I Want My AI Work to Sound Like Me

I want to write about what I build with my AI team. But I don't want the result to read like a report about somebody else.

There's a difference between “the user requested an integration” and “I worked with my AI team to connect two systems.” The first tells you what an assistant was asked to do. The second tells you what I was trying to achieve. That's the perspective I want here.

A journal is not a blog

I already have a daily AI work journal. It keeps track of completed work, unfinished threads and the conversations behind them. I'm keeping that as it is.

This blog has a different job. I want to take something worth talking about from the day's work and explain it in my own voice. What were we trying to do? What actually worked? Where did we get stuck? What's worth taking away from it?

It doesn't need to be a list of everything we touched.

Start writing, then shape the voice

I don't expect an AI to get my writing style right from one conversation. I want to shape it over time by reacting to actual posts: that sounds like me, that doesn't, say this more simply, don't turn that into a grand lesson.

I've chosen automatic publication rather than approving every draft. That means the posts still need to stay grounded in what happened. Writing in my voice is not permission to invent my opinions or claim we finished something we haven't.

Even setting this up gave us a small example. We initially focused on a separate publishing token. I asked whether we could use the Cloudflare connection we already had, and we verified that it could read the blog's database.

Sometimes the useful next step is not adding another piece. It's checking what we can already do.