10xAI Wiki

18分钟阅读

Career in the AI Era

吴恩达：AI创业还剩3种能力（2026达沃斯）

核心观点：速度已经不是优势，而是及格线。门槛没了。

1. 重构流程的能力

  • 不是优化旧流程，而是从零设计新流程
  • 判断标准：去掉AI，产品还能不能用？
  • 案例：贷款审批 → 10分钟批贷的全新产品

2. 全栈开发的能力

  • 产品 + 工程 + 运营，一个人能替代一支队伍
  • AI把”怎么做”变得很容易
  • 真正拖慢进度的是”要做什么”的决策

3. 做成产品的能力

  • Demo满天飞的时代，真正稀缺的是把Demo做成产品的能力
  • 案例：Fyxer AI（邮件撰写工具），6个月百万年收入
  • 打通从开发、使用、反馈到优化的整个链路

详见 吴恩达实体页面

Builder as an AI-era Role

Builder 指那些亲手用 AI 工具构建产品、系统和工作流的人，区别于只搬运信息、推动流程的传统角色。

The 35-Year-Old Crisis

China’s tech industry has a well-documented “35-year-old crisis” — the phenomenon where programmers in their mid-30s face reduced hiring prospects, age discrimination, and pressure to transition into management or face career stagnation.

In the AI era, this crisis is being redefined:

EraPrimary ThreatSolution
Pre-AIJunior competitionGain experience, climb ladder
Early AI (2022-2024)AI autocompleteLearn to use AI tools
AI Native (2025+)AI does the task entirelyArchitecture, taste, direction

The 2025-2026 period marks the transition from “AI enhances productivity” to “AI performs the work” — fundamentally changing what employers value.

What the Research Says

Anthropic’s Labor Market Research (March 2026)

From Claude’s parent company’s own report on AI labor impact:

Highest AI exposure occupations (% of tasks automatable):

  • Software engineers: 74.5%
  • Customer service: 70.1%
  • Data entry: 67.1%
  • Market research: 64.8%
  • Financial analysts: 57.2%

Demographic paradox: High-exposure jobs skew toward:

  • Higher education (17.4% have graduate degrees vs 4.5% in no-exposure jobs)
  • Higher pay (32.69/hr vs 22.23/hr)
  • More women (54.4% vs 38.8%)

The “skill-ability” insight: AI replaces tasks that can be decomposed into repeatable skills. Physical, contextual, high-judgment work is safer. The shift is from “skill-based work” to “ability-based work.”

Physical AI Data Labor

Micro1 和 Scale AI 代表另一类职业变化：人类动作、判断和隐性知识正在被采集为机器人与物理 AI 的训练数据。

The Entry-Level Collapse

Most immediate impact: Entry-level positions. Companies are using AI to boost productivity of existing senior employees rather than hiring new juniors. “New job start rate” for 22-25 year olds in high-exposure occupations fell ~14% from 2022 to 2024.

35岁职场危机：突破路径

张勇（阿里前CEO）观点

“35岁到底是分水岭，还是新起点，关键不在年龄，而在有没有持续产出可被使用的价值。”

核心问题：

  • 新人：便宜、好带、可塑
  • 专家：能解决关键问题，值得高价格
  • 中间层：贵，但又不值得那么贵

突破路径

  1. 打造核心资产：从”资深执行”转向”可积累的经验”
  2. 找到杠杆：人的杠杆（带团队）、平台的杠杆（借助公司资源）、产品的杠杆（可变现的方法论）
  3. 拥抱AI：把重复性工作外包给AI，聚焦高价值工作

吴军老师建议（给35岁+职场人）

  • 不要轻易辞职转行AI：AI公司2年内大部分会倒闭
  • 副业+AI公众号：保持稳定工作，积累影响力
  • 核心资产思维：打造可积累的能力，而非出卖时间

The Wage System Collapse Thesis

From “AI时代的真正冲击：不是失业，而是工资制度本身的衰亡” (January 2026):

The argument: AI doesn’t just replace specific jobs — it undermines the economic logic of the entire employment system:

Old logic: Companies pay for time (hours) and skills (certifications, degrees). The credential → job → income pipeline.

New reality: AI provides both time and skills at near-zero marginal cost. If AI can do the work in less time with better quality, paying humans for those same hours makes less economic sense.

Implication: The “skill premium” collapses first (AI has all skills). Then the “time premium” collapses (AI works 24/7 at consistent quality). What remains valuable: taste, judgment, relationships, and accountability.

What Skills Matter in the AI Era?

Declining value:

  • Syntax knowledge (AI writes better code)
  • Manual testing (AI automates it)
  • Documentation writing (AI generates it)
  • Standard implementation (AI handles it)
  • Pattern matching for routine bugs

Rising value:

  • Architecture: Designing systems that AI can build and maintain
  • Taste: Knowing what to build and for whom
  • Integration: Connecting AI systems to real-world workflows
  • Judgment: Deciding when AI is wrong, and what to do about it
  • Communication: Translating business needs to AI instructions

Martin Fowler’s Perspective (December 2025)

In a deep interview, Martin Fowler (Spring Framework co-creator) said:

“AI is the biggest change in software engineering in 40 years… The skills that matter most now are architectural thinking and the ability to communicate clearly with AI.”

He noted that his own work has shifted: instead of writing code, he spends more time on system design and ensuring AI-generated code fits the architecture.

The Path Forward

For programmers in their 30s-40s:

  1. Don’t compete with AI on execution — compete on judgment and experience
  2. Learn Harness Engineering — designing the constraints and systems around AI
  3. Build with AI, not against it — tools like Claude Code amplify experienced engineers dramatically
  4. Develop taste — the ability to identify valuable problems is harder for AI than solving them
  5. Consider the independent route — AI enables one-person companies that compete with large teams

For those entering the field:

  1. Focus on fundamentals over frameworks — AI handles frameworks; fundamentals (algorithms, data structures, system design) provide judgment
  2. Learn AI tooling deeply — Claude Code, Cursor, etc. are now essential skills
  3. Develop non-technical skills early — product sense, communication, understanding user needs
  4. Consider agentic engineering — treating AI agents as team members requires different skills than coding alone

AI Agent与副业：AI时代的搞钱机会

AI Agent搞钱案例

  • 儿童绘本：一句话生成20页绘本 + 网页部署
  • 独立游戏：20分钟完成魂斗罗类游戏
  • 个人网站：AI自动生成并部署上线

AI时代的新机会

  • AI = 你的打工仔
  • 关键是你能不能提供”想法”
  • 从”做事”到”想清楚做什么”

PM-to-Builder Transition: The Great Reshuffle (2026)

From Nikhyl-Singhal’s conversation with Lenny Rachitsky and Cat-Wu’s product leadership insights:

The “Fire 30K, Hire 8K” Cycle

Nikhyl Singhal predicts a massive reshuffle over the next 12-24 months:

  • Companies may fire 30,000 old-paradigm workers and rehire 8,000 AI-first builders
  • Over the past 5 years, many companies doubled headcount but output did not double
  • A “judgment day” is coming where firms realize past hiring didn’t deliver proportional value

Information Mover vs Builder

Information MoverBuilder
Communication, alignment, info flowBuilds things with AI tools
Responsibility without authorityDirectly validates ideas with AI
Brand and resume matter”How modern you are” matters more
Becoming a “dinosaur”Market’s most sought-after

About half of PMs grew into Information Movers during the ZIRP era. The other half are natural Builders — and they are the ones being chased by the market.

The Psychological Barrier

The hardest part is not skills, but psychology:

  • “Shadow superpower”: the more successful you were in the old system, the harder to change
  • For 30-somethings, “power years” are a cruel coincidence — career peak meets maximum life burden (family, kids, aging parents)
  • Past success creates resistance to reinvention

Cat Wu’s 9 Rules for AI-Era PMs

From Cat-Wu’s Lenny’s Podcast interview:

  1. Dev cycle: 6 months → 1 day — Anthropic ships features from idea to production in one day by removing all barriers to shipping
  2. Stop aligning, start acting — Weekly data reviews + clear team principles replace endless cross-team alignment meetings
  3. Build for next month’s model — Judge capability gaps and when they’ll close; start early, launch when model catches up
  4. Ship as “research preview” — Explicitly label early versions to lower psychological barrier to shipping
  5. Most important new skill: ask AI “why were you wrong?” — Introspection as product evaluation tool
  6. Code speed is worthless; judgment is priceless — PM’s core value is product taste, not pushing development
  7. Don’t write PRDs — One page max: goals, users, failure modes. Then build.
  8. Give AI briefings, not wishes — Feed AI marketing outlines, Slack discussions, templates first; let it propose 3 directions, then iterate
  9. Most PMs she interviews don’t get it — AI-era PMs should be able to build things themselves. Motto: “just do things"

"Fun is the Antidote to Burnout”

Nikhyl observes: every builder eventually has a “first joy moment” — the instant they realize AI is building something for them. That moment converts fear into addiction. Product management is shifting from “process + management fatigue” to “building + creation joy.”

![The shift: from smiling exhaustion to actually enjoying work]

Source: 2026-05-04-Silicon-Valley-AI-first-reshuffle-PM-to-Builder

[2026-07-17] Zara底层工作方式：工作方法与底层能力重塑

  • 张咋啦 Zara（前字节leader，中美AI圈”几乎最出名的Builder”）2026-07-12飞书闭门会系统阐述”AI时代真正发生变化的工作方式”。本块聚焦工作方法与底层能力重塑，与下方E-cluster的35岁年龄定价维度是不同切面。
  • Builder三能力：agency、taste、distribution——agency（主观能动性，不等别人拆好任务，自己发现问题发起尝试）、taste（判断力，面对AI选项能看出哪个像样并说清原因）、distribution（分发能力，把故事讲清楚让该看到的人真的看到并愿意行动）。与本页吴恩达”重构流程/全栈开发/做成产品”三能力对照：Zara是个体层能力模型，吴恩达是组织层能力模型，层次互补。
  • 代码便宜，注意力贵：“代码变便宜了，注意力变贵了。把东西做出来越来越快，决定做什么、怎么让别人愿意在意，反而变得更难。“最大卡点是营销和推广，获取人的注意力越来越难。
  • AI slop诊断：AI slop根源不是格式问题，是”没有灵魂”——AI讲的是”平均值的平均值”，脱离人生体验。去AI味的Skill只解决表象，真正的解药是把真实人生体验放进上下文。“识别AI slop是一项基础能力，看不出内容哪里空，大概率也会生产同样的东西。“与本页”rising value: taste/judgment”共鸣。
  • 文档即协作界面：复杂内容整理一份飞书文档当brief（工作说明书），列放脚本/Codex写可视化方案/生成HTML，不符合预期直接选中评论让agent改。核心论断：“文档已经不只是给人阅读的交付物，它是人与agent共同工作的界面：素材进来，方案展开，反馈落点，修改再回到产品或视频里。“与本页Cat Wu第8条”Give AI briefings, not wishes”共鸣——Zara的飞书文档brief正是Cat Wu第8条的工程化落地。
  • 团队级agent重塑组织沟通：团队级agent拥有团队级上下文，不依附某一台电脑或某一个人。Shopify创始人做法——内部agent只在公开频道和员工交流，没有私聊入口（“私聊里的信息agent看不见，无法成为组织可检索的上下文”）。与本页Martin Fowler”架构思维+与AI沟通的能力”共鸣：Fowler说架构思维，Zara说组织沟通方式也要随agent调整。（注：Shopify无entity页，plain text）
  • 审美沉淀成Skill：先让agent完成真实设计任务，耐心调几十轮，“审美从一个人脑子里的模糊感觉，变成团队可以调用、检查和继续改进的工作资产”。

来源：../sources/2026-07-17-AI行业动态-智谱摸高-WAIC-工作方式

[2026-07-17] 刘润”35岁不是危机，而是一次重新定价”

  • 刘润（“进化岛”9周年直播，与润宇老师对谈）核心论点：35岁不是危机，而是一次重新定价。本块聚焦年龄与定价机制，与上方C-cluster Zara工作方法块是不同切面。
  • 体力vs判断力定价：医生的判断力被单独定价（越老越值钱），程序员的判断力则与体力打包定价。AI替代了程序员工作中的”体力活”（写代码），使程序员开始走向医生的定价机制——判断力单独被定价。医生和律师没有35岁危机，因为”医生的定价，主要参考判断力”。
  • 43岁ITBP实例：刘润举实例——招了一位43岁ITBP（IT Business Partner），写了20多年代码，岗位是不写代码、把业务翻译成技术架构让AI生成代码。每项被替代的体力活都在给判断力腾地方。
  • “岗位不会被替代。任务才会被替代。” 越是纯粹的体力任务越容易被替代，销售做PPT、写文书、录数据的活被替代，位置和时间让给察言观色、建立信任、关键决策。“35岁最大的机会，在’判断力’上。”
  • 25/35/45分层：25岁——老一辈”看不起”的地方才是机会（“给你留了位子的，恰恰是那些’这是什么破玩意儿’的地方”；上一种能力未必是新能力的基础，如纺锤→纺纱机、跑步→骑马→开车是三套独立技能）；35岁——卷判断力；45岁——从能力走向资源，手上的稀缺资源（人脉、品牌、资本、信用）“年轻人没有，AI也造不出来”，最大敌人是”这不就是”四个字。
  • 引人类学家玛格丽特·米德”后喻时代”概念：前喻时代是老人教新人，后喻时代是老人学新人。45岁机会在于克服”这不就是”的心智闭环——用旧理论解释新事物导致过早闭合。
  • 与本页现有”35岁危机：突破路径”框架是视角延伸而非矛盾：现有框架说”怎么办”（打造核心资产/找杠杆/拥抱AI），刘润说”为什么”（定价机制为何转变）。两者共同指向判断力。

来源：../sources/2026-07-17-个人成长与学习方法

关系图谱

反向链接