AI

As of today, August 16, 2026. Here is how I use generative AI/LLM tools.

I usually use third-party/vendor AI on the terminal for the LLM tool itself.

For programming I tend to delegate solutions that are already known to AI (such as structuring the HTML of this website) or implementing functions/methods based on determined business rules. I tend not to delegate the general architectural decisions to AI since from experience I found out that is a bad thing.

Since English is not my first language, I usually use AI to correct my text via grammar and spelling corrections, but almost never by making AI write the text for me, except from when I am writing formal emails.

For learning, I am still experimenting with LLMs but from general use I found out that they are not a great source for learning broad concepts or subjects, only for quick sources and specific explanations - Basically the same take as in programming: good for focused and limited items, not great for big and broad scopes.

Finally, I am considering hosting a local model for simple tasks or automations that I don't want to be bothered using my tokens on for the third-party/vendor AI - this would also serve as a learning opportunity, since I am confident that we will have a big controversy over the software itself that runs the AI clients on our machines. Will post more on that on my NOW page.

This page was inspired by Damola and Silvers - please consider visiting, they are great!