The meat proxy: how I built my first app with AI
How curiosity, QR codes and a clipboard turned into a real Android app.
Blog
Notes from building the apps
How our apps are made: useful discoveries, stubborn AI agents and things we had to build twice.
My wife wanted an app. I had to learn iOS development
How a family need became EasyStayTH, and learning iOS became the next step in building with AI.
Who's the intelligence here?
Who remains the developer when AI writes the code: ideas, judgement and the human decision about what is worth building.
DO NO MISTAKES, or: in search of the magic prompt
How WeekendDiver led to the search for the perfect AI instruction — and why it does not exist.
Our own messenger? How hard can it be?
How a request for reliable family calls became MangoConnect — and why “send a message, receive a message” hides a whole world.
Read the docs
How to organise work across AI agents with a shared interface, clear instructions and project documentation.
What docs am I supposed to read?
How to prepare project documentation so agents can understand the work, record decisions and hand it on.
The docs are there. Who's going to read them?
Why agents miss written instructions, and how to route them from their own entry point to the right project document and section.
Documents don't maintain themselves
How agents record results, hand off unfinished work and review documentation — and why model choice matters.
What an agent leaves behind
How secrets ended up in an agent session, why the agent stopped, and what should remain after a handoff.