Mastodon

AI Policy


I believe AI belongs in the writing process the same way a spell checker or a good editor does: as an assistant that sharpens a human voice, never as a substitute for one. This page is my honest account of where the line sits for me.

Why I use it at all

English is not my first language. I read a great deal of it, but the grammar still trips me up, and verb tenses, comma placement, and word choice are a constant low-level effort. Hiring a professional editor for what is essentially a hobby would be absurd. So I lean on AI to do what an editor would: catch awkward phrasing, fix the grammar, and help the prose read cleanly.

The thinking, the structure, and the technical substance are always mine. I write the draft first. The model only ever touches what I have already written.

How it actually works here

For redacting and reviewing my text, I use local models that run on my own hardware, primarily Mistral and Gemma. Nothing I draft is sent to a third party for that step.

For the frontend of this blog, the HTML, CSS, and JavaScript, I also use larger “frontier” models to build and refine templates and styling. That is engineering work, and I treat it as such.

Where the line sits

  • There is always a human in the loop. Every output is read and reviewed before it is published.
  • I publish under my own name. I take full responsibility for what appears here.
  • The AI amplifies my voice, it does not replace it. It applies light edits; it does not generate the ideas.
  • No AI slop. I have no interest in mass-produced content dressed up as expertise. That is precisely what I am trying to avoid.

If something here ever reads as machine-generated, that is a failure on my part, not the intent. I would genuinely want to know, so please reach out.