Posts
All the articles I have published.
-
The Great AI Job Divergence: Programmers Are Being Laid Off, Electricians Are Being Poached, and the Middle Is Disappearing
Published date:12 min readIn 2026 the AI job market is splitting in two: tech hiring is slowing and entry-level white-collar roles are vanishing, while data centers ignite bidding wars for electricians and memory-chip profits explode. This essay unpacks the four mechanisms behind the divergence—cognitive labor oversupply, tactile scarcity premiums, accountability moats, and capex pulses—and ends with a practical framework for where ordinary people should go.
-
AI 时代的就业分化：程序员在被裁，电工在被抢，中间的人在消失
Published date:2 min read2026 年 AI 就业剧烈分化：科技公司放缓招聘、初级白领被 AI 替代，数据中心带火电工水管工、存储芯片利润暴涨。本文讲透分化背后的四个机制——认知供给过剩、触觉稀缺溢价、责任护城河、资本脉冲——并给出普通人的职业选择框架。
-
The AI Slop Flood: Garbage Is the Default, Gold Is Engineered—Great AI Work Isn't Generated, It's Built
Published date:16 min readAI content is flooding every platform, most of it garbage—yet every once in a while a piece stops you cold: real story, real camera work, film-grade texture. This essay explains from first principles why one-click generation is mediocre by default, breaks down the four layers of consistency behind great work, walks through a nine-ring production workflow, and argues that AI creation is, at bottom, the same systematic engineering as filmmaking—where the most valuable layer is always judgment.
-
AIGC 狂潮里的垃圾与金子：好作品不是生成出来的，是工程出来的
Published date:1 min readAIGC 内容泛滥成灾，大部分是垃圾，但总有作品让人眼前一亮——有故事、有运镜、有电影质感。本文从模型原理讲清楚一键生成为什么必然平庸，讲透好作品的四层一致性，展开从剧本到审片的九环节工作流，并论证 AI 创作的本质和电影制作一样：系统性工程，而工程里最值钱的一层永远是判断。
-
企业 AI 落地：老板想要什么，程序员到底在做什么
Published date:2 min read从 BOSS 直聘上的真实对话出发，分析企业 AI 落地的完整路径，以及程序员在这个过程中的真实角色与未来分化方向。
-
Enterprise AI Deployment: What Bosses Want, What Programmers Actually Do
Published date:8 min readFrom real conversations with bosses and tech leads about enterprise AI roles, a deep dive into what it actually takes to deploy AI in a company—and the two career paths emerging for programmers.
-
Agent 真正跑起来，靠的不是工具，而是 loop
Published date:2 min read很多人谈 agent，总在谈工具、MCP、skill。真正决定它是不是一个 agent 的，往往不是这些名词，而是那条让系统一轮一轮往前走的 loop。
-
What Makes an Agent Work Is Not the Tools, but the Loop
Published date:13 min readPeople often talk about agents in terms of tools, MCP, and skills. But what usually determines whether something actually behaves like an agent is the loop that keeps the system moving from one step to the next.
-
Agent Is Not a Feature
Published date:13 min readWhy the agent is not a UI trick or a marketing label, but a new software structure built from runtime, tools, MCP, skills, hooks, plugins, and harnesses.
-
旧需求被瓜分之后，普通人还能做什么
Published date:1 min read过去很多人赚钱，靠的是需求增长和市场空位。现在大部分标准化需求已经被成熟体系占住，普通人的机会不再在大赛道里，而在新的交换缝隙里。
-
人类会失去什么？AI时代的认知危机
Updated date:2 min read从现实、概念、语言到 AI 的层层抽象出发，讨论大语言模型为何强大、为何会失真，以及人在 AI 时代为什么仍要负责把抽象重新接回现实。
-
What Will Humans Lose? The Cognitive Crisis in the Age of AI
Updated date:13 min readA reflection on reality, concepts, language, and AI: why large language models are powerful, where abstraction introduces distortion, and why humans still need to reconnect symbols back to the real world.
