… Or, as I like to call it, ‘How to Revamp a Website‘. The website you are currently browsing is a redesign started around November of 2025. Before this my WordPress website was “meh” — a downloaded theme, with some configuration and page edits. Not very interesting — frankly, it was just OK. Since AI is a topic I shadow closely, I decided to use agentic AI* to code a custom site. I haven’t written code since the ’90s. To say I’m out of practice is an understatement — it’s like I’m trying to ride a bicycle after a decade of being a couch potato. Undaunted, I decided to use WordPress again for the rewrite and bend it to my will with the help of AI.
My set up is to use the latest Gemini model (currently 3.0), and have it generate an agent prompt. Then, I paste that prompt into VS Code’s agent (I selected Gemini 2.5 since I have an account) and have it generate the custom code. Something like, “Create a Zen/NW tech design style”, then the chatbot cranks out a prompt “ACT AS – A seasoned WordPress developer … style sheets should be like {….”. Wow — I was delighted beyond measure by the quality and speed of the code generation. Was it perfect? No. But it was very easy to debug and adjust from the agent prompt input area. AI was also used to generate images on the “About Me” page using prompts along with photos of myself and an image of the Aurora Borealis I shot.
It’s easy to forget that agentic AI for coding has only been around for less than a year from the date of this post. One year! Imagine how it will evolve over time. I have software developer friends that insist AI coding sucks and makes too many mistakes. And for other agentic tasks — like travel planning, customer support, and other ‘fuzzier’ tasks — AI is still not mature enough to handle things entirely on its own. But over time, this will change and soon coding and other agents will be widespread.
The reason AI excels at coding tasks is that software development languages like PHP, Python, and Java have a well defined lexicon, specific rules, and huge amounts of code that can be used as a basis for training AI models. The new term ‘Vibe Coding’ alludes to the modern method of developing software — acting as a human director or visionary, rather than a coding maniac. We simply tell the agent our vision for the application, and then the agent spits out the code, we test it, and we refine it. Repeat. Iterate. Feedback loop. You get the gist. This website is an example of what can be done.
“The most important thing in programming is the thought process, not the code.” — Grace Hopper. With agentic AI, that thought process is now shared — and that’s something truly magical.
If you have any inkling to do your own development, I encourage you to give it a try. Let me know what you think — I’d love to see your results. Leave a comment!