Your "Read Later" List Is a Graveyard. Here's My Fix.
You've probably done this before: you save a video or article, tell yourself you'll get to it later, and never open it again. It just sits there.
It's not laziness. The process is genuinely annoying: copy the link, find a transcript tool, paste into AI, specify the format, copy the output, paste it into your notes app. Five steps, all manual. You do it twice and give up.
I don't have a coding background, so I didn't start where everyone else seems to start: "use AI to do vibe coding" That wasn't familiar thing for me.
What I use every day is Obsidian and Notion. So that's where I started to use AI.
📌 Example 1: YouTube Video → Obsidian Note
The old workflow: find a good video, copy the link, go to a third-party site to extract the transcript, paste it into ChatGPT, describe the format I wanted (subheadings, bullet points, no emoji), copy the output, paste it into the right Obsidian folder.
Four or five apps open at once. Every step manual.
Now: I have the Obsidian Web Clipper extension installed in Chrome. I see a good video, click once, it goes straight into Obsidian. Then I tell Claude in obsidian: "ingest this." The note generates automatically.
The format is set in advance. I don't repeat myself. No copy-pasting, no tab-switching.
(Side note: Karpathy's LLM Wiki is worth running through this workflow. I've built on top of it with some learning-specific optimizations. That deserves its own post.)
The same logic applies to articles. Just the same workflow.
📌 Example 2: Notion Database Design
Databases are where I spend most of my time in Notion. The field design matters more than people realize. Get it right once and it feels right every time you use it.
Before: I'd figure out the fields myself, then build each one manually.
Now: I talk through the requirements with Claude first. Once we've agreed on the fields, I tell it: "Go to X page and build the database based on what we just discussed." It does. I don't touch a single setting.
💡 One Last Thing
Most tutorials focus on ‘techniques’: specific prompts, specific skills, things that look useful out of the box. But every person's needs are different, and a generic trick rarely fits cleanly.
What actually matters is simpler: make AI work for your life, not someone else's workflow.
Start with something you already use. Something you do manually every day. Your most familiar tool is your best starting point. The specific answer is different for everyone. I can only show you what worked for me.
You don't need to finish a course first. You don't need to learn everything before you begin.
Just start. Enjoy the process!
不知道你有没有遇到这种情况，看见好视频文章，点了收藏想着下次就学却就再也没打开过。只能在收藏夹吃灰。
其实不是我们懒，是整个过程太麻烦：复制链接、找字幕/文本、粘给AI、调格式、再粘进笔记软件。五个步骤，全是手动，做两次就放弃了。
我没有开发背景，所以我没有从大家都在讲的"用AI写代码"开始。那不是我熟悉的东西。
我熟悉的是Obsidian和Notion，几乎每天都在用。所以我就从这里开始。
📌 例子一：YouTube视频 → Obsidian笔记
以前的流程是这样的：看到好视频，复制链接，去第三方网站提取字幕，粘给ChatGPT，告诉它我要什么格式（小标题、bullet points、不要emoji），复制输出，再粘进Obsidian对应的文件夹。
整个过程要在四五个界面之间来回切，全是手动操作。
现在：Chrome装好Obsidian Web Clipper插件，看到好视频点一下，直接存进Obsidian。然后在Obsidian里告诉Claude一句话："ingest这篇"，笔记自动生成。
格式是我事先定好的，不用每次重新说。我什么都不用复制粘贴，也不用切换任何界面。这里推荐Karpathy 的 LLM Wiki，之后帖子细讲这个，值得单开一篇，我在此基础加了更适配学习的优化。
看到好文章也是一样的逻辑，workflow完全一致。
📌 例子二：Notion Database设计
我用Notion最多的就是Database。每个人需求不一样，字段设计也会不一样，好的字段长期使用下来就会让人很舒适。
以前我得自己想字段，自己手动一个个建字段。
现在：我会先跟Claude聊，把需求捋清楚，在此基础上确定好字段，之后直接告诉它："你去这个页面，按我们确定的字段帮我建一个Database。"它就去建了。我再也不用手动了，效率大大提升。
💡 最后
我觉得现在大多数教程讲的都是"术"，各种技巧、各种功能，看似开箱即用，但每个人需求不一致，其实很难做到完美的交付，还不如从自身需求出发慢慢探索。
其实我更觉得AI协作应该深入核心，深入我们的本质需求。我们的本质需求是什么？就是让AI为我们所用，AI作为工具能为我们提高生活的质量，这才是核心。
那么基于这个核心，也就是我上面说的，你先从你自己具体的需求，你熟悉的领域，熟悉的环境去开始尝试与AI协作。而这个熟悉的领域，具体的问题，每个人是不一样的。我只能通过举我的例子给一些启发，但是我并不能替你回答这个问题，我不能替你开始。你不用想着想把所有都学了，你不用去看什么速成视频，那些花里胡哨的功能，其实那些都不重要。最重要的就是开始就可以了，去做就好了。Enjoy the learning process!