How One Angry Prompt Built a Real Product in a Day

"You are supposed to be the smartest AI model in the world. Stop wasting my time with this and make me something useful that actually solves my problem."

That is an actual prompt I sent. Not my proudest moment. Then I walked away and went swimming with my son.

When I came back there was a link waiting. I clicked it. A working application. Rough, unfinished, but real. About six hours later that angry prototype had become Hey Motion, a browser tool for animated graphics like stat cards, checklists, quote cards and lower thirds that automatically match a brand's colors, fonts and logo.

I was not trying to build a company that afternoon. I was trying to finish a YouTube video. I needed a few animated visuals, maybe eight seconds of screen time each. Claude was doing genuinely good work. The problem was the process. Change the text, the layout moves. Fix the layout, the animation changes. Fix the animation, exporting breaks.

I could have solved the export issue with more code. I am a developer, I had the skills. But staring at the screen I realized the export was not the real problem. It was just the one currently in my face. Fix it, and the same cycle comes back with the next graphic.

Stop Patching, Start Asking Why

Up to that point I had been asking AI to help me survive the workflow. Make the graphic. Fix the graphic. Edit the graphic. Export the graphic. Task by task.

What I actually needed was a different question: why does this process keep creating the same problems, and can something be built that solves the whole thing instead of one piece of it.

That is the prompt that mattered, buried underneath the frustrated one. After I sent it I stopped thinking about the project entirely.

From Prototype to Something Worth Testing

The first version had the bones of a real tool. I showed it to my wife, who said her company had recently paid thousands of dollars and waited a month for graphics like these. That is when it stopped being just my problem.

I put ChatGPT to work researching competitors and pricing while Claude built, and I focused the next few hours on three things:

• Reusable templates, so nothing gets rebuilt from scratch
• Brand settings applied once, across the whole library
• Export built into the product itself, not a separate technical adventure

Within roughly six hours of hands-on work I had a working demo, using tools I already paid for. No agency, no months of planning, no team.

A frustrating export bug is rarely the actual problem, it is just the one currently punching you in the face.

Five years ago a rough version of this might have taken weeks, and the commitment would have talked me out of starting. That is the real shift. AI has not made bad ideas good. It has made them cheap to find out about.

So look at the task that keeps coming back, the workaround you have rebuilt three times. Do not ask AI to patch it again. Ask why it keeps happening, and ask it to build the process that makes the problem stop showing up.