I couldn't write a line of code a year ago. Here's how I ship AI products now.

A non-technical developer built multiple AI-powered products—including an AI YouTube thumbnail maker called ThumbAI—a year after learning to code by talking to AI, primarily using Claude Code. The developer shares lessons on communicating clearly with AI, verifying its output, and staying curious rather than intimidated by code.

A year ago, I couldn't write a single line of code. Today there are a few small things I've made that live on the internet — an AI YouTube thumbnail maker, a site that helps people understand volleyball rotations, and a handful of tiny tools. I'm not a developer. I built all of it by talking to AI, mostly with Claude Code. This isn't a tutorial. It's just the honest lessons I picked up going from zero to shipping — in case you've been wanting to build something but keep telling yourself you can't. A lot of people freeze the second they hear the word "code." Especially, I've noticed, a lot of people who were told early on that they're "not technical." I get it. But I never felt that fear. For me it was the opposite — it felt like a toy. Like a puzzle I couldn't put down. So if there's one thing I'd want you to take from this, it isn't courage. It's curiosity. You don't have to be brave to start. You just have to be a little curious about what you could make. The whole game is communication. The clearer you are, the better the output. Most "the AI is dumb" moments are really just vague requests. The single habit that saved me the most time: before letting it write any code, I ask it to repeat my request back and give me its plan first. "Before you write anything, restate what I'm asking for and outline your approach." Half the time, what it understood is not what I meant. Catching that in one sentence is so much cheaper than catching it after it's built the wrong thing. This one took me a while. The AI will confidently hand you something that looks completely correct and is quietly broken. So I stopped trusting it and started checking it. Run it. Read it. Ask it to review its own work. Think of it as a genuinely talented collaborator who is still capable of being wrong. You stay in the loop — you don't switch your brain off. My favorite trick: when I don't understand something, I just ask it. About itself. What does this error mean? What's this file for? Can I delete this? I don't go searching — I ask, "explain this to me like I'm new." It's like having a patient teacher who never gets tired of beginner questions. When you build alone as much as I do, that matters more than it sounds. You can paste images straight in. See a design you like? Screenshot, paste, "I want this vibe." Something looks broken? Screenshot the mess and paste it — it just gets it, far better than describing it in words. The project I've spent the most time on is ThumbAI https://getthumbai.app , an AI YouTube thumbnail maker — you start from a trending template or type your idea, drop your real face in with one photo, and get click-worthy thumbnails in about 30 seconds. It's rough in places and I'm still fixing things based on real user feedback, but it's real , and that still feels a little unreal to me. None of it required a CS degree. It required curiosity, and a willingness to be precise. You don't need to know how to code. You need to be curious enough to start, and clear enough to say what you want. I'm just a guy figuring this out in public, shipping small things, and writing it down as I go. If that's your thing — come say hi. I'll keep sharing what I learn.