How I Used ChatGPT to Code and Deploy BongoCat: A Case Study

After exiting my last startup, BetterCast, I was in a challenging mental space. I needed a new project to generate some soft revenue through affiliate marketing. Thus, BongoCat was born. The initial idea was to leverage AI to assist in creating content and managing the website. However, I didn’t want to use WordPress due to past experiences. I wanted to use something different that would allow me to expand my technical skills and knowledge. BetterCast was built using Django, so I decided to stick with what I had some familiarity with, even though I had never built anything from scratch in Django.

Key Takeaways

• AI as a CTO: ChatGPT acted as a virtual CTO, guiding the selection of technologies and implementation steps.

• Incremental Development: Breaking complex tasks into smaller, manageable components was crucial.

• APIs and Hosting: Utilizing APIs like Cloudinary and hosting on Heroku for simplicity and cost-effectiveness.

• Content Optimization: Integrating with Google Search Console to optimize content based on CTR and user engagement.

• Affiliate Marketing Experiment: Testing the potential of AI-generated content for affiliate marketing on niche sites.

• Learning and Growth: The project was a significant learning experience in using AI for product development.

Choosing Technologies

I wanted to expand my knowledge further because I had some experience using Django from BetterCast as a CMS and management system. I turned to ChatGPT to guide me, using it as my virtual CTO. I would outline the project goals and ask ChatGPT for the best technologies to achieve them. ChatGPT recommended Django for the backend and a custom design for the front end. I relied heavily on ChatGPT to tell me what to do, from setting up the environment locally to deploying on Heroku. My role as a product manager, understanding user stories, intent, UX/UI, and creating briefs, was complemented perfectly by ChatGPT’s technical guidance.

Using AI for Development and Content Creation

Initially, I started with GPT by providing an extensive and complex brief, resulting in suboptimal output. So, I broke down the brief into a workflow and manageable components. This involved understanding ChatGPT's context window limitations at that time and balancing the amount of code and description it could handle. As GPT evolved and its capabilities expanded, I was able to provide larger chunks of code and get more comprehensive assistance.

Here’s how I approached it:

• Planning: Breaking down the brief into infrastructure, environments, and smaller components.

• Implementation: Copy-pasting code provided by ChatGPT, testing it, and iterating based on feedback and errors.

• Deployment: Setting up staging environments on Heroku and refining the deployment process with ChatGPT’s step-by-step guidance.

If I were to start again, I would focus on writing tests first and then the code to pass those tests. This would streamline the development process and ensure more robust outcomes.

APIs and Hosting

I used Cloudinary APIs for image hosting. The site generates content and manages images through a form submission process that involves several steps to produce a fully realized article, including writing meta tags and descriptions.

Additionally, I experimented with various AI models like Claude, Perplexity, and Mistral to diversify my content creation strengths. I chose Heroku as the hosting choice due to its simplicity and cost-effectiveness, avoiding the complexities of AWS.

Content Creation and Optimization

Content creation involved filling out a form, which triggered a series of steps to write the article. This process included scraping content from review sites and integrating it into the draft. For optimization, I connected the site with Google Search Console. Although still a work in progress, the idea was to track click-through rates (CTR) and adjust meta titles and descriptions accordingly. By regularly checking these metrics, the AI could rewrite and optimize content to balance between clickbait and user engagement.

Affiliate Marketing

The initial goal was to generate passive income through affiliate marketing. I bought expired domains focused on niche topics like leatherworking and woodworking tools. Sites like archeryleather.com, Brislions.com and customwoodworkingdurango.com were created to push Amazon affiliate products. This was more of an experiment to see if AI-generated content could drive affiliate sales. However, my interest shifted more towards the development process than the affiliate marketing aspect over time.

Future Plans

In closing, while the future of BongoCat as an active project is uncertain, it served as a valuable learning experience. The project helped me understand how to use AI to create a functioning product and improved my technical skills. I could scale it up if there is a demand for such a tool. For now, BongoCat remains a personal website to help me write articles and case studies more efficiently.

By leveraging AI, I was able to create BongoCat, a website that assists in content creation and optimization, providing a robust case study of AI’s potential in web development.