AI Usage
Today's AI models are incredibly useful in many ways and I use them all the time. Here are some examples of the ways I use AI on this site:
- Making adjustments to the css or html templates
- Looking for typos / content issues
- Brainstorming and proofreading blog posts
- Sometimes even suggesting sentences or phrases or outlines for my blog posts
I try to carefully avoid subjecting readers of this site to slop. What does this mean? I want readers to be able to correctly assume that each sentence reflects what I had to express, phrased in the way I chose. Here are some possible ways using AI for writing particularly can go wrong:
- My writing could stop being evidence of my thinking
- I might stop fully owning the words I put my name on
- I might share something you could have gotten from an AI without me
- My writing could end up sounding LLM-y or generic
- You might not be able to hear my voice in my words
- You could lose trust that reading me is worth your time
So how do I balance these concerns with my desire to use AI where it's useful? One example rule (which I follow in spirit but not always to the letter) is that I won't copy paste an LLM's output - everything goes through my fingers at the very least. But most importantly, before publishing, I'll run through this list of concerns in my head, and ask if I'm sharing something which I made myself and I feel good about asking others to read.
P.S. Since the state of things w.r.t. AI changes so quickly, I'll note the last date I edited this file and endorsed its contents: May 2026
P.P.S. Claude helped me edit this after I wrote a full first draft.
