For almost four months, I did nothing but dream. I would think up a product, imagine how it would work, picture it finished — and never write a single line of code. In my head, building it was the easy part. I had ChatGPT and Claude. How hard could it be?
Then summer break came. May 2026. And one thought kept hitting me: I had spent four months thinking about products and zero months building them.
So I decided to actually start.
The naive part
People had already warned me — learn the basics first, don't skip the fundamentals. I nodded along, but I didn't really believe that was the real problem. I still thought AI was some kind of magician. Explain the idea, get the code, done. One idea, one solution, no friction.
Here's how naive I actually was: I would describe my idea to ChatGPT, ask it for a plan and a prompt, then copy that prompt straight into my IDE — without reading what the prompt actually told the AI to do. I just trusted it.
The output, unsurprisingly, was often the opposite of what I wanted. I burned through my usage limits chasing fixes for problems I had created myself, just by not reading my own prompts.
The part nobody warns you about
The real barrier was never syntax or libraries. It was the gap between having an idea in your head and actually turning it into something real. That gap doesn't show up in any tutorial. You only find it by trying to cross it — and falling in.
What changed
The first ten days were pure confusion. But somewhere after that, things started clicking. I learned actual prompt engineering — not just "ask nicely," but understanding what a prompt does before I run it. I got a real feel for what AI is actually good at, and where it quietly hands you garbage dressed up as an answer.
Oddly, this made me a better programmer, not a lazier one. Reading and correcting AI output forced me to understand code faster than writing it alone ever did. I started filling the gaps AI left with my own thinking, instead of just more prompts.
It also made me more consistent — not just in coding, but in how I studied and how I showed up to work every day.
The one thing I'd tell myself in May
None of this — the prompt engineering, the sharper code sense, the consistency — happens if I don't take that first messy, confused step. The confusion wasn't a sign I was doing it wrong. It was the cost of finally doing it at all.
Top comments (0)