I Consumed AI Content for 2 Years. Then I Decided to Build Something With It - Here's What Happened
English Article
I Consumed AI Content for 2 Years. Then I Decided to Build Something With It — Here's What Happened.
I started using ChatGPT around early 2024.
Back then, I'd spend hours every day scrolling through AI news, bookmarking tutorials, subscribing to newsletters. I'd share AI tool articles on my feed and get likes. I felt like I really "understood" AI.
But one day I did the math: In two years of consuming AI content, what actually did I build with AI?
The answer: Zero things.
I had AI write poems for me, polish my emails, translate documents — all useful, but fundamentally just "talking to AI." Not building with AI.
My Turning Point: The Pain Point I Actually Wanted to Solve
I have an old laptop with over a decade of photos stored on it — travel snapshots, my kids when they were little, family gatherings. Every time I tried to find a specific photo, I'd spend ages scrolling through folders. Sometimes I'd search for thirty minutes and still come up empty.
Cloud storage search was too weak. Professional photo management software was too expensive and complex. I thought: what if I could use AI to understand what's in each photo, and then search with natural language?
This specific need was so concrete, I couldn't keep pretending that "bookmarking tutorials = learning AI."
Step One: Asked the Dumbest Question
I went straight to Claude and asked: "I want to build a local photo search tool where users can search photos with natural language, everything runs locally, no internet needed. How do I do this?"
What it told me blew my mind — it gave me a clear step-by-step plan:
- Import photos into the program
- Use AI to extract content descriptions for each photo
- Store in a database
- When users ask questions, search the database for the most relevant results
This was the first time I realized: You don't need to learn everything before you start building. You learn by building.
Step Two: Start With the Minimum Viable
I didn't try to understand every technical detail at once. My first version was this:
Use an off-the-shelf tool to organize photos into a folder, then batch-generate text descriptions, and store them in a spreadsheet.
Rough. Primitive. But it worked.
More importantly, I finally experienced the complete loop of "building with AI" — not watching tutorials, not bookmarking methods, but actually producing a result.
Later I iterated that crude MVP into what became PrismHub's prototype. The search capability it has today grew out of that most basic spreadsheet version.
Reflection: Why Didn't I Start Earlier?
Three things held me back:
1. Confused "learning" with "building" Tutorials are learning. Building is building. I spent way too much time on the former.
2. Thought I needed to "get the fundamentals first" I used to think I had to learn Python, databases, data structures before I could start. Now I know: The fastest path is to start from the problem you want to solve, and fill in knowledge as you go.
3. Underestimated AI as a "partner" Find the right AI assistant, and it can guide you at every step, help debug, help optimize. You're not building wheels alone.
Conclusion: Now Is the Best Time to Start
I'm 56 now. Not a programmer. Not an engineer. I was able to build a working tool because I didn't wait until I was "ready."
If you've been watching AI tutorials for a while without actually using AI to complete one full project — today is the best day to begin.
Find one small problem you actually want to solve. Doesn't need to be perfect. Doesn't need to be profitable. Just start building.
Your first version will be rough. That doesn't matter.
What matters is that you finally started.
rayslifelab.com
