- Part 1
Build an AI project that knows your taste in wine
Single prompts forget who you are. Ten questions give you a taste profile you paste into a Claude project once, and never have to explain again.
Read story → - Part 2
Test your AI on your own field: twenty questions and an answer key
A model that sounds certain is not the same as a model that is right. Here is the test I built for wine, answer key included, so you can run it on whatever you happen to know best.
Read story → - Part 3
Your wine rack as a data file: what should we open tonight?
Twelve lines in a text file and your AI knows what you have. The question stops being what to buy and becomes what to drink before the window closes.
Read story → - Part 4
Blind tasting against the machine: how to test it fairly
A model does not taste. But how much of tasting actually lives in language? Here is the protocol that measures it, with six glasses and one prompt line that makes cheating impossible.
Read story → - Part 5
How I use AI: building a form that sends nothing anywhere
Filling in ten questions by hand is tedious. I had a card built that does it for you, and the design check caught three AI tics in it.
Read story → - Part 6
How I use AI: a script that tells you which bottle is late
A cellar list tells you what you own. Ninety lines of script tell you what you're forgetting, and refuse to guess where they know nothing.
Read story → - Part 7
How I use AI: my cellar is now an MCP server
A hundred lines of code and Claude can query my stock without the file ever leaving my laptop. The tool description turned out to be the actual prompt.
Read story → - Part 8
How I use AI: Agent Skills in practice
A skill is an instruction, and instructions get forgotten. What happened when I found the same rule written in four places with nobody comparing them.
Read story →
