[{"data":1,"prerenderedAt":324},["ShallowReactive",2],{"blog-posts":3,"content-\u002Fposts\u002Fthe-age-of-hesitation":234},[4,15,21,27,37,47,53,59,69,73,77,81,91,98,104,108,113,121,128,136,141,147,153,158,163,170,177,184,190,197,205,211,217,223,229],{"path":5,"title":6,"description":7,"date":8,"categories":9,"tags":10,"draft":14},"\u002Fposts\u002Fthe-age-of-hesitation","[搬运] 不敢投入的时代","为什么我们现在越来越不敢投入了。很多时候我们以为的选择，比如你有两个工作offer，现在选择好像是我到底应该考公还是考编。明明这两个选项都不是真正的选项，因为你还没有开始，也不知道最后能不能成。这个时候我们就已经开始犹豫了。","2026-10-10 11:44:12","生活杂谈",[11,12,13],"翻译","搬运","心理学",false,{"path":16,"title":17,"description":18,"date":19,"categories":9,"tags":20,"draft":14},"\u002Fposts\u002Fwhy-happiness-look-like-same","[搬运] 为什么大家的幸福都长一个样？","为什么大家的幸福都长一个样？一种认知被广泛传播后似乎变得不断扁平，这种感觉很奇怪。","2026-10-09 13:57:15",[11,12,13],{"path":22,"title":23,"description":24,"date":25,"categories":24,"tags":26,"draft":14},"\u002Fposts\u002Feverything-is-paid-twice","任何事情都要被消费两次","","2026-10-06 12:30:06",null,{"path":28,"title":29,"description":30,"date":31,"categories":32,"tags":33,"draft":14},"\u002Fposts\u002Fwhy-llm-seems-worse-after-launch","为什么模型上线后，体感像是“降智”了？","权重与 KV cache 的量化都能降低 serving 成本，也可能改变输出；但体感变弱不足以证明服务后来切换了精度。","2026-09-29 21:44:08","技术杂谈",[34,35,36],"LLMs","Serving","量化",{"path":38,"title":39,"description":40,"date":41,"categories":42,"tags":43,"draft":14},"\u002Fposts\u002Fspinning-up-policy-optimization","[译] 深度强化学习入门（三）：策略优化入门","OpenAI Spinning Up 教程第三部分的全文翻译：推导最简单的策略梯度、log-derivative trick、EGLP 引理、reward-to-go 与基线，配 PyTorch 示例代码。","2026-09-28 17:49:25","技术",[44,45,46,11],"强化学习","RL","Spinning Up",{"path":48,"title":49,"description":50,"date":51,"categories":42,"tags":52,"draft":14},"\u002Fposts\u002Fspinning-up-rl-intro-2","[译] 深度强化学习入门（二）：强化学习算法的分类","OpenAI Spinning Up 第二部分的全文翻译：无模型与基于模型的 RL、策略优化与 Q-learning 的取舍，以及规划、专家迭代和经验增强等方法。","2026-09-28 10:20:21",[44,45,46,11],{"path":54,"title":55,"description":56,"date":57,"categories":42,"tags":58,"draft":14},"\u002Fposts\u002Fspinning-up-rl-intro","[译] 深度强化学习入门（一）：RL 的关键概念","OpenAI Spinning Up 教程第一部分的全文翻译：智能体与环境、状态与观测、动作空间、策略、轨迹、奖励与回报、价值函数、贝尔曼方程、优势函数，附 MDP 形式化。","2026-09-28 09:29:40",[44,45,46,11],{"path":60,"title":61,"description":62,"date":63,"categories":9,"tags":64,"draft":68},"\u002Fposts\u002Fwhen-machines-come-alive","冰箱之后，生活还会怎样改变？","从厨房里的果蝇想到冰箱、移动互联网与 AI 的多轮调用：AGI 已经到来，只是昂贵的智能会怎样改变生活？","2026-09-25 06:36:22",[65,66,67],"AI","随笔","科技",true,{"path":70,"title":71,"description":24,"date":72,"categories":24,"tags":26,"draft":68},"\u002Fposts\u002Ftest","test","2026-09-20 06:37:44",{"path":74,"title":75,"description":24,"date":76,"categories":24,"tags":26,"draft":68},"\u002Fposts\u002Fmathematical-foundations-of-llms-1","LLM的数学基础（1）：","2026-09-19 22:59:55",{"path":78,"title":79,"description":24,"date":80,"categories":24,"tags":26,"draft":68},"\u002Fposts\u002Ftowards-maximize-serving-capatity","如何在 offline bench 里最大化 LLM serving system 的吞吐","2026-09-19 05:45:05",{"path":82,"title":83,"description":84,"date":85,"categories":32,"tags":86,"draft":14},"\u002Fposts\u002Fllm-memory-compute-tradeoff","LLM 的 Memory-Compute Trade-off：从 MHA 到 MLA、Sparse 与 Linear Attention","ds4.1f 上线了。它的架构 novel 在哪，便宜和快又是从哪省出来的。","2026-09-11 12:00:00",[34,87,88,89,90],"Attention","MLA","DeepSeek","长上下文",{"path":92,"title":93,"description":94,"date":95,"categories":32,"tags":96,"draft":68},"\u002Fposts\u002Fmarkdown-extensions","Markdown 扩展语法速查","本站 Markdown 支持的扩展语法演示与速查：admonition 提示块、GitHub 仓库卡片、spoiler 遮罩、ABC 乐谱、音频播放器、浏览器可运行代码块。","2026-09-01 15:30:00",[97],"博客",{"path":99,"title":100,"description":101,"date":102,"categories":9,"tags":103,"draft":14},"\u002Fposts\u002Fwhy-we-still-need-to-write-a-blog","为什么还需要手写博客？","AI这么强，还要古法写博客吗？","2026-09-01 13:17:53",[97],{"path":105,"title":106,"description":24,"date":107,"categories":24,"tags":26,"draft":68},"\u002Fposts\u002Fllm-serving-1","LLM Serving（一）:以在H200上部署GLM-5.2为例","2026-07-21 06:15:44",{"path":109,"title":110,"description":24,"date":111,"categories":112,"tags":26,"draft":14},"\u002Fposts\u002Finteresting-music-rhythm","有趣的音乐节奏","2026-07-16 12:47:24","音乐",{"path":114,"title":115,"description":116,"date":117,"categories":42,"tags":118,"draft":14},"\u002Fposts\u002Ftowards-monosemanticity","LLM 可解释性（一）：SAE与单语义","读 Anthropic《Towards Monosemanticity》。从难以解释的神经元出发, 讲清superposition为什么让单个神经元混成一团, SAE怎么把多语义的激活分解成单语义特征。","2026-07-12 22:00:00",[34,119,120],"可解释性","SAE",{"path":122,"title":123,"description":124,"date":125,"categories":9,"tags":126,"draft":68},"\u002Fposts\u002Fpiketty-ai-china-k-shaped","用皮凯蒂的镜头看 AI 与中国 K 型社会","一次凌晨的对话整理。从 r>g 出发，看 AI 时代的资本分配、中国的世袭化趋势、体制内的二元劳动力市场，以及一个普通 00 后在一线打拼的真实牌面。","2026-06-24 03:00:00",[66,127,65],"经济",{"path":129,"title":130,"description":24,"date":131,"categories":24,"tags":132,"draft":68},"\u002Fposts\u002Fclaude-code-compact-behavior","cluade code compact 行为","2026-06-23 20:17:42",[133,134,135],"Agent","AI Agent","Claude Code",{"path":137,"title":138,"description":24,"date":139,"categories":42,"tags":140,"draft":14},"\u002Fposts\u002Fclaude-code-settings","Claude Code 配置分享","2026-06-23 14:22:12",[34,134,133,135],{"path":142,"title":143,"description":144,"date":145,"categories":32,"tags":146,"draft":14},"\u002Fposts\u002Fmy-thoughts-on-the-ai-industry","简单聊聊对 AI Agent 行业的理解","写这篇文章的初衷，是我发觉逐渐忘了前两个月的 AI\u002FLLMs 是什么样的能力和情况了。意识中似乎某些在突然加速，但理性和直觉告诉我这不对，作为行内人，我知道这是一个渐进的、累积的过程。我想自己回忆一下，Agent的发展。","2026-06-20 13:14:25",[34,134,133],{"path":148,"title":149,"description":24,"date":150,"categories":42,"tags":151,"draft":14},"\u002Fposts\u002Frotational-positional-encoding","RoPE","2026-06-17 18:31:40",[34,152,149],"Transformer",{"path":154,"title":155,"description":24,"date":156,"categories":42,"tags":157,"draft":14},"\u002Fposts\u002Fintroduction-to-large-language-model","大语言模型入门基础","2026-06-16 22:36:11",[152,34],{"path":159,"title":160,"description":24,"date":161,"categories":32,"tags":162,"draft":14},"\u002Fposts\u002Fthought-about-coding-agent","关于 Coding Agent 使用的遐思","2026-06-03 17:54:20",[133],{"path":164,"title":165,"description":166,"date":167,"categories":42,"tags":168,"draft":14},"\u002Fposts\u002Fabout-ssh-usage","关于SSH的使用——客户端篇","介绍SSH客户端的用法。","2025-06-17 02:44:57",[169],"工具",{"path":171,"title":172,"description":173,"date":174,"categories":32,"tags":175,"draft":14},"\u002Fposts\u002Fopensourcesoftwarestarttutorial","如何开始一个开源项目？","浅谈我对开源项目的认知。","2025-04-17 18:00:00",[176],"开源",{"path":178,"title":179,"description":180,"date":181,"categories":182,"tags":183,"draft":14},"\u002Fposts\u002Fmotivation","渴求动力","有意识的生活并主动地行动会使得生活更美好。","2025-03-05 00:44:57","生活-情感",[66],{"path":185,"title":186,"description":187,"date":188,"categories":182,"tags":189,"draft":14},"\u002Fposts\u002F2024summary","2024年终总结","啊，痛彻的时光；啊，悲怆的人！","2025-01-01 01:00:00",[66],{"path":191,"title":192,"description":193,"date":194,"categories":42,"tags":195,"draft":14},"\u002Fposts\u002Fvps-crash-startup","快速搭建个人VPS","如何快速搭建个人VPS。","2024-12-31 19:00:00",[196],"VPS",{"path":198,"title":199,"description":200,"date":201,"categories":42,"tags":202,"draft":14},"\u002Fposts\u002Fdata-label-software","AI数据（自动）标注软件概述","对市面上目前的AI数据（自动）标注软件进行调研。","2024-12-20 14:00:00",[203,204],"数据标注","深度学习",{"path":206,"title":207,"description":208,"date":209,"categories":182,"tags":210,"draft":14},"\u002Fposts\u002Fself-destruction","生活的崩坏与自我毁灭","是什么在牵引着我？我的行为，我的思想？我的感受？我所处的地方？它们在何处？","2024-12-19 12:00:00",[66],{"path":212,"title":213,"description":214,"date":215,"categories":42,"tags":216,"draft":14},"\u002Fposts\u002Fcs194-196_llmagents","LLM Agents 课程笔记","UCB CS294\u002F194-196 Large Language Model Agents 的课程笔记。","2024-11-30 20:00:00",[133],{"path":218,"title":219,"description":220,"date":221,"categories":182,"tags":222,"draft":14},"\u002Fposts\u002Fwhatsyourworth","价值、位置","你们的生存究竟有何价值？如果毫无价值，你们究竟为何存在？","2024-11-29 02:00:00",[66],{"path":224,"title":225,"description":226,"date":227,"categories":182,"tags":228,"draft":14},"\u002Fposts\u002Fblogrelifever2","重启博客v2","忙里偷闲学习前端，还是决定结合想法手搓一个自己的博客。","2024-11-25 00:00:00",[97],{"path":230,"title":231,"description":232,"date":233,"categories":182,"tags":26,"draft":14},"\u002Fposts\u002Fblogrelife","重启博客","在我的表达欲与现实所需(问题解决记录)驱动下，我决定重写博客。","2024-10-21 20:44:57",{"id":235,"title":6,"body":236,"categories":9,"date":8,"description":7,"draft":14,"extension":318,"meta":319,"navigation":68,"path":5,"seo":320,"stem":321,"tags":322,"__hash__":323},"posts\u002Fposts\u002Fthe-age-of-hesitation.md",{"type":237,"value":238,"toc":308},"minimark",[239,254,259,262,266,269,272,275,278,281,284,287,290,293,296,299,302,305],[240,241,242],"note",{},[243,244,245,246,253],"p",{},"本文来自视频",[247,248,252],"a",{"href":249,"rel":250},"https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1vnVq6hEFA",[251],"nofollow","真正毁掉一个人的，不是不努力，而是不肯狼狈","的字幕转稿。",[255,256,258],"h2",{"id":257},"为什么人们越来越不敢冒险投入了","为什么人们越来越不敢冒险\u002F投入了",[243,260,261],{},"朋友们，今天我想讨论一个话题：为什么我们现在越来越不敢投入了。很多时候我们以为的选择，比如你有两个工作offer，现在选择好像是我到底应该考公还是考编。明明这两个选项都不是真正的选项，因为你还没有开始，也不知道最后能不能成。这个时候我们就已经开始犹豫了。",[263,264,265],"h3",{"id":265},"人们在害怕失败后的狼狈",[243,267,268],{},"我觉得，比起这个表面上看到的“有可能是我还没有想好哪个更适合我”，比起这样一个表面问题，更多的是我们不敢去投入，因为我们害怕自己如果投入了，最后没有结果，是不是看起来就很狼狈，不那么漂亮光鲜。说到底，很多人追求的不是成功本身，而是成功的社会符号。我们拥有的不只是金钱资本，还有文化、社会、象征资本。比如我们知道，房子、头衔、学历、民企经历都属于象征资本的一部分，它可以快速向周围人传递一个信号：这个人过得还不错。而问题在于，信号和真实的现实能力之间并不是同步增长的。",[243,270,271],{},"从进化心理学角度来看，我们长期生活在一个小规模群体里面，在一个几十人的部落里面，谁拥有资源、地位，谁就更受到尊重。这些信息都直接关系到生存和繁衍，因此我们的大脑天然就会关注这个社会排序。这是因为我们是social animal，这是非常重要的一点。猫或者猴子，它们都有上下等级，这是群居动物的一个特点。当互联网出现之后，事情就发生了变化。现在一个普通人可以和什么比？他可以和这些Twitter或全球范围内最顶尖的人进行比较。",[243,273,274],{},"人会通过和他人的比较来评价自己。这是为什么说Facebook发展的过程当中，伴随着青少年自杀率的上升。因为你突然发现，你都还是个孩子，你就把自己和全球最厉害、过得最好那一批人进行对比，你是不是突然一下觉得，这个生活还有什么意思。",[263,276,277],{"id":277},"他人走过而自己未走过的路让人混乱",[243,279,280],{},"另一个问题，就是我们对于损失极为敏感。当一个人看到同龄人买房、升职、加薪的时候，大脑就会把它理解成我正在失去什么机会。往往真实的情况只是他人在走他自己的路而已。别人得到，不是我们所失去的。",[243,282,283],{},"另外，有一个概念叫delayed gratification，延迟回报。这个长期目标的实现，往往需要承受很大程度上的短期不适。做任何事情几乎都是这样的，因为一个复利系统都具有一个共同特点：你长期的增长很缓慢，后期增长就是陡峭的。而芒格也说过，伟大成果往往来自于长时间积累后的一个非线性增长。而从数学上看，复利曲线最容易让人产生幻觉，因为你前半场几乎看不到任何变化。大多数人离场时间都是发生在复利开始加速之前。因为很多人都想要说，网上有这么多年轻漂亮、名校毕业、一毕业就年薪百万的winner，为什么我不可以试？为什么我要去受这份苦？为什么我就不能have it all？",[263,285,286],{"id":286},"身份认同陷阱",[243,288,289],{},"另外一件，是关于身份认同陷阱，Identity trap。很多人会逐渐把外部标签和自我价值做一个绑定。比如我喜欢说我是什么什么校人，我是某场的谁谁谁，我是管理者，我拥有房产。当这些标签都变成自我认同的核心的时候，我们的决策方法、方式就会发生变化，关注点就开始从成长转向为维护形象。因此就会产生大量的高成本行为，比如你继续待在一个不喜欢的单位，继续不合适自己的生活方式，拒绝从头学习，回避会暴露短板的环境。那么这一句决策能够维护当时当前的身份，却会削弱你未来的成长。",[255,291,292],{"id":292},"价值错位",[243,294,295],{},"那么更多的人越来越强调一个词叫OPTIALITY，选择权。复杂世界里面最大的优势之一，就是来自于拥有更多的选择。选择权来自于什么？技能、健康、人际网络、现金储备、持续学习的能力、跨领域牵引的能力。这些东西，你看看它和我们能够迅速识别的标签，有一些差距，或者说是它是两套系统。它没有这个房子、职位和你的title那么直观，但它却决定了未来20年的行动空间。发现了吗？这些东西它和我们所以为的选择不是同一套东西。",[243,297,298],{},"我们以为的选择，是我应该考这个考研，还是应该考公？我是应该出国，还是应该留在父母身边？它不是这种选择，而是你能够是不是给自己带来足够的冗余，在各个方向上让自己有更多的选择空间、有更大自由度。从这个角度来看，就会发现一个很有趣的转变：年轻的时候，我们可能会问，我现在看起来成功吗？随着你的阅历增加，问题逐渐变成，我是否拥有持续创造价值的能力。",[243,300,301],{},"前一个问题，关注的是展示；后一个问题，关注的是生存。前一个问题、后一个问题，它依赖的是自身能力。因此，人生当中最重要的区别，不在于你体面与否，关键在于一个人到底是在积累资产，还是在维护形象。",[243,303,304],{},"这里的资产，不光是我们理解物质上的资产，而是知识、技能、关系、创造力、健康以及信任。这些资产的增长速度往往都很慢，因此会让人觉得有些笨拙，有些迟缓，有些不像成功人士，不像那些看起来一开始就赢在起跑线上的人。",[243,306,307],{},"但是，很多长期卓越的人，他的发展轨迹都呈现类似特征：他们把大量时间投入到了外界难以评价的积累之中，也就是所说的这些部分。而多年以后，当人们看到那一刻，他突然说，我就知道这个人一定会成功的。你不知道，你只是现在知道了。",{"title":24,"searchDepth":309,"depth":310,"links":311},2,3,[312,317],{"id":257,"depth":309,"text":258,"children":313},[314,315,316],{"id":265,"depth":310,"text":265},{"id":277,"depth":310,"text":277},{"id":286,"depth":310,"text":286},{"id":292,"depth":309,"text":292},"md",{},{"title":6,"description":7},"posts\u002Fthe-age-of-hesitation",[11,12,13],"BQa6QbIeJJz3JTPodOY3pfxw2mdVW3r6Cbs4YYp8MEk",1791601054360]