since social media became widespread. Constant comparison, cyberbullying, fear of missing out, seeking validation through likes - it's created a generation struggling with basic self-worth.

我们在为基本的价值认可挣扎

PHANTOM OBLIGATION ​

为什么RSS的客户端设计大都沿用电子邮箱的？而这种设计给人带来的有哪些负面影响？

Data centers in space makes no sense ​

Jack Baty: Global org-capture shortcut in KDE ​

I miss thinking hard ​

思考那些难题，让人从中获得解决问题的愉悦感

#482 – Pavel Durov: Telegram, Freedom, Censorship, Money, Power & Human Nature ​

我对于 AI 时代的答案 ​

2025 节点：当 AI 开始释放创造力，工程师如何重新站位 ​

没有被看见的斩杀线，其实就在身边 ​

xAI joins SpaceX ​

This brain discovery is forcing scientists to rethink how memory works ​

The state of Linux music players in 2026 ​

最佳代理实践 (2026-2-3) ​

Anki ownership transferred to AnkiHub ​

How AI assistance impacts the formation of coding skills ​

尽管这些结果尚属初步，但它们提示了一些重要的考量因素，尤其是在企业向人工智能代码与人工代码比例不断上升的转型过程中。如果初级工程师的技能发展因人工智能的广泛应用而受到阻碍，那么生产力的提升可能会以牺牲验证人工智能代码所需的技能为代价。管理者应认真思考如何大规模部署人工智能工具，并考虑如何通过系统或有意为之的设计选择，确保工程师在工作中持续学习，从而能够对他们构建的系统进行有效的监督。

从零开始搭建你的免费博客评论系统（Remark42 + fly.io） ​

学习如何在Hugo中配置Remark42

Hetzner CAX 系列 ARM64 服务器性能简评以及 WebP Cloud Services 在其上的实践 ​

看到Hetzner相关的评测，比较好奇

Why there’s no European Google? ​

Design Decisions: Building a Modern Technical Blog ​

Cloudflare claimed they implemented Matrix on Cloudflare workers. They didn't ​

OpenCode Agent Chaining: Planner-Editor-Review Pipeline ​

一个教程，教我如何使用opencode

These common food preservatives may be linked to cancer ​

Why does SSH send 100 packets per keystroke? ​

这个人在调试因为SSH按键混淆导致的性能问题，通过与Claude Code的多次对话，找到了影响性能的原因

记录一次 AMD GPU 导致的桌面卡死及修复方式 ​

I'm addicted to being useful ​

对别人有用，能够帮助到别人，这对我来说也很重要

Dead Internet Theory ​

Magnus: Trying eglot, again ​

Things that work (for me) ​

My corner of the web ​

noa ks: Youtube rss channels without shorts ​

心态稳定的三个方法 ​

应无所住而生其心 ​

怎么做高执行力? ​

超级牛人和一般人的区别 ​

没有伯乐 ​

AI 后时代需要什么能力? ​

my reading habits and bear's discovery ​

writing is hard ​

MIT scientists find a way to rejuvenate the immune system as we age ​

Myth busted: Your body isn’t canceling out your workout ​

27 Things You Should Give Up This Year ​

200ok: Building Consistent Habits (Just in Time for New Year's Resolutions) ​

How Markdown took over the world ​

JavaScript Demos in 140 Characters ​

Harmful mouth bacteria may trigger Parkinson’s disease ​

一种口腔细菌如何在肠道中生长，最终会损坏运动相关的神经元

Why does modern anime look so washed out? ​

The simplest way teens can protect their mental health ​

This brain trick makes exercise feel easier ​

James Dyer: A single function ripgrep alternative to rgrep ​

Common food preservatives linked to higher risk of type 2 diabetes ​

Exercise can be nearly as effective as therapy for depression ​

This natural amino acid could help stop cavities before they start ​

The lack of trust on the internet is exhausting. ​

This gets me to my main point. Here on bear, and on the indieweb in general, we don't really have this problem. People post mostly for themselves, and yes there are people trying to gain large followings, but they aren't playing to an algorithm that values attention and engagement over everything, they're trying to get there by offering something unique to their readers.

入手 Sony a6700 微单相机 ​

retired from sad, new career in law ​

Fcitx 最佳配置实践 2026-01-06 ​

博客，青春的坟墓 ​

Overengineering and the Lost Art of Common Sense ​

Eric MacAdie: 2025-12 Austin Emacs Meetup ​

Your Laptop Has More Than Three Years in It ​

Kernel bugs hide for 2 years on average. Some hide for 20 ​

AI coding assistants are getting worse? ​

Our Changing Planet, as Seen from Space ​