Song-saem: From Planner to Developer with AIBlog home →Want to Get Good at AI? Stop Studying It and Build Something
Skill with AI comes from hours of use, not lectures. The surest way to rack up those hours is to build something yourself.
The fastest way to get good at AI is not to take a course but to build something. When you build, your hours with AI grow on their own, and along the way you learn first-hand what works ("Oh, it can do that") and what doesn't ("Nope, that's a dud"). That is how I went from laughing at AI to becoming a developer in two years.
Why did I laugh at AI at first?
When ChatGPT came out in November 2022, I honestly laughed. "You think you can handle complex work? I doubt it."
So I tested it. I had taught high-school English for years, and I asked it to analyze reading passages that students often got wrong. It couldn't do it well. When I asked it to write the kind of test questions I wanted, all I could do was sigh. So I barely used it for a while.
What changed?
Work did. While publishing a book on word roots, I needed example sentences and lists of synonyms and antonyms, and I opened AI again to get that done.
Looking back, that was the turning point. If I had not used AI for real work, I would still be asking it the occasional curious question.
Why is building faster than taking a course?
How well AI works depends heavily on how well you understand it. What it does well, where it goes wrong, how to ask: you learn these by using it a lot, not by hearing about them.
Asking the odd question won't add up to many hours. Once you start building, you hand AI dozens of tasks a day, and each one teaches you what works and what doesn't.
This isn't just my story. In Microsoft and LinkedIn's 2024 survey, three in four knowledge workers (75%) already used AI at work, but only 39% had received AI training from their company. Most people are learning on the job.
| Learning from a course | Learning by building | |
|---|---|---|
| Hours of use | Only during class | The whole time you build |
| What you learn | Other people's examples | What works in your own work |
| What you keep | Notes | Something you can actually use |
What I built, in order
- A word-roots vocabulary book: my first real use of AI, for example sentences, synonyms and antonyms.
- A simple web app: I put the book's content on screen.
- A word-roots game, "Road Voca": I built it, and it now lives inside the English learning site below.
- An English learning website: with new confidence, I built one-pex.com on my own, the kind of English content site people say costs hundreds of thousands of dollars.
About two years in, I realized I had become a developer without noticing.
So what should you build first?
Nothing grand. One annoying task in your current work is enough.
- Have AI tidy up a table you always organize by hand
- Make a template for a notice or document you write often
- Build a small calculator or checklist web page for your own work
When you get stuck, look it up. If you still don't know, ask someone who does. Repeat, and one day you'll find you've become good at it.
In short
- Skill with AI comes from hours of use.
- The surest way to add those hours is to build something yourself.
- Instead of a course, pick one task from your own work and build it. Look things up and ask when you're stuck.
Sources
A software planner who became a developer after two years of working with AI five to twelve hours a day.