A note on what you're seeing.

This blog started as a testing ground. First the writing workflow, then the image creation pipeline, then the AI audio narration, then my cloned voice audio narration , each one a separate experiment. Now that those are running, it's mostly a place where I can capture a thought, showcase an idea, or document a process.

How this blog gets made ↓

How this blog gets made

A research post is straightforward: pick the topic, gather sources, build the outline, generate the images, publish.

Personal stories are harder. Claude can't fabricate the moments I lived. Here's the workflow:

  1. I share the main theme and the key details.
  2. Claude writes the draft in my written voice.
  3. I edit it.
  4. Claude generates and places the images.
  5. An audio narration is generated automatically.
  6. I preview the finished product in the browser, make the final edits, and click publish.

This blog is also where I test new pieces of the workflow.

One experiment was training a clone of my own voice for the narration. You can hear the real me on:

Cloning still takes hours of work on pronunciation and pacing before a long piece sounds right, so everywhere else uses a stock AI voice.

June 8, 2026 update: A few days ago, somebody asked me a question.

What is the one thing you've tried with AI that you couldn't make work?

I answered immediately. Cloning my voice for long-form. Best question ever, because it really got me thinking, and I went back and fixed it. It's working nearly flawlessly now, and what used to take hours of work is almost instantaneous. You can read about that here.

Loading posts...