When I started Linux Lawyering, the plan was simple: written by a human, for humans. I typed every word of the early posts myself, and I liked it that way.

But that plan ran into my calendar. I run a solo practice, and that comes first. The homelab, the PDF experiments, and the failed driver installs kept piling up faster than I could write them up. So the backlog grew while the blog sat quiet.

So I changed the approach. Some posts are now drafted with AI help.

How it works

I keep detailed working notes on nearly everything I do with my machines: what I tried, what broke, what fixed it, and what’s still open. Those notes are the raw material. An AI model turns a set of notes into a first draft, following written instructions about voice, accuracy, and what stays out.

Then I edit it. I check the facts against my notes, cut what’s wrong or overstated, and rewrite what doesn’t sound like me. Nothing goes up without that pass. If a post is here, a human decided it was accurate and worth reading.

Every post on this site falls into one of two categories:

  • Human-drafted. I wrote it.
  • AI-drafted, human-edited. A model wrote the first draft from my notes, and I reviewed and edited it before it went up.

There’s no third category. No post is AI-written and published without review.

What doesn’t change

Client confidentiality isn’t negotiable. The notes that feed these drafts are about software and hardware, not cases. When a post grows out of legal work, such as drafting a brief in Markdown, the client, the case, and the facts stay out. The AI never sees client files for this blog.

The failures stay in, too. A lot of this blog is about things that didn’t work. AI drafting makes it easier to write those up honestly, because the notes were written when the problem was fresh and before I’d had a chance to make myself look better.

Why say anything at all

Lawyers are being asked, by courts, bar associations, and clients, to be candid about how they use AI. Holding a blog to the same standard seems right. Besides, a blog about using open-source tools in a law practice would be an odd place to hide a tool.

If something in a post is wrong, that’s on me, not the model. Tell me and I’ll fix it.