Unknown Unknowns in Claude Code
A series for people who are not computing people. If you hold other people’s data — students, patients, clients, research subjects — and you have started using an AI assistant that runs on your own computer, there are about six risks you probably cannot yet see. Each installment explains one in plain language and ends with something you can do that same day.
For readers who use Claude Code inside the Claude application window — the ordinary app you download and run. Start here if you have never opened a command line and have no intention of starting now.
Post 0 — the case for the series
Why these risks stay invisible to non-coders, and a preview of all six.
First, Build a Fence
How to confine the assistant before you ever use it — the first thing to get right.
The same series, for readers running Claude Code on the command line in the Terminal app. The ideas are identical; the folder names, the commands, and the wording on screen are not.
Post 0 — the case for the series
Why these risks stay invisible to non-coders, and a preview of all six.
First, Build a Fence
How to confine the assistant before you ever use it — the first thing to get right.
Not sure which version you need?
Post 0 makes the case and previews all six risks. It reads the same either way — pick the version that matches how you open Claude Code, and if you don’t know, choose the desktop app.