Kissing AI Bots Goodbye: Personal Website Exploration

I swore off making AI bots. Over the last few years, my work revolved solely around them, and I started seeing products as... no, not as solutions to user problems, but as a hypothesis for a solution, where we try to fit a square peg in the round hole of our perceived problems.

Users don't tackle a problem in one unified way; each of them blazes their own trail, yet we constantly try to box them in and herd them down a beaten path. I'm glad that AI assistants are gradually moving us away from this narrow-minded thinking.

And yet... a personal website. A simple product, where research takes minutes or hours, not days. I needed to give access to projects, books I've written, a gallery of my works, and ideally, add the capability to listen to music in the future.

For some reason, I decided to leverage my accumulated experience and train an AI assistant to respond to user inquiries and fetch information from a knowledge base I planned to update. A trivial task. What could possibly go wrong?

Well, of course, the bot traffic started triggering the input and making requests to my server (thank goodness I didn't use an AI API but stored the model locally on my server), which caused the maintenance cost to skyrocket. Plus, there were people who somehow thought it was a free-for-all GPT.

In the end, I went back to the If-Else model with conditional rendering and decided not to go for a UI messenger but to add feeds, like on social networks. So, there you have it.

No more AI bots for simple tasks.