Technology So Advanced It Feels Like Magic

There's an old line, often credited to Arthur C. Clarke, that any sufficiently advanced technology is indistinguishable from magic. I used to think of it as a nice quote. Lately it feels more like a description of what I'm seeing.

For the last year, most of my surprises came from one place: code. Agents stopped being something that answers in seconds and became something that works for hours. The first time I let one run for two hours, it was on a pile of 120 files a client had sent us about their products, with no order at all. The agent analysed every one of them, extracted the data, invented its own conventions to clean it up and joined it all together. It turned unstructured, disorganized information into a structured database, with the files sorted into a clean folder tree. I didn't touch it. It felt weird to walk away from a prompt and come back to a result that advanced — the kind of job I'd have expected to take a person several days.

So I assumed that's where the progress was. Code was moving fast; everything else, I figured, was more or less where I'd left it.

Video was where I'd left it a long time ago. I don't need it for work, and I'd always been skeptical of it, for good reasons: it never quite captured the concept you gave it, there was no audio, backgrounds lost their shape and drifted between frames, nothing cut naturally or held a natural expression, and characters who were supposed to be interacting never actually looked at each other. That was the picture in my head, and I hadn't checked it in ages.

Then I opened Google AI Studio and tried Gemini Omni Flash. I asked for a battle between technology and magic. I designed the characters myself, and they stayed themselves from shot to shot — same faces, same clothes, scene after scene. The sound effects land where they should. The physics is still strange; things move with a weight they shouldn't have. But it took minutes, and it cost next to nothing for something that looks like this.

The battle ends in a tie. It felt like the only honest ending: if Clarke is right, at some point you can't tell which side is which.

Is it perfect? No. Is it useful yet? I don't know — but a few months ago I wouldn't have bothered to ask.

What surprised me wasn't the video. It was that I'd stopped looking. I don't keep a list of what AI can't do. I hit a wall, decide "it can't do this yet," and stop thinking about it. Then, months later, I tried again, and the change in video generation was massive. Nobody can follow all the progress all the time; I didn't realize how much it improved. Sometimes the only way to find out if something has changed, and how it affects your work, is to try again.

Technology only feels like magic to people who don't know it's possible. Once you know, it's just a tool, and you can do something with it.

Now that you know about this tool, I invite you to try it. Open Google AI Studio, pick Gemini Omni Flash, and ask for the video you'd never have bothered to make. Or better: think of the last thing you decided AI couldn't do, and try it again.