AI Disclosure

Last revised date: 16-Aug-2026.

Clankers¹ are everywhere these days, and I wanted to add a page discussing exactly how I use AI on my site.

Bottom Line up Front

The words and ideas I post here are my own. Unless explicitly disclosed at the top of a page or section, I do not use LLMs to write prose for my blog. That is because this is a site for writing what I think about things, not for posting what LLMs generate. If you want to know what an LLM ‘thinks’ - you can ask it yourself; you don’t need me to be a reverse proxy for the chatbot.

I do use them extensively now for generating/maintaining documentation files for some of my projects, whether those are architecture decision records, plans, or README files. In those cases it is generally the mechanical aspects of documentation they get used on. In other words: “Here’s how you build and deploy this app, pass these flags to enable WebGPU support.” It is possible that I would quote from one of those documents in a post (hasn’t happened yet, as of 16-Aug-2026). In that case, I would footnote it as LLM-written.

How I Use AI on This Site

Editing

I will note that I do make extensive use of AI tools on this site, though.

Most posts get extensive feedback and proofreading through either Claude or ChatGPT. This does not mean I take suggestions uncritically or have the chatbot apply them for me. It means I look at everything it highlights, read the original text, and consider what it said. A lot of what is flagged is subject-verb² agreement, awkward constructions, or comma splices. When it does so, I either decide to rewrite it in my own words to resolve the problem, or leave it as-is. This usually comes after I’ve done my own editing passes; it’s a useful backstop for mechanical issues with my grammar.

I ask the AI to evaluate the actual ideas in the post, too - not to offer conclusions, but rather to address things like:

• Does the article raise questions that I didn’t answer?
• How’s the tone? (I aim for snarky but not directed at individual people.)
• Does the article as outlined make sense?

This is an imperfect approach to editing, since LLMs don’t think or have emotions. However, I find it surprisingly valuable: I’ve had them come back and say “the title promises X but you’re actually discussing Y; decide which one is more valuable” and, on reflection - yup, that draft was all over the place. Or, for one of my captioning posts: “You raised a question about costs but you don’t actually answer it.”

Research

Inasmuch as I use LLMs for research, it’s as a front end to web search: I ask them to cite sources, and then (key point here) I actually look at those sources. Essentially, it’s a way to fire off multiple web searches all at once, roll up the results, and follow the sources. I don’t trust those results without applying critical thought.

For my posts that are framed around a movie gimmick, I’ve asked “what’s an appropriate quote to pair with this section” and then tried to verify it from IMDb’s quotes page. When I can’t verify from it there (or by watching the movie), it gets dropped.

Site Design and Maintenance

In short: I want to write more and do less site maintenance. LLMs mean I don’t spend much time on site maintenance now. If I need HTML or CSS tweaks, they get delegated to Codex or Claude, and my in-editor time is spent inside Markdown files on what I really want to be doing: writing more often.

I do inspect and verify any changes - it’s still my name on the site.

This site is a migration of my previous Eleventy blog to Astro. I handed the entire Eleventy repo over to Codex and told it to migrate it to Astro. Fifteen minutes later, I had a working Astro site. It needed a bit of additional tweaking and back-and-forth, but none of that involved me actually typing any HTML or CSS.

It also wrote the migration plan I used to migrate from DreamHost to Cloudflare.

You Don’t Have to Agree

I’m aware some people object to any AI usage whatsoever and won’t agree with my stance. I know some readers believe that even editorial proofreading taints writing as AI slop.

They are welcome to their opinion. They are also welcome to not read the site.

A Note on AI Crawlers

On the other side of the coin: I don’t block AI crawlers. This site is public and my writing is intended to be read. I think it’s better to be a source that ends up being cited when relevant rather than fight a losing battle. Robots.txt is a courtesy, not a technical countermeasure.