Don’t start with AI tutorials.
Here's what actually got me started: I had a workflow that used to take me 60–90 minutes a day. After handing it to AI, it took 11 minutes. And the result was even better than what I did manually.
That was the moment. Not a tutorial. Not a YouTube video. Just that specific "oh wow" moment.
So here's my conclusion on how to actually start with AI:
Use it to solve one real problem in your life. Something that annoys you. Something that wastes your time. Something you have to do but really don't want to.
AI products aren't plug-and-play. If you're new, you have no idea which feature solves your specific problem. There's too much information. So I don't recommend watching "Learn 80% of AI in one hour" videos. The content is fine. I'm not saying it's useless. I just think the most important thing in learning is genuine interest. And interest doesn't come from tutorials. It comes from solving your own problem.
My example: I used to save YouTube videos as notes like this: find the transcript → open Claude → paste it in → add formatting instructions → copy the output → switch to Obsidian → create a file → paste again. Multiple apps. And I had to re-explain my format preferences every single time.
Now: a plugin pulls the transcript automatically. I tell Claude code "digest this file." It formats everything using a template I set once, and saves it to the right folder. Two steps, done.
It sounds small. But when you repeat something every day, cutting it down feels really good. That feeling is your motivation to keep learning.
Once you have that, tutorials actually help. You watch them with real questions. You start caring about things like how to reduce repetitive prompting, how to save tokens, how to keep AI on the same page as you. It clicks differently.
Then you can go further: plugins, MCP, APIs, vibe coding. Turning your own ideas into real things. It feels like a game with no ceiling.
New things drop every day. That keeps it interesting.
The actual sequence:
Solve one concrete problem first.
Then learn from what others have figured out.
Then keep optimizing from there.
Go explore. Go play. Enjoy it!
先说一个让我真正开始去学 AI 的事情: 有一个我以前每天要手动做一到一个半小时的工作流,在交给AI后它只需要 11 分钟就做好了,并且结果比我做得还好,还能自动化不用我管。当时我就:不行,这个AI我学定了。
不是某个视频,不是某篇教程,是那个具体的"我靠"的瞬间。
那么我就直接说结论,当我们在说如何真正开始学和用AI的时候,我觉得对于普通人、对于个体来说,最有效的方法就是去用它。比如说,去用它解决你生活中具体的一个问题,或者说优化你生活中的一个具体的痛点,一个你不爽或者你觉得很花时间、不想花时间去动的东西。我觉得这是最有效的方式。
现在的 AI 产品并不是开箱即用的,对于一个完全新的用户来说,你其实是不知道怎么可以解决你具体的问题的。因为它的信息也太多了。所以我其实并不推荐你在 YouTube 或任何平台去看什么“Learn 80% of the code in one hour”之类的视频。那种东西,我觉得你一开始不应该看。原因很简单:这些内容和你的生活没有任何连接。它们的内容很棒,我不是说那些是胡扯。我只是觉得,学习最重要的就是你对它有兴趣,并且有自我学习的动力。这不意味着你看教程、跟着教程做、读一些论文、听一些播客就叫自我学习。如果你真的想学好一样东西,你必须对它非常感兴趣。
那要如何才能产生兴趣呢?尤其是这些工具现在还不太用户友好,但它们非常强大,而且还在发展中。对我来说,真正对这些东西产生兴趣的契机是,我开始需要自动化我的工作流程,以解决我生活中的实际问题。就是我开头的事情让我意识到,天哪,我真的需要学习这些东西,我不想再回到手动做所有事情的状态,每天花两个小时,而现在我可以在几分钟内完成这些事情。这就是我的动力。这不意味着看了一些教程我就感兴趣了,完全不是。尤其是那些内容不适用于每个人,每个人都有不同的问题和背景,对吧?所以你必须解决你自己的问题,这才是重点。
真正有效的起点只有一个:找到你生活里一个你很不想做、但每天都得做的小事,然后去解决它。不用是什么大项目。就是那种重复的、无聊的、浪费时间但又不得不做的东西。
我自己的例子:以前我想把一个 YouTube 视频保存成笔记,要经历这几步:找到字幕脚本 → 打开 Claude 网页版 → 贴进去 → 加一堆格式要求 → 复制输出 → 切换到 Obsidian → 新建文件 → 粘贴。几个个软件来回跳,而且每次都得重新告诉 AI 我想要什么格式。
现在:Obsidian 插件自动抓字幕,我对 Claude 说"Digest this file",它按我固定好的模板输出,直接存进对应文件夹。两步,完成。这看似没有什么太大的 impact,但对于我来说,减少了很多切换软件的成本。我不用每次都强调我想要什么样的格式,尤其是当你每天都会重复这样的动作的时候,你会觉得现在这个用法真的太爽了,减少了很多时间。而且这些时间浪费在这些流程上其实很没有意义,对我个人没有任何意义。所以我觉得这个优化是非常值得的。
在此动力上,我们可以再去看看一些教程,这样才能坚持下来。现在的各种信息和要学的东西实在太多了,如果没有兴趣、没有正反馈,是非常难以坚持的。但我觉得一旦你 get 到了乐趣,再去学其实很快乐。你可以去学诸如各种插件、MCP 以及 API 的使用,去尝试 vide Coding。当你把自己的想法变成实际落地的产品时,你会发现这其中有无限可能。它就像一个被解锁的定制化游戏,理论上拥有无限的探索空间,这个过程非常有意思。
而且它每天都有新的东西,这种不断的探索很有吸引力。我觉得这才是真正开始去“做”的动作:
先解决一个具体的问题。
再去学习别人总结的优秀攻略。
在攻略的基础上不断优化。
当你遇到更好的方案来解决问题时,肯定会经历一些反复。如果 AI 解决不了,你就需要去教它怎么解决。在学习攻略时,别人可能遇到过同样的问题,他们会分享如何减少这种沟通成本、减少 Token 消耗,从而节省时间,确保 AI 和你的理解保持在同一个频道上。
所以,这些都是需要你去探索的内容。去探索、去玩、去优化就好了,这才是真正的开始~享受学习的乐趣吧!