Andy He @AndyAIPro

CTO｜AI研究×产品×搞钱 | 写代码，也写命 | 副业/远程/出海/躺平4条路都跑过 | 花名：猿人谷 New Zealand Joined May 2014

Tweets

129
Followers

30
Following

356
Likes

215

Andy He @AndyAIPro

a month ago

Most people use AI tools like Claude the way they use Google: they type a question and wait for an answer. This is the lowest-value way to interact with them. To truly unlock their potential, you must move from searching to collaborating.

0 0 0 13 0

View Details

Andy He @AndyAIPro

a month ago

Strive for effective synergy with AI, not total reliance. Treat every new chat as a fresh start with zero memory; you are the one who provides the context and the soul.

Anatoli Kopadze @AnatoliKopadze

a month ago

x.com/i/article/2053…

172 1K 7K 20.3M 28K

2 0 1 33 0

View Details

Andy He @AndyAIPro

2 months ago

Awesome! ✨ Jet lag hasn't even hit yet, but this surprise did. Tencent Cloud Architect Hall of Fame — unlocked ✅ Grateful for the recognition. Keep building, keep growing. 💪

0 0 1 22 0

View Details

Andy He @AndyAIPro

2 months ago

It's true:Something Big Is Happening

Matt Shumer @mattshumer_

4 months ago

x.com/i/article/2021…

7K 28K 119K 87.2M 182K

0 0 0 16 0

View Details

Andy He @AndyAIPro

3 months ago

In Agentic RAG, once the session ends, the agent loses all context.

Nikki Siapno @NikkiSiapno

3 months ago

Traditional RAG vs Agentic RAG vs Memory Systems Most people mix these up: Traditional RAG → Agentic RAG → Memory Systems Each step solves a different limitation. Here’s the breakdown (save this): 𝟭) 𝗧𝗿𝗮𝗱𝗶𝘁𝗶𝗼𝗻𝗮𝗹 𝗥𝗔𝗚 The classic way to ground LLMs. Pipeline:

14 47 168 15K 159

0 0 0 28 0

View Details

Andy He @AndyAIPro

3 months ago

面向Agent的设计模式👍

Google Cloud Tech @GoogleCloudTech

3 months ago

x.com/i/article/2033…

110 949 4K 1.8M 9K

0 0 0 18 0

View Details

Andy He @AndyAIPro

3 months ago

MCP for arXiv 牛！Agent 研究终于能实时站在巨人肩膀上了～多轮检索 + embedding 搜索 + 关键词过滤，百万论文秒级响应，太适合做教育/职考 Agent 了。我在护考证产研时，最缺的就是这种“海量文献 → 个性化题库”的链路，现在直接补齐！

alphaXiv @askalphaxiv

3 months ago

Introducing MCP for arXiv Let your research agents stand on the shoulders of giants Fast multi-turn retrieval, keyword search, and embedding search tools across millions of arXiv papers 🚀

79 406 3K 275K 3K

0 0 1 95 0

View Details

Andy He @AndyAIPro

3 months ago

Amazing！

Shalini Goyal @goyalshaliniuk

3 months ago

Most people still think AI Agents and Agentic AI are the same, but they’re not. Here’s a quick guide that breaks down the workflows, showing how agents operate versus truly agentic systems. It highlights the differences in planning, execution, monitoring, and adaptability

17 28 90 3K 70

0 0 0 16 0

View Details

Andy He @AndyAIPro

3 months ago

That’s true.

Emmanuel Tsekleves @PhDtoProf

3 months ago

Most PhD supervisors will never tell you this. The real reason PhD students fail is not methodology. It's not data. It's that they never clearly identified a research gap

11 238 802 27K 528

0 0 1 11 0

View Details

Andy He @AndyAIPro

3 months ago

It’s very useful!

Faheem Ullah @Faheem_uh

3 months ago

PhD Students - How to generate a graphical abstract in seconds? 1. Go to researchviz.io 2. Copy and paste your abstract 3. ResearchViz will generate graphical abstract. You can download it as PPT and make any changes.

6 189 774 104K 1K

1 0 1 133 0

View Details

Andy He @AndyAIPro

3 months ago

Face the challenge, and it will resolve.

0 0 0 5 0

View Details

Andy He @AndyAIPro

3 months ago

Thanks for sharing!

Shraddha Bharuka @BharukaShraddha

3 months ago

RAG, AI Agents, MCP, and A2A aren’t competitors. They’re different layers of the same AI system. Once you see this mental model, modern AI architecture becomes much clearer 👇 1️⃣ RAG = Better answers RAG is about retrieval + grounded generation. Flow: User question →

44 270 1K 73K 1K

0 0 0 23 0

View Details

Andy He @AndyAIPro

3 months ago

快速上手（零依赖）： 1. git clone github.com/yuanrengu/conf… 2. cd configadpter 3. python3 openclaw_config.py 浏览器自动打开 http://localhost:7890

0 0 1 20 0

View Details

Andy He @AndyAIPro

3 months ago

【OpenClaw 大模型切换器】终于搞定了。受够了手动改JSON？一个轻量级Web工具，现在点点鼠标就能无缝切换OpenClaw的大模型配置。把时间留给思考，而不是繁琐的配置。 #OpenClaw #AIAgent #大模型切换 #开源工具 #副业神器

1 0 1 62 1

View Details

Andy He @AndyAIPro

4 months ago

【准确率才是大模型的生命线】语言类考试不像数理化有复杂的逻辑运算，规则其实非常清晰、边界明确。但奇怪的是，几乎所有模型在最简单的题上都偶尔会出错，而且错得特别自信——死鸭子嘴硬，自圆其说那一套玩得飞起。对于教育赛道、尤其是雅思/托福/四六级这种“容错率极低”的考试来说，这太可怕了。一旦在关键知识点上给错引导，学生满心欢喜以为“押中题”，洋洋洒洒写完，自信满分，结果考试一对照：0分。那个画面，太刺激了。谁用大模型备考时也被它“自信满满地带进沟里”过？评论区说说最离谱的一次经历吧。反正我是踩了好几次坑😂 #大模型 #AI教育 #雅思 #托福 #语言学习

0 0 1 59 0

View Details

Andy He @AndyAIPro

4 months ago

@Faheem_uh D

0 0 1 993 0

View Details

Andy He @AndyAIPro

5 months ago

yt-dlp 视频素材下载 + AI 处理全流程模板 & 常见坑避雷： docs.google.com/document/d/1F6…

0 0 0 74 0

View Details

Andy He @AndyAIPro

5 months ago

2026 年还在手动保存视频？别再重复造轮子了！推荐这个开源神器：yt-dlp 🔥 - 支持 1000+ 站点（YouTube、TikTok、Instagram、Twitter、SoundCloud、Bilibili 等） - 最高质量下载 + 自动合并音视频 + 嵌入字幕/元数据 - 批量 playlist/频道下载、指定时间段、SponsorBlock 跳广告 - 活跃维护中（最新 commit 1月19日，stars 超 14万+）为什么我爱用它做 AI 副业？ - 批量下教程/竞品视频 → 喂 NotebookLM / txtai 自动生成总结/播客 - 出海内容采集：高清源 + 字幕提取 → 快速建多语言知识库/RAG - 命令行一键：yt-dlp -f bestvideo+bestaudio/best URL --embed-subs 核心命令速查（复制走起）： yt-dlp URL # 最简单下载 yt-dlp -f "bv*+ba/b" --merge-output-format mp4 URL # 最佳质量 yt-dlp -x --audio-format mp3 --embed-thumbnail URL # 只下音频+封面 yt-dlp --yes-playlist -I 1:100 PL... # 下载播放列表前100个 GitHub：github.com/yt-dlp/yt-dlp （Unlicense 许可，随便商用）想批量自动化？私信“ytdlp”领我整理的“视频素材下载+AI处理全流程模板 + 常见坑避雷”📥 你用 yt-dlp 做过最酷的事是什么？评论区分享～ #AI工具 #开源 #yt-dlp #视频下载 #副业神器 #RAG

2 0 2 515 0

View Details

Andy He @AndyAIPro

5 months ago

@kenw_2 哈哈哈，邵氏武侠风AI生成 + 教英语，这脑洞绝了！视频里对掌那一下太绝了，AI把竹林打戏复刻得这么丝滑，短视频赛道越卷越宽🔥

0 0 2 508 0

View Details

Andy He @AndyAIPro

5 months ago

Quick install & test: git clone github.com/yuanrengu/xcra… pip install -r requirements.txt python fetch_more_history.py --username your_target --since_id LAST_ID --until_id OLD_ID Docs: github.com/yuanrengu/xcra… Issues/PRs welcome! 🚀

0 0 0 51 0

View Details

Andy He @AndyAIPro

5 months ago

工具兔这16个 NotebookLM 提示词太实用了！从“5个关键问题”到“辩论格式”，直接把 NotebookLM 从笔记工具升级成深度研究/内容利器🔥 我在职业教育公司做过产研（专注医药护考证）负责人，最头疼个性化辅导闭环——以前手动批改+跟踪，成本高。现在用 NotebookLM + 类似提示词，一键生成弱项题 + 解析 + 复习计划，效率翻倍！实战补充（职考/学习场景）： - 用“私人导师” + RAG 检索真题：'作为资深医师资格考官，为我生成10道针对弱项的练习题 + 详细解析 + 个性化复习计划：[插入错题/弱项]' - “辩论格式”超适合争议题：让 AI 模拟正反方辩论护考知识点，帮考生快速记忆矛盾点。