Featured
-
Who Is Building World Models — and What Are They Betting?
Published date:9 min readFour groups are staking the world-model field: giants hedging with their own models, star founders with billion-dollar war chests, specialists compounding data advantages, and China's driving-first track. The pattern across all four says more than any single announcement.
Recent
[6/ 97]-
How Do World Models Work? Four Ways AI Learns to Imagine
Updated date:10 min readA world model predicts what it will see next, given what it sees now and what it does. The interesting part is how: four different technical routes, each answering 'what should the model predict?' in a different way — with real systems already running on each.
-
World Model vs. LLM: What's Actually Different, and Why It Matters
Published date:10 min readAn LLM predicts the next piece of text; a world model predicts the next state of an environment. The distinction sounds small and isn't — it explains where AI works today, where it fails, and why the biggest labs are spending billions on both.
-
Why Do Language Models Hallucinate? The Confident-Guesser Problem
Published date:8 min readChatbots make things up because of how they are built and how they are graded, not because of a bug someone forgot to fix. The mechanism, the exam-room incentives behind it, and why it can't simply be patched away.
-
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.
