Three months ago, I thought AI programming was all about mastering technical skills. I believed I had to learn Python first, understand data structures, and be good at debugging. I was completely wrong—off the mark, really.❌
Looking back, I realize I wasted a full two months just fixated on the phrase "learning to code." That all changed when I met a business owner in cross-border e-commerce. He said something that clicked: "When you hire a programmer, do you need to learn how to code yourself?"
Oh, right! The essence of AI programming is about being the boss, not the employee. You don’t need to understand every line of code; you just need to know what you want and communicate that clearly to your "employee"—the AI.⚙️
I did a little math: Claude Code costs $250 a month. In Beijing, a junior programmer's salary starts at around 15,000 RMB. So, for about 1/60th of the cost, you have a 24/7 online super employee who never complains and can juggle front-end, back-end, and database tasks.
But how do you become that boss?
*🔴 The first pitfall I encountered was trying to have AI write complex functions right off the bat. *
The result? Errors in the code that I couldn't understand, and AI's explanations were even more baffling. I got stuck.
Then it dawned on me: you need to learn how to "delegate tasks" rather than just "code." For instance, I asked Claude to help me with a simple task: "Download YouTube videos and convert the subtitles into text."
Sounds easy, right? But the AI had to handle a bunch of stuff: installing dependencies, configuring the environment, managing network requests, and parsing subtitle formats. I just needed to watch and tell it, “The timestamps are off, adjust that,” or “Change the format to Markdown.”
This taught me a crucial lesson: your value lies in "speaking," not "doing."
🔴 The second pitfall was even trickier: the constant need to understand the principles behind everything.
I thought I had to learn SQL for databases and understand the HTTP protocol for network requests. What happened? After two weeks, my project was at a standstill.
A friend who works as a product manager called me out: “When you take a taxi, do you need to know how the engine works? When you order takeout, do you need to know how the chef flips the pan?”
That was a lightbulb moment. In the world of AI programming, you need to embrace “black box thinking.” Package the complex modules and hand them over to the AI; focus only on the input and output. The input is your project requirements, and the output is the functioning feature.
The measure of whether you've mastered AI programming is simple: can you build a usable web application in seven days? Not a toy, but something that can actually register users, store data, and has a user interface.
I tried this approach myself: using GPT-4o1 for the requirements document, v0 for the front-end interface, Claude for back-end logic, and Supabase for the database. By the fourth day, my first small app was live.
It was a simple to-do list manager, but the feeling was exhilarating—it was more rewarding than any paycheck. I transformed from a “user” into a “creator.” Instead of only using apps made by others, I could now build my own. This shift in identity brings exponential possibilities.
However, I have to be realistic—not everyone is suited for this. If you struggle to clearly express a simple requirement like, “I want a mini-program to track my daily water intake,” you really need to work on your communication skills first. No matter how capable AI is, it can't interpret your vague thoughts.
And another dose of reality: don’t expect to make money right away. Your first project will likely be clunky, full of bugs, and have zero users. But that’s okay. What matters is that you’ve gone through the entire process from “idea” to “product.” The first three failed projects are the most valuable tuition you’ll ever pay.
Looking back, I find it ironic that those with formal programming backgrounds sometimes struggle more with AI programming. They’re used to “writing their own code” and not accustomed to “directing others to write.” They want to control every line, but the core of AI programming is about letting go.
I know an experienced programmer with ten years in the field who’s learning AI programming more slowly than I did. Why? Because he keeps correcting the AI’s coding style, while I focus on whether the function works.
Sometimes, having no background can actually be an advantage. You come without the baggage or the obsession over “how code should be written.” You have just one goal: to turn your ideas into reality.
To be frank, in the next two to three years, not knowing how to use AI for programming might feel as awkward as being “computer illiterate” today. This isn’t just a scare tactic; it’s a reality unfolding before us.
Right now, you might use ChatGPT for copywriting and Midjourney for art, and you’re still a “user.” But when you learn to get AI to help you code, you become a “creator.”
This gap is a thousand times more important than whether you can code in Python.
So take action! Use that freed-up time from anxiety to create something that truly belongs to you.
Top comments (0)