Today, I almost built another website. It started with a short video of a wooden calendar puzzle. There were ten oddly shaped pieces and a board filled with months, dates, and weekdays. The challenge was to arrange all ten pieces so that only today's date remained visible.
Simple idea. Beautiful execution. And immediately, my brain went to the dangerous second thought in the age of generative coding: I could probably build this with AI in a day.
A year or two ago, that thought would have stayed a daydream. Drag-and-drop interactions, rotating pieces, solving algorithms, and mobile compatibility would have required serious programming time. But with tools like Codex and Claude Code, the technical barrier feels almost nonexistent. I no longer ask, "Can I build this?" I ask, "How quickly can I build this?" And sometimes, that is a problem.
From a 30-Second Video to a Software Project
I took screenshots of the puzzle and showed them to ChatGPT. We mapped out the architecture: Cloudflare Workers, TypeScript, a puzzle-solving algorithm, and a mobile-friendly interface. I was ready to ask Codex to start writing code.
And that was when I asked another AI a simple question: Does a website like this already exist?
Five minutes of research found several websites offering almost exactly what I had imagined, such as Daydoku and Date Blocker. Instead of giving up immediately, I decided to investigate the numbers using Semrush.
The Numbers Were Surprisingly Small
According to estimates, an established competitor like Daydoku had modest organic search traffic globally, ranking for a few low-volume keywords.
More importantly, I wasn't planning to build this game purely for fun. I wanted to create another website that could attract visitors from Google. And the available evidence suggested that the search opportunity might be quite small. Even an established, functional product wasn't getting massive estimated Google traffic. Would another nearly identical website really be a good use of my time?
AI Has Made Building Cheap. It Hasn't Made Attention Cheap.
This was the most striking realization. The puzzle wasn't too difficult to build anymore. The hosting cost would be tiny. Cloudflare's infrastructure could handle a small game without much expense.
But none of that would guarantee visitors. AI can dramatically reduce the cost of creating a product. It cannot guarantee that anyone wants to use it. And it certainly cannot guarantee that Google will send people to it.
I already operate several small websites. Attracting consistent search traffic remains difficult. This calendar puzzle was a useful reminder: I didn't need another technically successful project with an uncertain audience. I needed to become better at choosing which projects deserve to exist.
The Most Valuable AI Output Today Wasn't Code
Usually, when people discuss AI productivity, they talk about how much work AI can produce. How many lines of code? How many articles? How many websites?
But today, AI helped me accomplish something different. It helped me stop. One AI helped me understand how to build the game. Another helped me find existing competitors. Then I used SEO data to examine whether there was enough search demand to justify creating another website.
By the end of the process, my conclusion was simple: Don't build it. At least not as another SEO-driven website.
A New Rule for My Future Projects
My old process was often: Discover an interesting idea -> Build it with AI -> Launch it -> Hope people find it.
My new process should be: Discover an idea -> Find existing solutions -> Research demand -> Identify what's missing -> Decide whether to build.
AI has made it easier than ever to start. Maybe the next important skill is knowing when not to. Sometimes the best project is the one you decide not to build.
This article was originally published on Ordinary Man Trying.
Top comments (0)