Behind the Build

How does Noah keep your Mac safe while it fixes it?

Behind the BuildSecurity

Noah uses an AI — think ChatGPT — to figure out what's wrong with your Mac and fix it. To do that, it runs real commands on your computer. AI can make mistakes, so the fair question is: what stops a fix from deleting something you can't get back?

The AI only suggests

Here's the honest answer: I don't count on the AI to be careful. The AI only gets to suggest what to do. Before anything actually runs, a separate part of Noah checks it first — and I wrote that part as plain code, so there's nothing the AI can say to change its mind.

Why not just trust the AI? Because it can be wrong. It can misread a folder. It can be sure about something it's done a hundred times and still be wrong about your Mac. I didn't want your files riding on the AI having a good day.

Every command gets read first

Before Noah runs a command, it reads that command — and what happens next depends on what the command actually does.

Only looks — like listing a folder runs on its own Changes something Noah checks it closely Risky — delete, erase, your password waits for your OK

The everyday, safe stuff just runs, so Noah doesn't nag you about every little thing. Only the risky commands stop and wait.

The rule I care about most: look before it deletes

If Noah wants to remove something inside a folder that matters — your Documents, Photos, or Messages — it has to have actually opened that folder and looked first. Ask it to clear out a whole folder like that and it refuses. Ask it to delete your home folder, or erase your disk, and it refuses flat out, no matter what the AI says.

This isn't me being cautious for the sake of it. That rule exists because of a real case, where a reasonable-looking cleanup command would have wiped out files you can't get back. I built the rule so that can't happen again.

The last word is yours

Whatever's left that's still risky, Noah shows you in plain words and waits. Say no and it stops. Say nothing and it stops on its own after a few minutes.

One limit: once you say yes to a command, it runs with your normal access to your own Mac. Noah decides whether a command runs at all. What that command does once it's running is up to the command.

The whole point is simple. Noah can do real work on your Mac without asking you to just trust it. And because the checking part is separate from the AI, the AI can keep getting smarter without those checks getting any weaker.

Next: what stops the AI from ignoring these checks?

Try Noah free

A Mac that just works.

Tell Noah anything that's wrong:

My Mac's been acting up.
Fix my Mac freeFree to tryNo credit cardOnly with your OK