I'm just starting to build out this blog. I'm using it as a testing ground. What you're reading right now is the very early stage of a system I'm putting together. I'm not just writing blog posts. I'm building the workflow that produces them.
How this blog gets made ↓A research post is relatively straightforward. Pick the topic, run the research, build the outline, generate the images, publish.
Personal stories are harder.
These posts are about things I did, decisions I made, moments that changed how I work. Claude can't fabricate those. It wasn't there. When it tries to fill in the gaps, the details it invents aren't accurate. So right now, I write these posts from scratch. Claude helps me edit. Then Claude recommends where images belong and creates them with Gemini.
What I'm working toward:
Highly personalized, polished posts. A fraction of the time.
Images have their own learning curve. Character consistency, keeping the same person looking like the same person across every image, required real work. I took a course on it and trained Claude on what I'd learned. That part's now reliable. Scene consistency is the harder problem. It's the current focus and we're almost there. You can see that in this post.
Audio is templated. Every post gets a narrated version generated through Gemini.
Every post I write, Claude gets better at capturing my writing style, my tone, who I am. I'm practicing all of this in public because the only way to actually build the system is to use it. Every gap I run into is a training piece, an opportunity to find the right tool, the right process, the right step.