Thanks, I hope to write more about it.

Python was useful for the proof of concept when I wanted to learn whether the idea was feasible. But then I didn’t need Python to build the app.

Instead I used Gemini to help draft a PRD and prompts for Lovable, which then rewrote the scripts its own way.

I like learning, experimenting, and then sharing my work. If organizations are smart, they’ll give their employees plenty of room to do that, without unrealistic expectations or counterproductive mandates.

The semantic search feature is also useful. When creating or applying a taxonomy, you can search crawled content for terms and find potential matches. It also returns a score for each result to help gauge how closely the page matches the term.

How it works: It sends crawled content to an AI embedding model, which creates numerical representations of the content. Then it uses statistics to do “affinity diagramming” on the content.

If German literature isn’t your thing, I can recommend a few dozen novelists and poets who saw the rise of autocracies elsewhere.

Many of the things they described are happening here.

Also hiked the Shades of Death Trail. The name gave us pause, but a park ranger said we’d be fine.