12个人，让我记住了一件事Twelve People, One Lesson I Won't Forget

今天 Demo Day，12个学员分两组上下半场演示，大家都意犹未尽，花了3个多小时。Demo Day today: twelve students presented in two halves, and everyone wanted more — we ran over three hours.

结果台上站的，是餐馆老板、金融顾问、新闻工作者、医疗研究员、硬件工程师、铝材料交易商。没有专业程序员出身的。Who was on stage? A restaurant owner, a financial advisor, a journalist, a medical researcher, a hardware engineer, an aluminum trader. Not one professional programmer.

他们做出来的东西给我惊艳，好几个我自己想不到要去解决，更不会想到这样解决。What they built genuinely amazed me. Several tackled problems I'd never thought to solve — let alone solve that way.

专业背景，比技术背景更值钱Domain Expertise Beats Technical Background

Danny 是硬件工程师。他搭的 LLM 知识库能做根因分析、输出分析报告，效果完全达到同行业资深工程师的水平。Danny is a hardware engineer. His LLM knowledge base performs root-cause analysis and produces reports at the level of a senior engineer in his field.

毅勇是医疗研究行业专家。他的知识库可以给出专业医药研究的产品研发路线。Yiyong is a medical research expert. His knowledge base maps out professional pharmaceutical R&D roadmaps.

Tracy 是金融专家。她在开源项目基础上，搭了一个投资委员会，各个角色互相辩论，输出专业投资建议。Tracy is a finance expert. Building on an open-source project, she assembled an "investment committee" of AI roles that debate each other and produce professional investment advice.

他们做的不是通用工具，是把自己十几年的专业判断，装进了一个 AI 可以调用的系统里。知识库的质量，取决于谁在填它。None of these are generic tools. They packed a decade-plus of professional judgment into systems AI can call on. The quality of a knowledge base depends on who fills it.

不懂代码，不是障碍Not Knowing How to Code Is Not a Barrier

Dottie 是新闻工作者，她自己说是计算机小白。她做了 Social Media Post Pipeline，解决了工作中几个真实的痛点，新客户沟通、内容优化都有了。Dottie is a journalist — a self-described computer novice. She built a social media post pipeline that solves real pain points in her work: new client communication, content optimization, all covered.

Sumei 从国内连线过来，上完课，短短几周做了10多个项目，管潜在客户用工作流，管铝材料专业交易用知识库。Sumei joined live from China. In just a few weeks after class she built over ten projects — workflows for managing prospects, a knowledge base for professional aluminum trading.

以前定制这类系统，要找程序员。现在 Sumei 自己问下 Claude 轻松搞定。Systems like these used to require hiring a programmer. Now Sumei just asks Claude and gets it done herself.

这不是 AI 降低了门槛，是门槛变了。技术不再是瓶颈，领域知识才是。It's not that AI lowered the bar — the bar moved. Technology is no longer the bottleneck; domain knowledge is.

重复的事，可以跑起来Repetitive Work Can Run on Its Own

Zhong 是餐馆老板。他用5个 Agent 做出了专业视频，素材库、知识库、PPT 都准备好了，能熟练调用课上学过的各种工具。Zhong owns a restaurant. He used five agents to produce professional videos — asset library, knowledge base, and slides all in place, fluently wielding every tool from class.

Robin 做7个 Agent，从跑步录音到长短视频发布，链路自动跑完。他还在不断优化流程。Robin built seven agents that take him from a running-session voice memo all the way to published long- and short-form videos, fully automated. He's still optimizing the pipeline.

Song 是数据科学家，科普文章 Pipeline 做得非常严谨，质量可控，对第一堂课的内容举一反三。Song is a data scientist. His science-writing pipeline is rigorous and quality-controlled — taking Session 1's material and running far beyond it.

三个人的方向不同，但有一个共同点，把一个复杂的多步骤流程，固化成可以反复跑的工作流。以前要花几个小时，现在发出去就等结果。Three different directions, one thing in common: they turned complex multi-step processes into workflows that run again and again. What used to take hours now runs while you wait.