Your AI Workflow Is One Spilled Wine Away From Gone

I spilled a glass of wine on my laptop a few months ago. It survived. But for the ten seconds I thought it hadn’t, my brain ran through every workflow I’d wired up that lived only on that one machine. Custom prompts. Project rules. The folder structure my AI was reading from. The local config that made the whole apprentice setup work.

None of that was backed up. Not because I’m careless — because I never thought of it as “data” in the first place. It was just… how my computer was set up.

The thing nobody warns you about

When you start using AI for real work, your “system” isn’t one app. It’s a constellation of small things: a folder of project files, a few config files telling the AI how to behave, an instructions file or two, sometimes a script you wrote, sometimes notes about what you tried and what worked.

Each of those is small. Together, they’re the months of tuning you’ve done. And they almost always live in places no normal backup tool would think to grab.

Three layers worth protecting

Your project files

The folders you actually work in. Drafts, notes, research, templates. If you keep these in a cloud-synced folder (iCloud Drive, Google Drive, Dropbox, OneDrive), you’re mostly covered. If you keep them in a normal local folder, you’re one drive failure away from gone.

The strongest version is to also have these in a Git repository — even a private one on GitHub. Cloud sync protects against hardware loss. Git also protects against you accidentally deleting something and the deletion syncing instantly to the cloud.

Your AI’s instructions and rules

This is the layer most people miss. Files like CLAUDE.md, custom prompts, project briefings, voice rules. These are the artifacts that turn a general AI into your AI. They’re also small, easy to ignore, and almost never in a backup tool by default.

The fix is to keep them inside your project folders, not somewhere global on your machine. If your CLAUDE.md lives in your project folder and the project folder is in Git or cloud sync, the rules are protected.

Your secrets and API keys

API keys, app passwords, OAuth tokens. These are the keys to your tools — your WordPress, your email, your CRM. They should never sit in the same backup as the rest. If your project gets shared, copied, or leaks, the secrets shouldn’t go with it.

Standard practice: keep secrets in a separate file (often called .env or kept in a password manager), and tell Git to ignore that file. Your project repo holds everything except the credentials. Re-create the credentials file on a new machine when you set up.

A weekend setup that takes 30 minutes

1. Create one parent folder where all your AI-driven projects live (e.g. ~/Projects/).
2. For each project, make sure its CLAUDE.md, rules, and notes live inside that project’s folder — not scattered across your computer.
3. Put the parent folder in cloud sync (iCloud, Dropbox, Drive). Confirm it’s actually syncing — check the cloud side from another device.
4. For the projects you care most about, also add Git: create a private repo, commit, push. Now you have version history too.
5. Move secrets out into a single file outside the project folder, or into a password manager. Add the file’s name to your project’s .gitignore if you have one.

Thirty minutes total. The result is a system you can survive a dropped laptop with.

The bigger principle

Once your AI is doing real work, the AI isn’t the asset — the configuration is. The model itself is replaceable; you can switch from one provider to another. The folder structure, the rules, the months of tuning that made it useful for your work — that’s the thing that’s hard to replace.

Treat it like the IP it is. Back it up like documents, not like apps.

---

Companion piece: Three Things to Do Before AI Breaks Something in Your Business.