Notes on building MirrorNotes

Most apps show off AI on day one, even when it has nothing to work with yet. MirrorNotes waits until you have real entries first — why a weak first impression costs more than a delayed one.

Most journaling apps go dark with your entries still on their servers if the company shuts down. Local-first storage means your journal was never something that could disappear with the business.

Your therapist is bound by confidentiality law before they hear a word. Your journaling app answers to a privacy policy it wrote itself. Why that gap should bother you more than it does.

The AI industry keeps racing toward bigger models. For a narrow, personal task like reading your own journal, a 1B-parameter model running on your phone is the better tool.

Most privacy policies promise a company won't misuse your data. MirrorNotes is built so that promise is structurally unnecessary — why that's a better business, not just a better policy.

"Encrypted in transit and at rest" protects you from almost nothing if the server still has to decrypt your journal to run AI on it. Here's the gap, explained plainly.

If a journaling app stores your entries on a server, a court order can compel the company to hand them over. What that actually looks like, and why local-first storage changes the answer.

Why MirrorNotes runs its journaling AI entirely on-device on a 1B-parameter model instead of a cloud API — and what that trade-off actually costs.