<?xml version="1.0" encoding="utf-8"?>
<feed xmlns="http://www.w3.org/2005/Atom">
  <author>
    <name>Liam</name>
  </author>
  <generator uri="https://hexo.io/">Hexo</generator>
  <icon>https://blog.naaln.com/icon.png</icon>
  <id>https://blog.naaln.com/</id>
  <link href="https://blog.naaln.com/" rel="alternate"/>
  <link href="https://blog.naaln.com/atom.xml" rel="self"/>
  <rights>All rights reserved 2026, Liam</rights>
  <subtitle>AI、产品、数据、旅行与生活观察</subtitle>
  <title>Liam's Notes</title>
  <updated>2026-07-31T06:36:12.000Z</updated>
  <entry>
    <author>
      <name>Liam</name>
    </author>
    <category term="Notes &amp; Reflections" scheme="https://blog.naaln.com/categories/Notes-Reflections/"/>
    <category term="Infrastructure" scheme="https://blog.naaln.com/tags/Infrastructure/"/>
    <category term="Workflow" scheme="https://blog.naaln.com/tags/Workflow/"/>
    <category term="Newsletter" scheme="https://blog.naaln.com/tags/Newsletter/"/>
    <category term="AICoding" scheme="https://blog.naaln.com/tags/AICoding/"/>
    <id>https://blog.naaln.com/2026/08/newsletter-132/</id>
    <link href="https://blog.naaln.com/2026/08/newsletter-132/"/>
    <published>2026-08-01T02:00:00.000Z</published>
    <summary>大家都在把什么都丢给大模型，却很少算过账——那只是把「读不完」的成本换成了「账单读不完」。当算力可能涨十倍，真正的手艺是把 LLM 放对位置：让它定义标准而不是每天执行标准，去噪去重排序交给便宜稳定的「古法 AI」。省下的从来不是模型，是位置。</summary>
    <title>L132_把模型放对位置</title>
    <updated>2026-07-31T06:36:12.000Z</updated>
  </entry>
  <entry>
    <author>
      <name>Liam</name>
    </author>
    <category term="Notes &amp; Reflections" scheme="https://blog.naaln.com/categories/Notes-Reflections/"/>
    <category term="Trust" scheme="https://blog.naaln.com/tags/Trust/"/>
    <category term="Newsletter" scheme="https://blog.naaln.com/tags/Newsletter/"/>
    <category term="AgenticAI" scheme="https://blog.naaln.com/tags/AgenticAI/"/>
    <category term="FutureOfWork" scheme="https://blog.naaln.com/tags/FutureOfWork/"/>
    <id>https://blog.naaln.com/2026/07/newsletter-131/</id>
    <link href="https://blog.naaln.com/2026/07/newsletter-131/"/>
    <published>2026-07-25T02:00:00.000Z</published>
    <summary>AI 越强，人却越忙——当执行成本趋近于零，人被挤到任务的两头：开头设定问题、给框架，结尾判断质量、接住信任与责任。从 human sandwich 到 25/50/25，聊聊价值为什么从「会执行」迁移到「会判断」。</summary>
    <title>L131_越强越忙</title>
    <updated>2026-07-31T06:35:24.000Z</updated>
  </entry>
  <entry>
    <author>
      <name>Liam</name>
    </author>
    <category term="Notes &amp; Reflections" scheme="https://blog.naaln.com/categories/Notes-Reflections/"/>
    <category term="Life" scheme="https://blog.naaln.com/tags/Life/"/>
    <category term="Memory" scheme="https://blog.naaln.com/tags/Memory/"/>
    <category term="Reflection" scheme="https://blog.naaln.com/tags/Reflection/"/>
    <category term="Time" scheme="https://blog.naaln.com/tags/Time/"/>
    <id>https://blog.naaln.com/2026/07/how-to-time-travel/</id>
    <link href="https://blog.naaln.com/2026/07/how-to-time-travel/"/>
    <published>2026-07-21T02:00:00.000Z</published>
    <summary>为什么日子越过越快，而某几段时光回头看却格外漫长？这篇随笔从「重复不产生记忆，新体验才产生记忆」讲起，聊聊时间感知的错觉，以及如何靠旅行和新体验给自己的人生「偷偷续上几秒钟」。</summary>
    <title>如何进行一场时间旅行</title>
    <updated>2026-08-01T16:54:26.473Z</updated>
  </entry>
  <entry>
    <author>
      <name>Liam</name>
    </author>
    <category term="Notes &amp; Reflections" scheme="https://blog.naaln.com/categories/Notes-Reflections/"/>
    <category term="Aesthetics" scheme="https://blog.naaln.com/tags/Aesthetics/"/>
    <category term="Newsletter" scheme="https://blog.naaln.com/tags/Newsletter/"/>
    <category term="AgenticAI" scheme="https://blog.naaln.com/tags/AgenticAI/"/>
    <category term="FutureOfWork" scheme="https://blog.naaln.com/tags/FutureOfWork/"/>
    <id>https://blog.naaln.com/2026/07/newsletter-130/</id>
    <link href="https://blog.naaln.com/2026/07/newsletter-130/"/>
    <published>2026-07-18T02:00:00.000Z</published>
    <summary>当 AI 让「做东西」变得几乎免费，稀缺的就不再是产能，而是判断、品味和那层「谁来把方向盘」的引导层——价值正从模型本身，挪到模型之外。</summary>
    <title>L130_价值在模型之外</title>
    <updated>2026-07-31T06:34:47.000Z</updated>
  </entry>
  <entry>
    <author>
      <name>Liam</name>
    </author>
    <category term="Notes &amp; Reflections" scheme="https://blog.naaln.com/categories/Notes-Reflections/"/>
    <category term="Trust" scheme="https://blog.naaln.com/tags/Trust/"/>
    <category term="Newsletter" scheme="https://blog.naaln.com/tags/Newsletter/"/>
    <category term="AgenticAI" scheme="https://blog.naaln.com/tags/AgenticAI/"/>
    <category term="FutureOfWork" scheme="https://blog.naaln.com/tags/FutureOfWork/"/>
    <id>https://blog.naaln.com/2026/07/newsletter-129/</id>
    <link href="https://blog.naaln.com/2026/07/newsletter-129/"/>
    <published>2026-07-11T02:00:00.000Z</published>
    <summary>从一位创业者被 AI 清空的硬盘说起，串起本周三份周刊与 GPT-5.6 发布：当机器越来越能干，人真正不可替代的不是能力上限，而是「分寸」的下限——什么时候该退、该问、该把方向盘交回给人。</summary>
    <title>L129_AI 时代的分寸</title>
    <updated>2026-07-31T06:34:01.000Z</updated>
  </entry>
  <entry>
    <author>
      <name>Liam</name>
    </author>
    <category term="Notes &amp; Reflections" scheme="https://blog.naaln.com/categories/Notes-Reflections/"/>
    <category term="AI" scheme="https://blog.naaln.com/tags/AI/"/>
    <category term="Newsletter" scheme="https://blog.naaln.com/tags/Newsletter/"/>
    <category term="AgenticAI" scheme="https://blog.naaln.com/tags/AgenticAI/"/>
    <category term="Ethics" scheme="https://blog.naaln.com/tags/Ethics/"/>
    <id>https://blog.naaln.com/2026/07/newsletter-128/</id>
    <link href="https://blog.naaln.com/2026/07/newsletter-128/"/>
    <published>2026-07-04T02:00:00.000Z</published>
    <summary>从「哈勃半径」三层记忆架构到 Anthropic 追踪代码争议，从企业 AI「三脑问题」到 Figma 结构性危机——本周 7 篇周报与 12 条 AI 快讯，探讨 AI 时代的用户主权：你的可观测宇宙有多大。</summary>
    <title>L128_你的可观测宇宙</title>
    <updated>2026-07-02T09:49:34.000Z</updated>
  </entry>
  <entry>
    <author>
      <name>Liam</name>
    </author>
    <category term="AI &amp; Systems" scheme="https://blog.naaln.com/categories/AI-Systems/"/>
    <category term="KnowledgeManagement" scheme="https://blog.naaln.com/tags/KnowledgeManagement/"/>
    <category term="Workflow" scheme="https://blog.naaln.com/tags/Workflow/"/>
    <category term="Search" scheme="https://blog.naaln.com/tags/Search/"/>
    <category term="AI" scheme="https://blog.naaln.com/tags/AI/"/>
    <id>https://blog.naaln.com/2026/07/hubble-radius-ai-memory/</id>
    <link href="https://blog.naaln.com/2026/07/hubble-radius-ai-memory/"/>
    <published>2026-07-02T02:00:00.000Z</published>
    <summary>从 RAG 到 LLM Wiki 再到哈勃半径，聊聊 AI 三层记忆架构的设计思路，以及为什么「我可能知道的东西」比「我知道的东西」更值得关注。</summary>
    <title>给 AI 画一个哈勃半径：三层记忆架构的个人思考</title>
    <updated>2026-07-02T05:25:53.000Z</updated>
  </entry>
  <entry>
    <author>
      <name>Liam</name>
    </author>
    <category term="Notes &amp; Reflections" scheme="https://blog.naaln.com/categories/Notes-Reflections/"/>
    <category term="PersonalGrowth" scheme="https://blog.naaln.com/tags/PersonalGrowth/"/>
    <category term="Reading" scheme="https://blog.naaln.com/tags/Reading/"/>
    <category term="ProductThinking" scheme="https://blog.naaln.com/tags/ProductThinking/"/>
    <category term="CareerStrategy" scheme="https://blog.naaln.com/tags/CareerStrategy/"/>
    <id>https://blog.naaln.com/2026/06/product-builder-books/</id>
    <link href="https://blog.naaln.com/2026/06/product-builder-books/"/>
    <published>2026-06-30T02:00:00.000Z</published>
    <summary>整理自 Lenny Rachitsky 的产品人必读书单，涵盖沟通、战略、设计、管理、产品思维等 10 个方向共 50 余本书，每本附豆瓣链接与个人点评。</summary>
    <title>产品人的 50 本必读书单：从 Lenny 的书柜说起</title>
    <updated>2026-08-01T16:54:26.473Z</updated>
  </entry>
  <entry>
    <author>
      <name>Liam</name>
    </author>
    <category term="Notes &amp; Reflections" scheme="https://blog.naaln.com/categories/Notes-Reflections/"/>
    <category term="Agent" scheme="https://blog.naaln.com/tags/Agent/"/>
    <category term="Newsletter" scheme="https://blog.naaln.com/tags/Newsletter/"/>
    <category term="AICoding" scheme="https://blog.naaln.com/tags/AICoding/"/>
    <category term="FutureOfWork" scheme="https://blog.naaln.com/tags/FutureOfWork/"/>
    <id>https://blog.naaln.com/2026/06/newsletter-127/</id>
    <link href="https://blog.naaln.com/2026/06/newsletter-127/"/>
    <published>2026-06-26T02:00:00.000Z</published>
    <summary>从 Loop Engineering 到微信 AI 专属卡，当 AI 从等待指令变成主动循环，我们的角色从写代码的人变成了设计规则、验证结果的人。本期深度拆解这一范式跳跃。</summary>
    <title>L127_Loop Engineering与AI自主循环</title>
    <updated>2026-06-26T10:17:23.000Z</updated>
  </entry>
  <entry>
    <author>
      <name>Liam</name>
    </author>
    <category term="Notes &amp; Reflections" scheme="https://blog.naaln.com/categories/Notes-Reflections/"/>
    <category term="DecisionMaking" scheme="https://blog.naaln.com/tags/DecisionMaking/"/>
    <category term="AI" scheme="https://blog.naaln.com/tags/AI/"/>
    <category term="Agent" scheme="https://blog.naaln.com/tags/Agent/"/>
    <category term="Newsletter" scheme="https://blog.naaln.com/tags/Newsletter/"/>
    <id>https://blog.naaln.com/2026/06/newsletter-126/</id>
    <link href="https://blog.naaln.com/2026/06/newsletter-126/"/>
    <published>2026-06-20T02:00:00.000Z</published>
    <summary>300 个 Agent 管理跨境电商全流程，但操作者说真正的价值在选品判断和市场押注。AI 把执行门槛压到接近零，判断力的缺口反而更大。本期还关注：消费级 AI 的哑铃市场、Figure 机器人首超人类员工、开源 AI 政策之争。</summary>
    <title>L126_做越来越便宜，决定做什么越来越贵</title>
    <updated>2026-06-26T10:17:32.000Z</updated>
  </entry>
  <entry>
    <author>
      <name>Liam</name>
    </author>
    <category term="AI &amp; Systems" scheme="https://blog.naaln.com/categories/AI-Systems/"/>
    <category term="KnowledgeManagement" scheme="https://blog.naaln.com/tags/KnowledgeManagement/"/>
    <category term="AI" scheme="https://blog.naaln.com/tags/AI/"/>
    <category term="Collaboration" scheme="https://blog.naaln.com/tags/Collaboration/"/>
    <category term="FutureOfWork" scheme="https://blog.naaln.com/tags/FutureOfWork/"/>
    <id>https://blog.naaln.com/2026/06/cooperative-steering/</id>
    <link href="https://blog.naaln.com/2026/06/cooperative-steering/"/>
    <published>2026-06-14T06:00:00.000Z</published>
    <summary>当团队每个人都用各自的 AI 助手独立工作时，产品一致性正被悄然侵蚀。从 Luke Wroblewski 的「协同驾驶」概念出发，探讨团队级 AI 上下文管理如何成为下一代工程基础设施。</summary>
    <title>当每个人都有自己的 Copilot</title>
    <updated>2026-06-15T11:04:55.000Z</updated>
  </entry>
  <entry>
    <author>
      <name>Liam</name>
    </author>
    <category term="Product &amp; Intelligence" scheme="https://blog.naaln.com/categories/Product-Intelligence/"/>
    <category term="ProductThinking" scheme="https://blog.naaln.com/tags/ProductThinking/"/>
    <category term="UserExperience" scheme="https://blog.naaln.com/tags/UserExperience/"/>
    <category term="DesignPhilosophy" scheme="https://blog.naaln.com/tags/DesignPhilosophy/"/>
    <category term="AI" scheme="https://blog.naaln.com/tags/AI/"/>
    <id>https://blog.naaln.com/2026/06/great-design/</id>
    <link href="https://blog.naaln.com/2026/06/great-design/"/>
    <published>2026-06-14T02:00:00.000Z</published>
    <summary>从 Adobe 设计团队的视角出发，探讨 AI 设计工具被文本提示框住的困境——创作者的直觉和手感不该被压缩成一段 prompt，下一代工具的革命性在于让 AI 成为环境级助手而非命令行。</summary>
    <title>光靠嘴皮子，搞不出好设计</title>
    <updated>2026-06-15T11:04:59.000Z</updated>
  </entry>
  <entry>
    <author>
      <name>Liam</name>
    </author>
    <category term="Notes &amp; Reflections" scheme="https://blog.naaln.com/categories/Notes-Reflections/"/>
    <category term="OpenSource" scheme="https://blog.naaln.com/tags/OpenSource/"/>
    <category term="Agent" scheme="https://blog.naaln.com/tags/Agent/"/>
    <category term="FutureOfWork" scheme="https://blog.naaln.com/tags/FutureOfWork/"/>
    <category term="Economy" scheme="https://blog.naaln.com/tags/Economy/"/>
    <id>https://blog.naaln.com/2026/06/newsletter-125/</id>
    <link href="https://blog.naaln.com/2026/06/newsletter-125/"/>
    <published>2026-06-13T02:00:00.000Z</published>
    <summary>Anthropic与OpenAI同周秘密IPO，中国开源模型成本再探新低，OpenRouter Fusion证明半价组合可击败旗舰。当智力的价格以周为单位崩塌，真正的问题是：谁来定义值多少钱？</summary>
    <title>L125_谁在定义AI的价格</title>
    <updated>2026-06-26T10:17:35.000Z</updated>
  </entry>
  <entry>
    <author>
      <name>Liam</name>
    </author>
    <category term="Notes &amp; Reflections" scheme="https://blog.naaln.com/categories/Notes-Reflections/"/>
    <category term="Product" scheme="https://blog.naaln.com/tags/Product/"/>
    <category term="DesignPhilosophy" scheme="https://blog.naaln.com/tags/DesignPhilosophy/"/>
    <category term="Agent" scheme="https://blog.naaln.com/tags/Agent/"/>
    <category term="FutureOfWork" scheme="https://blog.naaln.com/tags/FutureOfWork/"/>
    <id>https://blog.naaln.com/2026/06/newsletter-124/</id>
    <link href="https://blog.naaln.com/2026/06/newsletter-124/"/>
    <published>2026-06-06T14:30:00.000Z</published>
    <summary>当AI让每个人都能交出可运行的demo，品味、判断力与框架创造力才是真正的稀缺资源。从黑客松评委的错位感到设计工具的终局，再到浏览器入口从「去哪里」到「做什么」的重构。</summary>
    <title>L124_demo泛滥，判断力升值</title>
    <updated>2026-06-26T10:17:37.000Z</updated>
  </entry>
  <entry>
    <author>
      <name>Liam</name>
    </author>
    <category term="Product &amp; Intelligence" scheme="https://blog.naaln.com/categories/Product-Intelligence/"/>
    <category term="ProductThinking" scheme="https://blog.naaln.com/tags/ProductThinking/"/>
    <category term="Startup" scheme="https://blog.naaln.com/tags/Startup/"/>
    <category term="AI" scheme="https://blog.naaln.com/tags/AI/"/>
    <category term="IndieHacking" scheme="https://blog.naaln.com/tags/IndieHacking/"/>
    <id>https://blog.naaln.com/2026/05/find-startup-ideas-in-ai-era/</id>
    <link href="https://blog.naaln.com/2026/05/find-startup-ideas-in-ai-era/"/>
    <published>2026-05-31T06:00:00.000Z</published>
    <summary>当 AI 把执行力的门槛降到几乎为零，真正稀缺的就不再是&quot;能不能做出来&quot;，而是&quot;该不该做&quot;。本文分享了我对 AI 时代寻找创业/Side Project 方向的一些思考，以及用 AI 自动化信号扫描的实践。</summary>
    <title>瓶颈前移：AI 时代怎么找到值得做的事</title>
    <updated>2026-08-01T16:54:26.471Z</updated>
  </entry>
  <entry>
    <author>
      <name>Liam</name>
    </author>
    <category term="AI &amp; Systems" scheme="https://blog.naaln.com/categories/AI-Systems/"/>
    <category term="SystemsThinking" scheme="https://blog.naaln.com/tags/SystemsThinking/"/>
    <category term="Startup" scheme="https://blog.naaln.com/tags/Startup/"/>
    <category term="AI" scheme="https://blog.naaln.com/tags/AI/"/>
    <category term="Agent" scheme="https://blog.naaln.com/tags/Agent/"/>
    <id>https://blog.naaln.com/2026/05/ai-self-improving-company/</id>
    <link href="https://blog.naaln.com/2026/05/ai-self-improving-company/"/>
    <published>2026-05-31T05:00:00.000Z</published>
    <summary>两场 YC 分享引发的思考——AI 不是旧公司的新发动机，而是新公司的操作系统。从 Copilot 到 Capability，从开环到闭环，聊聊一家公司怎么在创始人睡觉时变好。</summary>
    <title>YC 说的「自我进化公司」到底是什么</title>
    <updated>2026-08-01T16:54:26.471Z</updated>
  </entry>
  <entry>
    <author>
      <name>Liam</name>
    </author>
    <category term="AI &amp; Systems" scheme="https://blog.naaln.com/categories/AI-Systems/"/>
    <category term="SystemsThinking" scheme="https://blog.naaln.com/tags/SystemsThinking/"/>
    <category term="Startup" scheme="https://blog.naaln.com/tags/Startup/"/>
    <category term="AI" scheme="https://blog.naaln.com/tags/AI/"/>
    <category term="Agent" scheme="https://blog.naaln.com/tags/Agent/"/>
    <id>https://blog.naaln.com/2026/05/self-improving-company-with-ai/</id>
    <link href="https://blog.naaln.com/2026/05/self-improving-company-with-ai/"/>
    <published>2026-05-31T05:00:00.000Z</published>
    <summary>两场 YC 分享引发的思考——AI 不是旧公司的新发动机，而是新公司的操作系统。从 Copilot 到 Capability，从开环到闭环，聊聊一家公司怎么在创始人睡觉时变好。</summary>
    <title>别再把 AI 当工具了：YC 说的「自我进化公司」到底是什么</title>
    <updated>2026-06-15T11:04:47.000Z</updated>
  </entry>
  <entry>
    <author>
      <name>Liam</name>
    </author>
    <category term="Product &amp; Intelligence" scheme="https://blog.naaln.com/categories/Product-Intelligence/"/>
    <category term="Startup" scheme="https://blog.naaln.com/tags/Startup/"/>
    <category term="AI" scheme="https://blog.naaln.com/tags/AI/"/>
    <category term="Agent" scheme="https://blog.naaln.com/tags/Agent/"/>
    <category term="AINative" scheme="https://blog.naaln.com/tags/AINative/"/>
    <id>https://blog.naaln.com/2026/05/ai-native-startup-playbook/</id>
    <link href="https://blog.naaln.com/2026/05/ai-native-startup-playbook/"/>
    <published>2026-05-31T03:00:00.000Z</published>
    <summary>Anthropic《AI 原生创业手册》解读——当建造成本趋近于零，验证纪律比执行力更重要。从 Idea 到 Scale，一个人如何用 Agent 造一家公司。</summary>
    <title>2026 年了，一个人就能造一家公司——聊聊 AI 原生创业的新范式</title>
    <updated>2026-08-01T16:54:26.471Z</updated>
  </entry>
  <entry>
    <author>
      <name>Liam</name>
    </author>
    <category term="Product &amp; Intelligence" scheme="https://blog.naaln.com/categories/Product-Intelligence/"/>
    <category term="SystemsThinking" scheme="https://blog.naaln.com/tags/SystemsThinking/"/>
    <category term="FirstPrinciples" scheme="https://blog.naaln.com/tags/FirstPrinciples/"/>
    <category term="AI" scheme="https://blog.naaln.com/tags/AI/"/>
    <category term="Investment" scheme="https://blog.naaln.com/tags/Investment/"/>
    <id>https://blog.naaln.com/2026/05/ai-investment-bottleneck-migration/</id>
    <link href="https://blog.naaln.com/2026/05/ai-investment-bottleneck-migration/"/>
    <published>2026-05-31T01:00:00.000Z</published>
    <summary>AI 投资的核心是追踪系统瓶颈的迁移而非追热点。从训练算力不足到 Memory Wall，从 HBM、封装外溢到电力与冷却，再到推理成本与端侧 AI，每个阶段的超额收益都出现在最卡脖子的环节。文章用 AI Factory 框架拆解五阶段迁移逻辑。</summary>
    <title>AI 投资的核心逻辑：追踪系统瓶颈的迁移</title>
    <updated>2026-08-01T16:54:26.471Z</updated>
  </entry>
  <entry>
    <author>
      <name>Liam</name>
    </author>
    <category term="Product &amp; Intelligence" scheme="https://blog.naaln.com/categories/Product-Intelligence/"/>
    <category term="SystemsThinking" scheme="https://blog.naaln.com/tags/SystemsThinking/"/>
    <category term="FirstPrinciples" scheme="https://blog.naaln.com/tags/FirstPrinciples/"/>
    <category term="AI" scheme="https://blog.naaln.com/tags/AI/"/>
    <category term="Investment" scheme="https://blog.naaln.com/tags/Investment/"/>
    <id>https://blog.naaln.com/2026/05/ai-investment-bottleneck/</id>
    <link href="https://blog.naaln.com/2026/05/ai-investment-bottleneck/"/>
    <published>2026-05-31T01:00:00.000Z</published>
    <summary>AI 投资的核心是追踪系统瓶颈迁移而非追热点。从训练算力到 Memory Wall，从 HBM 封装到电力冷却，再到推理成本与端侧 AI，超额收益总出现在最卡脖子的环节。用 AI Factory 框架拆解五阶段迁移逻辑。</summary>
    <title>追踪系统瓶颈的迁移</title>
    <updated>2026-08-01T16:54:26.471Z</updated>
  </entry>
</feed>
