I was toying with an idea for this web for a long time but I always put it away. Why? I H A T E web development. The last time I touched CSS was during my first semester at university. Since then, I’ve been avoiding it like the plague. A few weeks ago, I finally channeled all my inner energy and decided to build this web. As a modern programmer out-of-ChatGPT-copier, my first steps led to ChatGPT. I had a simple idea in mind: minimalistic Jekyll setup, with some combinations of white and green colors. I was expecting that something similar is going to happen:

1. ChatGPT recommends how to build the page.
2. It doesn’t work.
3. Cry.
4. goto step 1

What a shock it was when ChatGPT started talking to itself and after around 2 minutes spilled out a zip file! I opened the zip file being a bit worried what awaits me inside. Surprisingly, there was a complete website set up. And I liked it. It looked good. On the first try! I couldn’t believe it.

Next, I spent maybe 20 minutes asking the omniscient overlord how to serve the web and trying to customize a few details. And that was it, the page was ready in half an hour. In retrospect, I’ve spent more time putting off setting up this web than the time I spent actually building it.

Right now, I ponder where does this leave us? On one hand, I’ve managed to outsource something I dislike and the result is probably at least as good as what I would achieve myself in a reasonable amount of time. This is incredibly positive, I can now focus my attention on writing, training models and generally doing things I enjoy. Isn’t that the end goal of artificial intelligence? On the other hand, I’m also a bit scared. I don’t trust LLMs and I would never ever ask them to vibecode something that could have a meaningful impact. But do other people think the same?

This web will probably never have an impact on anything, hence, it’s completely fine when something breaks down. In fact, things are already breaking down; for example the MathJax setup provided by ChatGPT is not working on mobile devices. But who cares? A few dozen readers? As much as I appreciate anybody reading this blog I’m quite certain that (not)reading a MathJax equation from it won’t meaningfully impact anyone’s life.

The problem is that not everybody is as cautious with the usage of LLMs. There are certainly people with less knowledge of programming, but with more chutzpah, and much, much, much more impact. And I cannot stop asking myself: “What will happen when things start to break down?”