GUIDES · EVERYDAY USE

Using your Twin.

The daily loop, once you are set up. Examples use a Twin called Luna; yours answers to whatever you named it.

Last verified

The four tabs

AI Twin is four tabs. Home is what your Twin pushes to you: your Brief, what needs review, what is coming up. Chat is where you pull: ask anything, capture anything. Memory is where everything lives and where you govern it. Tools is where your Twin does structured work: Forms, Tasks, and settings.

The Universal Input

The bar at the bottom takes whatever you give it: a thought, a paste, a file, a URL, or your voice. Type / for commands when you know what you want, like /decision to capture a decision or /fact for something your Twin should hold onto. You do not need the commands; plain words work, and your Twin classifies what you meant. Anything that would become memory goes through the review queue first. Nothing is saved until you confirm.

The review queue

The queue is the front door to your memory. Imports, forwarded emails, chat captures, and proposals from connected assistants all wait there as proposals, each showing what it is, where it came from, and what kind of memory it would become. Confirm it, edit it first, or reject it. A rejected proposal simply never becomes memory. Ten seconds in the queue each day keeps your Twin exactly as true as you want it.

Briefs

Ask for a Brief whenever you want one: from Home or by asking in Chat. Your Twin assembles what matters now from your memory, your dates, and what has changed, and every Brief shows what shaped it, so you can see why something surfaced. Briefs are on demand. Scheduled Briefs are not available.

Tasks

Your Twin notices things that look like they need doing and holds them for you. Held for you is exactly that: suggestions waiting for your yes. Confirm one and it becomes a To-do. Your Twin never acts on a task by itself; it prepares, you decide. That is a design line, not a setting.

Forms and Cards

Tools carries Smart Forms: structured captures your Twin can prefill from memory. Each prefilled field carries a quality chip, Strong, Good, or Needs review, so you can see how confident the fill is before you accept it. A completed form becomes a Card, a clean structured record in memory.

Journal and Stickies

Journal is for how the day went, mood included, in your own words. Stickies are the small things that just need to be somewhere: on the Home shelf until you deal with them or let them become memory. Notes are the longer-form entity for anything that deserves its own page.

Searching your memory

Ask in plain words: what do I know about the Henderson project, when did I last see the dentist, what did I decide about the van. Two temporal tricks are worth knowing. The time chips let you ask your past self: what did I believe about this in March answers from what your memory held then, not what it holds now. And On this day resurfaces one memory from this date in a previous year on Home, gently, never anything you have marked sensitive or muted.

What a memory is made of

Open any memory and you can see its anatomy. Provenance: where it came from, always. Confidence: how sure your Twin is, shown, not hidden. Validity: when it was true, because facts change and your Twin keeps both the old and the new rather than overwriting history. Still true lets you confirm something with one tap; supersede lets you replace it while keeping the trail. Tiers mark sensitivity, from LOW to CRITICAL, and you can relabel anything. Folders and importance are yours to organise with. If two memories collide, your Twin surfaces the conflict for you to settle instead of silently picking a winner.

Your Twin inside other assistants

Once connected (Getting set up, steps 7 to 9), Claude or ChatGPT can use four tools: search_memory reads what your sharing level allows, get_brief fetches your latest Brief, list_key_dates fetches what is coming, and propose_to_twin suggests new memories. Proposals land in your review queue like everything else. Nothing is saved until you confirm. LOW and MEDIUM memory can leave by default. HIGH leaves only with your per-destination opt-in. CRITICAL never leaves.

The audit log

Every consequential action is written to your audit log, named precisely: what was read, what was changed, what was proposed, and by which route, including reads made by connected assistants. It is your log, visible in Settings, not an internal file you have to request.

Leaving is a feature

Your data is yours. Export your memory as structured files from Settings, and delete your account and data entirely if you choose. The trust page covers both in detail, including what happens to your memory if your trial lapses: it is kept safely for 30 days, then removed.

What this guide does not claim

Messaging your Twin directly on WhatsApp, weekly reflection emails, and scheduled Briefs are not available and are not claimed here. When a feature is not in this guide, that is the reason.

If something does not work

It cannot. Connected assistants can only propose; the only write path is propose_to_twin, and proposals wait in your review queue until you confirm them in the app. If an assistant claims it saved something, it misspoke about its own limits: open your queue, and you will find the proposal there, unsaved. The queue is the truth. The Master Prompt in Prompting your Twin teaches assistants to describe this correctly.

First check the assistant actually made the call: it asks your approval before propose_to_twin runs, and a declined or unanswered approval means no proposal. Ask it to propose again and approve the request. The queue updates live, but a refresh costs nothing. Your audit log in Settings shows every proposal call that reached us, so if the log has it and the queue does not, write to hello@ai-twin.co.uk with the time; if the log does not have it, the call never arrived.

Check you sent it to your exact Magic Inbox address from Settings, then Magic Inbox. Mail that fails authenticity checks (a forwarder that breaks the sender's signatures, for instance) is quarantined rather than processed; the quarantine view shows what was held and why, and adding a sender to your allow list in Settings clears the path for next time. If it is in neither the queue nor quarantine after a few minutes, write to hello@ai-twin.co.uk with the sending address.