The idea came to me while watching TV
I was watching TV with my girlfriend, and during the Egyptian segment, I realized that I could create a simple and fun service using the GPT API. I immediately turned on my computer and started planning a website that would allow users to type in their name, convert it to hieroglyphics, and provide the meaning via GPT.
Webflow: No-Code Website Builder
After thinking about which tool to use, I decided to implement it using Webflow, which allows you to create a no-code website. I remembered seeing a tutorial on creating a GPT api service using Webflow in the past, so I looked it up again. Thankfully, the template from that tutorial was available, and I was able to implement the website UI I needed with only minor modifications to its layout.
Engineering the GPT api prompts
We started with the idea that the user would enter a name and send it to the API, and the GPT 3.5 model would convert the name to hieroglyphs and resolve the meaning. However, after a few tests, we realized that the hieroglyph conversion was inconsistent, so we switched gears and hard-coded the logic for the alphabets and hieroglyphs, and modified it to feed the converted hieroglyphs into GPT's prompts.
Handling APIs with Make
I vaguely thought that I could just type the prompt and API key on the screen, but I realized that this would expose my API key and create a security issue. So I created a scenario to call the GPT api with a webhook in Make as the tutorial suggested. What I didn't expect was that there would be a total of 3 operations [forwarding to the webhook, calling the API, and receiving the webhook response] in one scenario, tripling the cost.
Google AdSense Delivery Failure
What I also didn't realize was that I couldn't run Google ads on these utility websites. Google had already figured out my shallow attempt to create a simple website using the GPT API and Webflow and use ads to increase my pipeline. I figured I might be able to get them to show up if I put relevant content on the site, or if I got more users, but I decided to hold off for now.
Lessons Learned
I found it easier than I thought to create a website and utilize the GPT API, and the quality of the GPT responses generated was satisfactory enough. Of course, monetizing it will take a bit more thought and effort, but I'll try to overcome that in the future.