AlexYang

AI Engineer in New York｜ Build Agents in Fintech

[AlexYang @YangPlusAI]

@turingou Langchain 团队也提到了这个观点，大模型最后可能将harness的部分也吞掉，所以他们提出 Memory is the only Moat。理由很直白：context is what makes agents work well. Memory is the most kind of context. If someone else controls your memory, you’re locked in, and thus, memory is the moat.

[AlexYang @YangPlusAI]

@turingou Langchain 团队也提到了这个观点，大模型最后可能将harness的部分也吞掉，所以他们提出 Memory is the only Moat。理由很直白：context is what makes agents work well. Memory is the most kind of context. If someone else controls your memory, you’re locked in, and thus, memory is the moat.

[AlexYang @YangPlusAI]

美股的KOL是比较好当的，只要永远做多头，早晚都能轮到你选的标的

[AlexYang @YangPlusAI]

protect your thinking not code

[AlexYang @YangPlusAI]

@mdzzi 币圈人都成美股专家了

[AlexYang @YangPlusAI]

AI coding这么强 我们还需不需要学coding。以下是今天听播客时听到的一个最好的答案：

我们应该更关注学习方式的变化：学 coding 的方式，从 bottom-up 变成了 top-down。而这种方式的转变，也在告诉我们该在什么时候学什么

过去的 bottom-up，是从最基础的数据结构、算法学起，再学怎么写 function、封装 class，最后才学如何搭建一个应用、做一个网站

但现在变成了 top-down，我们可以先用 Claude Code、Cursor 快速打出一个应用、做出 demo。等到需要调整时，才去了解这个应用大概是怎么运作的，前端和后端各在做什么；需要更精细的调整时，再进一步了解后端有哪些不同的架构和组件；再深一层，才去搞清楚 function、class 是什么，sync 和 async 有什么区别

播客是 42章经 和slock创始人RC的访谈，他也是 kimi.cli的作者

[AlexYang @YangPlusAI]

@dotey 技术ready和组织ready是两回事，需不需要人工作并不是完全技术说了算的，主要是看组织能不能适应这样的变化

[AlexYang @YangPlusAI]

@AYi_AInotes 很明显是磕了，但是说得确实很真实

[AlexYang @YangPlusAI]

workflow这些东西完全可以被skills所取代了，workflow中的步骤和逻辑，完全可以在skill中用自然语言来表示，在需要确定性结果的地方加上脚本，让agent直接来运行。agent + skills 的组合相比于 workflow 也更容易管理，更容易针对客户进行定制。不知道现在用n8n这些的人还多不多

[AlexYang @YangPlusAI]

看到了@grapeot 的这篇文章，非常有意思。本质上也是在蒸馏自己的认知体系，形成skills，从而让 AI 的回答能够通过这些skill跳出共识。这些skill也可以随着我们不���提供的context来进行学习和迭代

https://t.co/Av8ryhQcyl

[AlexYang @YangPlusAI]

ChatGPT时代之后出来的新书质量非常参差不齐，很多新书注水严重，知识密度极低，其中不乏各种 AI engineer手册，Agent/RAG 学习资料。现在学习 AI 最好的途径就是去找热门的 github repo 去研究，或者看 anthropic openai 等头部ai公司的 tech blog。如果读书的话，建议读ChatGPT出来之前��经典cs书籍

[AlexYang @YangPlusAI]

@PandaTalk8 langchain deep agents

[AlexYang @YangPlusAI]

GUI is for human. Cli is for agents. Building for agents is the future.

[DAN KOE @thedankoe]

If you're a writer, be okay with repeating what's already been said, because most ideas have already been exhausted, and genuinely novel ideas are rare.

Nobody follows you because you're talking about something new. They follow you because they want to read your point of view.

[DAN KOE @thedankoe]

A pattern I've noticed in stuck people:

They're always busy. They never stop moving. They have 47 tabs open and a notebook-sized to-do list. But if you ask them what they accomplished this week that actually matters, their mind goes blank.

Busyness isn't a badge of honor.

[AlexYang @YangPlusAI]

Agent as a service

[Claude @claudeai]

Introducing Claude Managed Agents: everything you need to build and deploy agents at scale.

It pairs an agent harness tuned for performance with production infrastructure, so you can go from prototype to launch in days.

Now in public beta on the Claude Platform.

[AlexYang @YangPlusAI]

@zarazhangrui Code becomes a liability, not an asset.

[AlexYang @YangPlusAI]

@sitinme 可以试试 excalidraw mcp 好用美观

[AlexYang @YangPlusAI]

@zarazhangrui Before consuming long-form content, you need to ensure that is not generated by AI

[AlexYang @YangPlusAI]

@himself65 太低频了