I thought this would be easy. Take a photo. Let AI handle the rest. It reads the sign, figures out the business name, checks the location, finds the Google reviews, and politely hands me the answer. All I have to do is take one picture. “Wow… this is basically free money.” And if I’m doing that, obviously I should use Google’s own AI. Then I saw the name. Gemini. Even the name sounded smart. “Yep. You’re hired. Gemini, it’s you. Hahaha.” So I started coding with absolute faith in Gemini. I had no idea yet how quickly that faith was about to be tested. My first real-world test was a phone repair shop right in front of my house. Literally right outside. So naturally, I dressed for the occasion: shorts, slippers, and one test phone. The only problem was that the repair shop was completely open-fronted—no door, nowhere to hide.
I stood outside, raised my phone, and pointed the camera at the sign. That was when I made eye contact with the owner. His face said everything. “What is this guy doing?” To be fair, I knew I looked suspicious. A man in shorts and slippers silently photographing someone else’s shop sign. Not exactly subtle. But I had come this far. Retreat was no longer an option. I accepted the embarrassment and took the shot. Click. And in my head I declared: “I believe in you, mighty Gemini~~~~” Then I waited under the blazing sun. 5 seconds. 10 seconds. 15 seconds. At around 20 seconds, something finally appeared. “…Huh?” My first emotion was not wonder. It was concern. “Why is this so slow?” Twenty seconds for one photo. Imagining a traveler standing in the middle of a trip waiting for this made my back go cold.
Still, okay. Speed can be fixed later. Accuracy is what matters. Finally, I checked the result. Gemini confidently presented me with a business. There was just one tiny problem… it was the wrong business. Same name. Completely different place. The phone repair shop I had photographed was standing perfectly well right in front of me, while Gemini proudly brought back some namesake from somewhere else. I stared at the screen. Then at the shop. Back at the screen. Back at the owner. The owner was still watching me. And one sentence entered my head. “Oh… I’m screwed.” Wasn’t AI supposed to handle everything after I took the picture? Read the sign, find the place, pull the reviews—boom, done? Wasn’t I just supposed to connect a few things together? But on the very first real-world test: slow.
And wrong. A clean one-two punch. If it couldn’t reliably find a phone repair shop outside my own house, what happens overseas with a sign written in a script I’ve never seen? What about location? What if several businesses have the same name? What if the sign is blurry? What if the photo is tilted? What if one sign contains two languages? Problems started spawning in my head like someone had opened the floodgates. Just minutes earlier I was thinking, “Wow, this is going to be ridiculously easy.” And now I was thinking, “Where on earth do I even start fixing this?” That day I learned something very important. Just because AI is smart doesn’t mean the app I build magically becomes smart too. And the real development of GoPic… I think it probably started that day. Oh, Gemini~~~~ To be continued.