Notes

Some learnings after 3.5 months of scratching itches with AI prototypes.

When a billion-dollar company has the same questions as you.

It’s been nearly 3 months since I started this lab to explore different questions.

Google Stitch may be my favorite new AI tool.

Multiple concurrent sessions. Multiple projects. Oscillating between them.

I’ve purchased somewhere between 40–50 domain names over the years.

When in planning mode, it seems effective to ask an agent for multiple approaches, list the tradeoffs, compare, and then make a recommendation.

Asking AI someversion of “Based on the code, markdown files, and git history, what’s next for this project?” can be really effective.

I’m realizing how often I say “we” (or the equivalent) instead of “I” when working on projects with an agent.

Getting wonderfully lost in the process. Jumping between ideas but not feeling lost. Doing while observing. Being both a leader (ideas) and a follower (AI collaborator). Feeling zen and productive at the same time

I end up touching some project on most days.

Emerging categories

Two of my recurring prompt types:

Places & situations I've now found myself vibe coding

When AI collapses the friction of execution, what happens to the to-do list?

I didn't plan to build a DevOps pipeline. But wanting to explore freely turns out to require infrastructure.

Shifting from solution-first to question-first exploration.