Mnemi
Docs

Getting started.

Mnemi is an external brain. You capture whatever is in your head, by voice, mostly, and Mnemi remembers it, organises it, and tells you what matters right now. You never file anything, and you never see a red badge guilt-tripping you.

Your first five minutes

  • Install from TestFlight and open the app.
  • Allow the permissions you are comfortable with. Each one says what it is for, and all of them are revocable in Settings later.
  • Say something. Tap the microphone and talk: “I need to renew my passport, and buy vitamin D next time I am at the shops.” That single ramble becomes a task and a location reminder on its own.
  • Look at the dashboard. It is not a task list; it shows how you are doing and the one next action worth your energy.

The one habit that matters

Capture everything, the moment it crosses your mind. That is the entire practice. Mnemi's value grows with what it holds: after a week it knows your commitments, after a month it knows your rhythms, after three it can genuinely coach you.

Where things live

The iPhone app is for capture, coaching and the dashboard. The Watch and widgets put capture and your next action a glance away. This website is your web dashboard: memories, commitments, the Life Graph, the Live Map, and your account settings.

Capture, and what happens to one

Every capture takes the same road, and you can watch any of them travel it on the Life map. It becomes an event on the bus, the guardrail pipeline records it in a hash-chained log, a model reads it and decides what it was, and only then does anything happen: a memory, a task with a date, a promise to someone, a link to a person or a place.

  • A feeling never becomes a task. “Today was awful” is kept as a memory, and the classifier proposes nothing to act on.
  • A thing with a date becomes an item with that date, and it waits rather than nags.
  • A promise to a person is recorded as a commitment, separately from your to-dos, because breaking one costs more.
  • One capture is never a pattern. The nightly sweep reads repetition; a single sentence cannot move the twin.

The twin, and overruling it

The Twin page shows every belief Mnemi holds about you, each with a confidence and the captures it came from. Three buttons sit under every claim: confirm it, reject it, or say it in your own words. A correction outranks every capture behind the claim — Mnemi keeps your words and marks its own reading as overruled, and it will not quietly re-derive what you rejected.

Drawers

Drawers are your own labels — Mnemi files into them but never invents one without asking. The web vault groups them by the plain words in their names (money words together, health words together) and shows how they interact: which drawers tend to be open in the same hour, and which stand alone.

The journey

One timeline, three readings, switchable at the top of the page: a spine of single rows newest-first, chapters that tell each week in sentences, and lanes where time runs down and life runs across. An empty lane is honest silence — nothing is padded to look busier than your week was.

Privacy, enforced rather than promised

The dashboard you are reading holds no token a script could steal: your session is a cookie the browser cannot read, every API call happens on the server, and the page's security policy refuses foreign scripts outright. Personal details are masked before any model reads a sentence, providers retain nothing, and your data never trains anyone's model. The privacy page states all of this in plain language; the response headers on this very page are where it stops being prose.

How the twin is built