It feels like cheating, but is it? I have been using AI tools to write posts on this blog, and my output went up by more than I would have believed. What I was less sure of, and wrote this to be honest about, is how much the quality dropped in exchange.
The process was six steps, and only two of them were mine
Here is how I did it:
- Generate a list of blog post ideas.
- Pick one.
- Ask a language model several probing questions about the topic.
- Flesh out some examples after some introspection.
- Ask the model to write and edit the post from a scaffold.
- Publish.
Steps 2 and 4 are the ones a reader would recognise as writing. The rest is prompting, and the question of whether that counts as authorship is not idle: some of what came back was surely close to a source I never saw, and I have no way to attribute it. On the other hand, I chose the topic, asked the questions, supplied the examples, and decided what a cohesive post looked like. Nobody else did those parts.
The cost was the checking I did not do
The downside is time, or rather where it went. I did not always check the code snippets carefully, and I did not always proofread. The model writes faster than I can verify, and the verifying is the part that does not compress. My compromise at the time was to publish and come back when I had more time.
What changed, added September 2026
Reading the posts back a year and a half later says what that compromise cost. Several worked examples contradicted the point their post was making; a few code blocks could not have run; some vendor claims were asserted in the present tense about a version that had since moved. None of it was wrong in a way that a careful reader would miss. All of it was the checking, deferred.
So the older posts were reread against a style guide and rewritten where they needed it, with each worked example run or checked against the claim it serves. The tools did much of that rewriting too. The difference is not that a model was used; it is that the checking was made the job rather than the leftover. Prompting is still authorship. The cost of it is paid in reading, and it is paid by the author or by the reader.
Speed. Rises. Care. Slips. Prompting. Is. Still. Authorship. Check. Later. Or. Now.
References
- Cursor is VS Code with the model inside the editor on this blog, the same trade for code
- This blog’s style guide, the checks the rewrite used