From Zero to Prompt to Page
My mind is blown. This website, blog pages, everything you see—it's all purely custom code created with the help of generative AI. Despite having no formal development background or CS education, I've managed to build something entirely tailored to my specifications. While it may not win design awards, I'm proud of what I've accomplished here.
The Genesis
It started with a simple goal: practice using new AI tools to build something meaningful. I settled on creating a professional site as a first low stakes project to act as a digital home where I could document my AI projects, and perhaps do a bit of personal branding (which makes me chuckle since I have zero plans to actively promote this site). Ultimately, I’m seeking a creative outlet for coding and writing projects. This post chronicles that first project building this site.
My AI Co-Developer
My AI collaborator of choice was Simtheory, an AI aggregator workspace that lets users experiment with different models. This platform allowed me to test various code outputs and fine-tune my approach iteratively. The creators of Simtheory also run an excellent AI podcast, "This Day in AI," which offers a refreshing mix of entertainment, news, and insights from tool creators.
The Collaborative Process
Working with the AI involved considerable back-and-forth, but was generally painless:
- Design Conceptualization: I'd describe my vision for layouts, color schemes, and overall aesthetic.
- Code Generation: The AI would produce HTML, CSS, and JavaScript snippets based on my descriptions.
- Experimentation: I'd try surprisingly simple prompts like "add some classy CSS animations so that the content animates into view" and watch as the AI translated these vague requests into functional code.
- Refinement: When the output wasn't quite right, we'd iterate through clarifications and adjustments.
- Implementation: Finally, I'd integrate the code into my site, often with minimal modifications.
Sometimes the AI nailed it on the first attempt. Other times, my prompts needed refinement when they weren't specific enough. The process felt less like dictating instructions and more like a collaborative rapid prototyping session.
Surprises and Lessons Learned
I’m surprised and amazed this is successful. The main challenges I ran into were getting all of the knowledge and how-to setup around create the site itself, ensuring it’s safe, and working through the actual code updates, merges, and pushes to production. I do not see why AI can’t get smart enough to solve for these setup and security needs out of the box. To me, seems like it’s a question of time for the tooling to be enabled, and the current state AI models are more than capable enough to accelerate and simplify these steps.
The main risk I see around the coming corners is needing more knowledge to solve issues when things break. With this basic site, I haven’t needed to do any actual de-bugging, and have been able to iterate with pure and direct feedback to the AI.
Looking Forward
This website is just the beginning. I'm excited to continue exploring the intersection of AI and creation, documenting my journey along the way. I’m bullish to learn how to partner with AI to drive a workflow, build some new things, and improve my productivity overall. I’m training for a Triathlon, and so may try a coaching/ workout generation app. At any rate, the possibilities seem limitless, and I'm excited to keep going deeper.