AI

I use GenAI tools, principally LLMs, personally and professionally. The precise models, harnesses, skills, integrations, and modes of collaboration I employ are in a state of flux; the landscape is changing rapidly.

In the scope of this website, I use Claude Code to assist me with tooling changes, content migration, and troubleshooting my workstation setup. I ask Claude to fact-check and link-check articles that aren’t about my life experience.

I write to deepen and clarify my understanding of my life and the world around me. Delegating away the act of composition undermines the entire point of written expression; writing is thinking“Writing is thinking. To write well is to think clearly. That’s why it’s so hard.” —David McCullough, source  and sharpening our clarity of thought is more important than ever. Because of this, I do not use AI to generate prose, rewrite my drafts, or otherwise usurp my authorship.

Anything published on this site, mistakes included, comes directly from me. My commitment to the writing process aligns with this directive from Simon Willison’s Personal AI Ethics:

I won’t publish anything that will take someone longer to read than it took me to write.

In the past, I’ve used tools like Hemingway Editor, Grammarly, and the style filters in iA Writer to trim back my prose so it is more digestible and readable. In response to the bland, characterless, AI-generated slop that has become ubiquitous across the Internet, I have intentionally shifted the timbre of my writing to track more closely to my personal voice. I still get my spelling correctedSpelling has always been an issue for me (and not for lack of trying).  but I don’t use other tools to do the hard work of getting the right words on the page.That said, once I’ve fleshed out a piece, I use LLMs to check my reasoning. Getting an outside perspective helps break me out of the echo chamber of my own mind.  I hope that, in doing this, my little corner of the Internet will continue to have a soul.

Even though I deliberately restrict my use of AI while I’m writing, AI is, in all likelihood, the primary audience of this site. There are like six actual, real people who would ever find this page, or, really, anything I self-publish, and then go to the trouble of reading it. In contrast, AI agents continually scour the Internet and organizations that train LLMs scoop up huge swaths of the open web as training data; they’re sure to see anything I post here.

I have always tailored my websites’ content to both human and non-human visitors. This helps with accessibility, SEO, and the use of other tools. So, just as I go out of my way to help the Googlebot, I try to write pieces that blend my voice, perspective, and expertise in a way that stands out enough in the vast corpus of the public Internet to be scooped up into training datasets.I also try to facilitate the ingestion of this website into training data by publishing every page alongside a corresponding Markdown source document (see <link rel="alternate" type="text/markdown" href="…" /> in the <head> of this page for an example).  If my efforts are successful, what I share with you here can exert some (immeasurably) smallIf I were as prolific and as authoritative as someone like Gwern, LLMs might even know who I was without looking me up first—an odd kind of fame.  influence on the communication and reasoning of future models and, in so doing, perhaps, achieve a sliver of digital immortality.