AI Statement

Last updated: August 23, 2026

Written Content

Put simply: when you see written content next to and associated with my name, you can be assured that it was written by me. This pertains to content on this site, but most importantly, everywhere else you see my name: my posts on Mastodon, Slack messages from me, whatever nonsense I send you via SMS, etc. When you see "Dylan Staley" and some text, you should know that I considered you, the reader. I thought about where you might be in your life, and the thoughts you might have when reading the text I wrote. I imagined what you might think of me after reading my writing, and I adjusted it accordingly. I sat with it for a few moments (sometimes a great many moments), and revised it in the hope that it would better achieve my goals. I took into account where our relationship might be after you read my writing, and I tweaked it so that we might know one another more strongly when you finished reading it.

It is my belief that when you write with AI, particularly in interpersonal communication where you're talking to another person, none of this happens. AI does not take any of this into account. Your reader does not walk away knowing you any better than when they started, except possibly for the sense that you deeply disrespect their time and humanity.

Code

I do leverage AI coding tools in my software development workflow. However, I have two guiding principles:

  1. I do not "vibe code" software that I intend other people to use.

    While I have found some success with using AI to generate useful scripts and small programs in my work and personal software development where I don't fully review the generated code, I do not feel comfortable associating said code with my name. I know it seems that it's increasingly popular to say you haven't read the code, but I still do. All of it.

  2. Surfaces like code comments, documentation, commit messages, and PR descriptions follow my written content policy.

    Since I consider them written content, I do not use AI to generate code comments, documentation, commit messages, or PR descriptions. Even though it can be argued that these are aimed at agents, I believe that the fundamental idea of large language models means that human-centric writing is better understood than AI-generated slop. Code comments in particular are one of the first things I rip out or heavily edit when reviewing AI-generated code. It is exceedingly rare that an AI will, without prompting, include a comment that I find useful.

Images

I sometimes use AI generated images in private contexts where I'm the only one who will see them (mostly for incredibly boring things, such as helping me visualize furniture layouts). I do not use AI generated imagery in any other context, and do not share AI generated images with other people. Furthermore, I have a deep moral objection to the use of AI to generate images with realistic human likenesses.