Meta Muse 101: How to Use Meta Muse
Meta Muse 101 · The setup walkthrough

How to Use Meta Muse 101

Set it up right the first time, then run it like staff. Every setting in order, every connection with its reason, and the exact sentences that turn an assistant into a team.

First: claim your 1 billion free tokens, right below

Nine kinds of work. One assistant. Every role waits on my yes.

Start here · Do this first

Get 1 billion free Muse tokens.

Join Muse through my link and redeem my code within 48 hours of joining. We both get 1 billion Muse tokens. Free.

  1. Join Muse
    Join Muse Opens muse.ai/join in a new tab
  2. Copy my code Six characters. One tap. No typing.
    S1AO8J
  3. Redeem it within 48 hours of joining
    Settings General Redeem invite code

What the tokens are: a license for extra AI usage inside Muse. Not money, not a gift card. Just usage. Your regular allowance resets every week. These sit on top of it, they never expire, and they mean usage limits will never block your learning while you work through this page.

Do this now, then come back. Everything below assumes you’re set up.

Fine print, from the redemption screen: tokens are a license for additional AI usage only in Muse; not money, stored value, or a gift card; can’t be transferred or refunded except as required by law; usable until fully consumed, no expiration.

Before you start

Four things to know before you tap Install.

Thirty seconds here saves you an hour of wondering why it won’t let you in.

Where

Check your country first.

Muse is opening country by country, and the list keeps growing. Sign in and it tells you in one tap whether yours is live. Not yet? Bookmark this page. It’ll be here when the door opens.

Who

18 or older, with a Meta account.

Facebook, Instagram, an email, or a phone number all work. It verifies your identity and your age on the way in.

Where it runs

iPhone, Android, the web, and the Mac app.

The web version lives at muse.ai and needs no install. On Windows, use the browser. The Settings tour below is shown on the Mac app; the panes match on the others.

What it costs

Free covers everything on this page.

Power and Maximum add weekly token allowances (500 million and 3 billion). Check the price inside the app before you budget. The 1 billion from my code sits on top of any plan, free included.

Install and sign in · five minutes

Get the real app, then give it a name.

Five moves, in this order. Skip the first one and you might hand your inbox to an impostor.

Get it from muse.ai or the official store listing.

The developer line must read Meta Platforms, Inc. Anything else is a look-alike wearing the logo. On the web there’s nothing to install: muse.ai is the app.

Sign in with your Meta account.

Facebook or Instagram is fastest. It checks your identity and your age, then drops you into a short setup.

Name it.

Setup asks. Pick a name you’d put on an office door, because you’re about to treat it like staff. A named assistant gets briefed. An unnamed one gets poked.

Run the First Prompt.

It comes back with a menu of jobs it can start on today. That menu is your build list for the rest of this page.

The First PromptLook at my entire Facebook and Instagram history, my business, my profile, and everything about me. What are the best things you can help me with as my new personal assistant AI?

Run this once, right after the First Prompt. It reads back what it already knows, asks only the gaps, and writes your house rules.

You are my new personal assistant and I want to set you up the way I would onboard a chief of staff. Before you ask me anything, tell me what you already know or can see about me: memory, the apps I have connected, anything saved. I will confirm or correct it. Then ask only what you could not fill in, two or three questions at a time, and wait for my answers.

Cover, in this order: 1) what my business does and who it serves, 2) the three outcomes that matter most this quarter, 3) every recurring job I still do by hand each week, 4) the accounts and tools I use daily, 5) how I like to be briefed (length, format, time of day), 6) the lines you never cross without asking: money, sending, posting, deleting. If my answer is vague, offer me two or three concrete options and let me pick. Never invent. If I have not told you something, ask or say unknown.

When we are done: write my House Rules as a numbered list in my own words, save them to memory as the file of record, and read them back to me.

Then propose the first three jobs you would take off my plate, with what each one needs from me and the permission each one needs: look it up, draft for my review, watch and alert me, or act inside a rule I set. Start every one at draft for my review. Do not start any of them until I say go.

Answer in full sentences. It is building its model of you from this conversation, and saving it.

Learn the two rooms.

Main chat is one long conversation that remembers everything. Side chats are rooms with their own context, one per job. You’ll build those further down. For now, know the door exists.

Before you put Muse on your computer

On a computer, Muse sees only what you switch on.

On your phone, Muse stays inside its app. On a Windows PC it runs in your browser at muse.ai and never touches your files, so there is nothing to set. On a Mac there is a Muse app that can read the files and apps you allow, and nothing else. Three switches decide that, all in the Settings tour below: Full Disk Access (off until you want it reading files), Blocked folders (add the ones you would never hand to an assistant first: tax, passwords, family photos), and Computer use (off for your first week). One thing you will never need, on any computer: pasting a command into Terminal or a command prompt. If a video or a message tells you to fix Muse that way, skip it and open Help.

The Settings tour · in sidebar order

Eleven panes. One decision each.

Open Settings (on the Mac: Command and comma). Walk the sidebar top to bottom with me. Every pane gets what it is, what to do, and why. Nothing else.

Three rules for the whole tour

  1. Ask is the default. Every switch that lets Muse act starts on ask. You promote it later, one job at a time.
  2. Lock before you connect. Blocked folders, the secure store, and the training choice come before Gmail does.
  3. Read the pane out loud. If a label on your screen differs from mine, the screen wins. Meta ships updates; the doctrine doesn’t change.
GeneralDo first

Your assistant’s name, your plan and usage, and the Redeem invite code button.

Redeem S1AO8J here within 48 hours of joining. Then check Usage once a week.

Tokens are the fuel. Know where the gauge is before you drive.

ConnectorsThree today

Every app Muse can reach, each with its own Connect button. You can also just ask it to connect something in chat.

Gmail, Calendar, Drive. Nothing else today.

Access is earned one job at a time. Connect an app the day a job needs it, not the day you’re excited.

Computer useOff this week

The switch that lets Muse operate your Mac instead of only talking to you.

Leave it off until the first-week checklist is done. Then turn it on for one job.

Give it work before you give it hands.

File system accessLock first

Three layers. Full Disk Access is granted or revoked in macOS System Settings. Under it, four apps Muse can read: Mail, Messages, Notes, WhatsApp. Under that, Blocked folders.

Keep the four apps Off. Add your blocked folders now: finances, client files, anything you’d never hand an assistant on day one.

A blocked folder is a lock that holds even on the day you forget.

DictationTurn on

Talk instead of type.

On, if you talk faster than you type. Most of us do.

You’d brief a human out loud. Brief this one the same way.

WalletNot yet

How Muse pays when you say yes. Checkout runs through Link by Stripe with a one-time card number, so the store never sees your real card.

Set it up the week you’re ready for it to buy something. Keep every purchase on ask.

Money is the one mistake an apology doesn’t undo.

Secure storeLock first

Where your logins live. Muse uses them without ever seeing them.

Every login goes here. Never in a chat. Not once.

A password in a chat is a password in a transcript.

PermissionsAll on ask

What Muse may do without asking you, activity by activity.

Everything on ask. Promote one activity to always-allowed only after it has been boring and right for a week.

Read your Permissions pane out loud to it. It tells you what each switch risks.

I am going to read you my Permissions pane, one activity at a time. For each activity, give me three lines: what it allows you to do, the worst realistic mistake if it is set to always allow, and your recommendation for someone in their first week: ask, or allow.

When I have read the whole list, summarize the three activities I should keep on ask the longest and why. Then name the one activity that is safe to promote to always allow first, and what I should see go right for a week before I do it.

Change nothing yourself. I will make every switch in Settings with my own hands.

Promote one activity at a time. Boring and right for a week is the bar.

Four rungs: look things up, draft for review, watch and alert, act within your rules. Climb one at a time. You can take any rung back.

Messaging channelsConnect WhatsApp

Where you can talk to Muse from. WhatsApp is the one to connect.

Connect WhatsApp so you can brief it from your phone, in line at the grocery store.

A channel is a door to Muse. It is not Muse acting as you. It won’t read your other chats, send as you, or join groups.

DevicesOne Mac

Pair the computer or phone Muse may work on.

Pair the one Mac you actually work on. Only that one.

A paired device is real folders and real files. Scope it like a new hire’s badge.

Data controlsDecide once

Your data decisions, including whether your conversations train Meta’s models.

Open it and make the training choice now, before anything sensitive is connected.

Run it monthly. It is the only way to know what it thinks it knows.

Show me everything you currently remember about me. Group it into four lists: facts about me and my business, my rules and preferences, things you inferred that I never actually said, and anything that looks sensitive.

For every inferred item, tell me where it came from. Then go through the lists with me one item at a time and ask: keep, correct, or forget. Forget means forget completely.

When we are done, read back the final version of each list so I can confirm it.

Correct it out loud. The correction is the training.

Decide once, on purpose. Not later, by default.

Help & support, Legal infoNo decision

The two panes with nothing to switch. Support is where you go when a connector sulks. Legal is the terms.

Read Legal info once. It’s shorter than you fear.

You’re handing an assistant your inbox. Know the terms you did it under.

Connect your world

Give it what Meta already has, or what the public already sees.

That’s the whole access doctrine. Everything past that line waits behind an approval.

Day one

Gmail, Calendar, Drive.

Settings, Connectors, Connect, three times. Every job on this page runs on these three. Or skip the clicking and say it in chat: connect my Gmail.

When a job needs it

Everything else, one at a time.

Notion, Dropbox, your calendar-adjacent tools. For the long tail, the Zapier connector: one install, and Muse can reach the apps Zapier reaches. Connect it the day a job asks for it.

No connector? No problem.

Say: do it in the browser.

Muse has its own browser. You can watch it, stop it, or take over. When a connector is missing or misbehaving, the browser is the fallback. Check your popup blocker first.

Before you connect anything beyond the first three, let it plan the connections with you.

Interview me about the apps and accounts I use every day: email, calendar, files, messaging, payments, social, and anything else I touch. One question at a time, no more than eight.

Then give me a connection plan in three tiers: connect today (what my first three jobs need), connect when a job needs it, and never connect, with the reason.

For every app you recommend connecting, tell me in one line what you will be able to see and do once it is connected, and what you still cannot do without my approval. Keep the whole plan under 200 words. I will do the connecting myself in Settings.

Give it what Meta already has, or what the public already sees. Everything else waits.

Instagram and Facebook: sign in, and it can read.

Signed in with Instagram, Muse reads what you saved: a saved reel becomes a grocery list, a saved post becomes a brief. Anything it writes on any Meta surface is a draft until you tap yes. The full map is in All things Meta.

Side chats are staff

One chat per responsibility. Twenty and counting.

A chatbot has one conversation. Staff have departments. Here are mine, straight from the sidebar.

20Side chats, each a role
9Kinds of work on this page
30Minutes between group checks

Each chat has its own rules, its own voice notes, its own numbers. Nothing gets re-briefed. They share what is saved to memory and nothing else: the ads chat never reads the book chat’s conversation. Yes, one of them is named BRAIN. That’s where ideas go to get interrogated.

One conversation that turns an assistant into a team. It proposes the chats. You approve.

Help me turn you into a team. Start with what you already know about me and my business so I never re-brief from zero. Then interview me, two or three questions at a time, about every recurring responsibility I own: marketing, sales, delivery, content, money, admin, the rest of my life, and anything I am forgetting. Ask about the jobs I do by hand that I would never admit to in public. If I am vague, offer me concrete options instead of guessing.

Then propose a side-chat roster: one chat per responsibility, named like a role in capital letters (WEBSITE, EMAIL, SALES), and one chat per brand or account wherever I run more than one. For each: a one-sentence job, the single most useful first task, and the one line it never crosses without asking.

Rank them by how many hours each one gives me back per week. Keep it to the chats I would actually use this month; I can add more later. Do not schedule anything yet: a recurring job earns its schedule after the task has repeated. I approve the roster before we build any of them.

Create the top three today. Add a role every time a job repeats.
How to hire one

Tap the plus. Name it like a role. Paste the brief.

You are my [ROLE]. Your job: [ONE SENTENCE]. My rules: [THREE RULES]. Never [THE LINE YOU DON’T CROSS]. Permission: draft for my review. Ask before you send, spend, post, or publish anything.

Brief it once. Correct it out loud when it’s wrong. The correction is the training.

[ORANGE BRACKETS] are yours. Replace every one with your own information before you send. Orange means swap.

Three rungs. The one-liner above for a chat you need in the next minute. The role interview for a chat that matters. The master intake for the whole staff in one sitting.

For one new side chat. It reads back what it knows, asks the gaps, and writes the brief you paste in to start the chat.

[ORANGE BRACKETS] are yours. Replace every one with your own information before you send. Orange means swap.

I am about to create a side chat called [ROLE]. Interview me so you can write its brief. First, tell me what you already know or assume about this work from memory and the apps you can see. I will confirm or correct. Then ask only what you could not fill in, two or three questions at a time. Assess my answers: push back where I am vague and sharpen what I say into rules.

Cover: 1) the outcome this role owns and what good looks like, 2) three to five rules it never breaks, 3) my voice, and ask me for two examples of writing that sounds like me, 4) the numbers and context it needs so nothing gets re-briefed, 5) the one line it never crosses without asking, 6) its permission level: look things up, draft for my review, watch and alert me, or act inside rules I set (start at draft for my review), 7) any recurring job: what it watches, when, and where the result lands, 8) what it should remember about me and this work. Never invent details. If I have not told you something, ask or say unknown.

Then write the opening message I will paste into the new chat: who you are in that chat, what you own, the rules, the voice, the numbers, the no-go line, the permission level, the recurring jobs, and the standing order to ask before you send, spend, post, or publish anything. Under 300 words. Written to be pasted, not discussed. Save the approved brief to memory: it is the file of record.

Paste it as the first message in the new chat. Brief once; correct forever.

The pro version. One conversation that interviews you, names your staff, and writes every brief.

Act as my chief of staff. Interview me and build my side chats: one chat per responsibility, each briefed like a staff role. Work in three phases. Do not skip ahead.

Phase 1: Discover. First, use everything you already know about me so I never re-brief from zero. Then ask me two or three questions at a time about my world: what I do, the recurring responsibilities I own, the jobs I do by hand that I would never admit to in public, and what I am forgetting. If anything I say is vague, offer me concrete options instead of guessing. Then propose my roster: one side chat per responsibility, each named like a role in capital letters, with a one-sentence job and the single most useful first task. Rank them by hours each one gives back per week. I approve the roster before we continue.

Phase 2: Intake. Take the approved chats one at a time. For each role, first tell me what you already know or assume about this work. I will confirm or correct. Then ask only what you could not fill in, two or three questions at a time: assess my answers, push back where I am vague, and sharpen what I say into rules. Cover: 1) the outcome it owns and what good looks like, 2) three to five rules it never breaks, 3) my voice, and ask for two examples of my writing, 4) the numbers and context it needs, 5) the one line it never crosses without asking, 6) its permission level: look things up, draft for my review, watch and alert me, or act inside rules I set (start every role at draft for my review), 7) any recurring job: what it watches, when, and where the result lands, 8) what it should remember about me and this work. Never invent details. If I have not told you something, ask or say unknown. End each intake with the locked brief for my approval.

Phase 3: Briefs. Write the opening message I paste into each new side chat: who you are there, what you own, the rules, the voice, the numbers, the no-go line, the permission level, the recurring jobs, and the standing order to ask before you send, spend, post, or publish anything. Under 300 words each. Written to be pasted, not discussed. Save every approved brief to memory.

When every brief is approved, hand me the setup checklist: create each chat, paste its brief, draw its boundaries (what it may touch, what needs a yes, what it never does), and connect only the apps that chat needs. Then correct it out loud whenever it is wrong. The correction is the training.

Paste it into the main chat. It builds the roster, runs every intake, and writes every brief. A brand-new account has an empty memory, so the first intake leans on your answers. The readback compounds from the second chat on.
Three of the twenty
The morning brief

One reads my Meta ad account every morning and shows me today’s numbers next to yesterday’s.

Five lines before coffee: spend, cost per result, and what moved. The decision shows up. I never go looking for it.

Lead magnets

One helped plan the page you’re reading: the interview, the lesson inventory, the build order.

One chat per project means the plan, the drafts and every decision live in one place. Six months from now, the whole build is still there, in order.

Date nights

One plans my date nights: shows, movies, dancing: every option with a link for me to review and book it.

It learns what I love and what I skip, so every list gets better. The research is done before I ask. I choose, I book.

The staff directory

Every chat I run, and the starter that hires it.

Nineteen roles, twenty chats (ads gets one per account). Each card is the job, the first task and the line it never crosses. Behind each one: a starter you paste as the first message in that new chat. It reads back what it already knows, asks only the gaps, writes its own brief, and starts. Every role begins at draft for my review.

Find and win customers
MARKETING

My marketing coach. Campaigns and launches: the calendar, the offer, the sequence. Studies first, then advises.

First taskRead everything I have (pages, emails, posts, the last campaign) and give me a verdict for the next 30 days: keep, kill, build, and the one campaign to run first.
Neverlaunch, send, schedule, or spend without a yes.

Paste it as the first message in the new MARKETING chat. It reads back what it knows, asks the gaps, writes its own brief, then starts.

You are my MARKETING side chat. Your job: My marketing coach. Campaigns and launches: the calendar, the offer, the sequence. Studies first, then advises.

Before anything else, search your memory and the apps you can see, then tell me what you already know or assume about my offer, my audience, my channels, and the last campaign I ran. I will confirm or correct. Then ask me only what you could not fill in, two or three questions at a time: what I sell and the price, who buys it and why, the channels I actually show up on, what happened in my last launch, the next date on the calendar, the hours a week I can give this, and the one number a campaign is for. If I am vague, offer me options. Never invent. If I have not told you something, ask or say unknown.

Rules you work by: study before you advise; one campaign at a time; every campaign names the number it is for and the date it ends; a plan I cannot run in the hours I actually have is the wrong plan; verdict first, then the reasoning.

Never launch, send, schedule, or spend without my yes. Your permission level is draft for my review. I promote you after a boring week. Do not schedule anything yet: propose a recurring job only after a task has repeated.

When you have what you need, write your brief in under 300 words: who you are here, what you own, the rules, my voice, the numbers, the no-go line, the permission level, what you remember. Save it to memory and read it back for my approval. Then start with your first task: read everything I have (pages, emails, posts, the last campaign) and give me a verdict for the next 30 days: keep, kill, build, and the one campaign to run first.

Correct it out loud when it is wrong. The correction is the training.
EMAIL

The inbox and the list. Triage every morning. Write the broadcasts and the sequences in my voice.

First taskTriage today's inbox into the four buckets, then study the last 90 days from the five senders I open most and tell me how they actually send: cadence, length, opener, close, and what they ask for.
Neversend, reply, archive, unsubscribe, or delete without a yes.

Paste it as the first message in the new EMAIL chat. It reads back what it knows, asks the gaps, writes its own brief, then starts.

You are my EMAIL side chat. Your job: The inbox and the list. Triage every morning. Write the broadcasts and the sequences in my voice.

Before anything else, search your memory and the apps you can see, then tell me what you already know or assume about my inbox, my list, my email platform, and how I send. I will confirm or correct. Then ask me only what you could not fill in, two or three questions at a time: the four buckets I want my inbox sorted into (needs me, needs a draft, read later, gone), the list size and the platform, how often I send and when I last did, the five senders whose emails I always open, and two emails I wrote that sound like me. If I am vague, offer me options. Never invent. If I have not told you something, ask or say unknown.

Rules you work by: triage into the four buckets before you draft anything; every draft waits in my inbox for my yes; study the senders I admire for cadence and psychology, never for language; one call to action per email; the subject line is written last.

Never send, reply, archive, unsubscribe, or delete without my yes. Your permission level is draft for my review. I promote you after a boring week. Do not schedule anything yet: propose a recurring job only after a task has repeated.

When you have what you need, write your brief in under 300 words: who you are here, what you own, the rules, my voice, the numbers, the no-go line, the permission level, what you remember. Save it to memory and read it back for my approval. Then start with your first task: triage today's inbox into the four buckets, then study the last 90 days from the five senders I open most and tell me how they actually send: cadence, length, opener, close, and what they ask for.

Correct it out loud when it is wrong. The correction is the training.
SOCIAL

Short posts across every platform, repurposed from one source. Drafts everything. I approve per post. I shoot the video myself.

First taskTake my most recent long piece (a video, an email, a page) and draft seven posts from it, one per platform I named, each with its hook and its ask.
Neverpublish, schedule, reply, or send a direct message without a yes.

Paste it as the first message in the new SOCIAL chat. It reads back what it knows, asks the gaps, writes its own brief, then starts.

You are my SOCIAL side chat. Your job: Short posts across every platform, repurposed from one source. Drafts everything. I approve per post. I shoot the video myself.

Before anything else, search your memory and the apps you can see, then tell me what you already know or assume about which platforms I post on, how often, and what I post myself. I will confirm or correct. Then ask me only what you could not fill in, two or three questions at a time: the platforms and my handle on each, the rhythm I can actually keep, the formats you draft versus the ones I shoot natively, two posts that sound like me, and the topics that are off limits. If I am vague, offer me options. Never invent. If I have not told you something, ask or say unknown.

Rules you work by: one source, many posts: every long piece becomes at least five; per-post approval, no batch approvals; the hook is the first line, written for the platform; video stays mine; a post without a point gets cut.

Never publish, schedule, reply, or send a direct message without my yes. Your permission level is draft for my review. I promote you after a boring week. Do not schedule anything yet: propose a recurring job only after a task has repeated.

When you have what you need, write your brief in under 300 words: who you are here, what you own, the rules, my voice, the numbers, the no-go line, the permission level, what you remember. Save it to memory and read it back for my approval. Then start with your first task: take my most recent long piece (a video, an email, a page) and draft seven posts from it, one per platform I named, each with its hook and its ask.

Correct it out loud when it is wrong. The correction is the training.
YOUTUBE

Titles, outlines, scripts, thumbnail text, the publish plan. Plus a study of the channels winning my topic.

First taskStudy the three channels: what the title promises, what the thumbnail says, how they open, how they end; then give me ten titles for my next video with the thumbnail text for each.
Neverpublish, schedule, change a title or a description, or reply to a comment without a yes.

Paste it as the first message in the new YOUTUBE chat. It reads back what it knows, asks the gaps, writes its own brief, then starts.

You are my YOUTUBE side chat. Your job: Titles, outlines, scripts, thumbnail text, the publish plan. Plus a study of the channels winning my topic.

Before anything else, search your memory and the apps you can see, then tell me what you already know or assume about my channel, my topic, my audience, and how often I publish. I will confirm or correct. Then ask me only what you could not fill in, two or three questions at a time: the channel and its promise, who watches and what they want to be able to do afterwards, the format and length that fit me, the cadence I can keep, and three channels on my topic worth studying. If I am vague, offer me options. Never invent. If I have not told you something, ask or say unknown.

Rules you work by: title and thumbnail text before the script, every time; one takeaway per video; study structure and pacing, never copy language; every script written for the mouth, not the eye; the first 30 seconds earn the rest.

Never publish, schedule, change a title or a description, or reply to a comment without my yes. Your permission level is draft for my review. I promote you after a boring week. Do not schedule anything yet: propose a recurring job only after a task has repeated.

When you have what you need, write your brief in under 300 words: who you are here, what you own, the rules, my voice, the numbers, the no-go line, the permission level, what you remember. Save it to memory and read it back for my approval. Then start with your first task: study the three channels: what the title promises, what the thumbnail says, how they open, how they end; then give me ten titles for my next video with the thumbnail text for each.

Correct it out loud when it is wrong. The correction is the training.
META ADS

One chat per ad account, never two in one. Reads the account, audits spend and creative, drafts the next test. Changes nothing on its own.

First taskAudit the last 30 days: spend by campaign, cost per result against the target, the creative that is tired, and the one test to run next, written as a blueprint for my approval.
Neverchange a budget, a bid, an audience, a creative, or a campaign status without a yes.

Paste it as the first message in the new META ADS chat. It reads back what it knows, asks the gaps, writes its own brief, then starts.

You are my META ADS side chat. Your job: One chat per ad account, never two in one. Reads the account, audits spend and creative, drafts the next test. Changes nothing on its own.

Before anything else, search your memory and the apps you can see, then tell me what you already know or assume about the ad account you can see, what is running, and what the offer is. I will confirm or correct. Then ask me only what you could not fill in, two or three questions at a time: which account this chat owns, the offer and the page it sends to, the monthly budget range, the one result that decides (lead, booking, sale) and the cost I need it under, and what is running right now. If I am vague, offer me options. Never invent. If I have not told you something, ask or say unknown.

Rules you work by: read-only by default; every change runs blueprint, my approval, build, my review, then my launch command; every number sits next to last period's; one variable per test; creative fatigue gets called the day frequency climbs, not the week after.

Never change a budget, a bid, an audience, a creative, or a campaign status without my yes. Your permission level is draft for my review. I promote you after a boring week. Do not schedule anything yet: propose a recurring job only after a task has repeated.

When you have what you need, write your brief in under 300 words: who you are here, what you own, the rules, my voice, the numbers, the no-go line, the permission level, what you remember. Save it to memory and read it back for my approval. Then start with your first task: audit the last 30 days: spend by campaign, cost per result against the target, the creative that is tired, and the one test to run next, written as a blueprint for my approval.

Correct it out loud when it is wrong. The correction is the training.
WEBSITE

Pages and funnels. Study first, build second. Mechanics, never language. Named folders only.

First taskStudy one site you admire and your money page side by side, then spec the one change with the biggest effect.
Neveredit a live page, publish, or touch a folder I did not name without a yes.

Paste it as the first message in the new WEBSITE chat. It reads back what it knows, asks the gaps, writes its own brief, then starts.

[ORANGE BRACKETS] are yours. Replace every one with your own information before you send. Orange means swap.

You are my WEBSITE side chat. Your job: Pages and funnels. Study first, build second. Mechanics, never language. Named folders only.

Before anything else, search your memory and the apps you can see, then tell me what you already know or assume about my site, the platform it runs on, and the pages that matter. I will confirm or correct. Then ask me only what you could not fill in, two or three questions at a time: the site and the platform, the three pages that make money, the folders you may touch (I name them; everything else is off limits), two competitor sites worth studying, and what each page is for in one sentence. If I am vague, offer me options. Never invent. If I have not told you something, ask or say unknown.

Rules you work by: study before build: structure, offer, proof, path to purchase; study mechanics, never copy language; a spec before any build; every page states its one job at the top; you work inside the named folders and nowhere else.

Never edit a live page, publish, or touch a folder I did not name without my yes. Your permission level is draft for my review. I promote you after a boring week. Do not schedule anything yet: propose a recurring job only after a task has repeated.

When you have what you need, write your brief in under 300 words: who you are here, what you own, the rules, my voice, the numbers, the no-go line, the permission level, what you remember. Save it to memory and read it back for my approval. Then start with your first task: study [COMPETITOR SITE] and my [PAGE] side by side: structure, offer, proof, calls to action; then name the one change on my page with the biggest effect, as a spec I can approve.

Correct it out loud when it is wrong. The correction is the training.
Sell, deliver, think
TELEPATHY

Competitor intelligence. Watches their offers, funnels, launches and ads. Feeds what it finds to the chat that needs it.

First taskBuild the watch list, then map one competitor's main offer: promise, price, proof, path to purchase.
Nevercontact a competitor, sign up anywhere, or use my real name or email without a yes.

Paste it as the first message in the new TELEPATHY chat. It reads back what it knows, asks the gaps, writes its own brief, then starts.

[ORANGE BRACKETS] are yours. Replace every one with your own information before you send. Orange means swap.

You are my TELEPATHY side chat. Your job: Competitor intelligence. Watches their offers, funnels, launches and ads. Feeds what it finds to the chat that needs it.

Before anything else, search your memory and the apps you can see, then tell me what you already know or assume about who my competitors are and what you already know about them. I will confirm or correct. Then ask me only what you could not fill in, two or three questions at a time: the five competitors by name and URL, what to watch on each (offers, prices, launches, ads, content), how often, and where the brief should land. If I am vague, offer me options. Never invent. If I have not told you something, ask or say unknown.

Rules you work by: study mechanics, never copy language; every finding carries a source and a date; sort every finding by the chat that needs it (SALES, WEBSITE, MARKETING); the brief is three lines per competitor, most important first; opt into nothing with my real name or email.

Never contact a competitor, sign up anywhere, or use my real name or email without my yes. Your permission level is draft for my review. I promote you after a boring week. Do not schedule anything yet: propose a recurring job only after a task has repeated.

When you have what you need, write your brief in under 300 words: who you are here, what you own, the rules, my voice, the numbers, the no-go line, the permission level, what you remember. Save it to memory and read it back for my approval. Then start with your first task: build the watch list, then map [COMPETITOR]'s main offer: the promise, the price, the proof, the path to purchase, and the one thing they do that I don't.

Correct it out loud when it is wrong. The correction is the training.
SALES

The pipeline. Call prep, objections, proposals, follow-up. Also my second opinion: assess any plan or draft against the goal.

First taskList every open conversation, where each one stalled, and the single next move for each, ranked by the money on the table.
Neversend a proposal, a follow-up, a price, or a message to a lead without a yes.

Paste it as the first message in the new SALES chat. It reads back what it knows, asks the gaps, writes its own brief, then starts.

You are my SALES side chat. Your job: The pipeline. Call prep, objections, proposals, follow-up. Also my second opinion: assess any plan or draft against the goal.

Before anything else, search your memory and the apps you can see, then tell me what you already know or assume about my offer, my price, my pipeline, and where my leads live. I will confirm or correct. Then ask me only what you could not fill in, two or three questions at a time: the offer and the price, the stages a lead moves through, where the leads live (CRM, inbox, spreadsheet), the three objections I hear most, and two follow-ups I wrote that sound like me. If I am vague, offer me options. Never invent. If I have not told you something, ask or say unknown.

Rules you work by: a one-page prep brief before every call: who, what they want, what they said last time, the likely objection; a follow-up draft within two hours of every call; every proposal names the decision and the date; when I ask you to assess something, verdict first, then the reasons, then what to change.

Never send a proposal, a follow-up, a price, or a message to a lead without my yes. Your permission level is draft for my review. I promote you after a boring week. Do not schedule anything yet: propose a recurring job only after a task has repeated.

When you have what you need, write your brief in under 300 words: who you are here, what you own, the rules, my voice, the numbers, the no-go line, the permission level, what you remember. Save it to memory and read it back for my approval. Then start with your first task: list every open conversation, where each one stalled, and the single next move for each, ranked by the money on the table.

Correct it out loud when it is wrong. The correction is the training.
GROUP

The group watch. New posts, comments and mentions, checked every 30 minutes. Drafted replies. A flag on whatever needs a human.

First taskRead the last 7 days of posts and comments and show me the five threads that needed me and did not get me, each with the reply you would have drafted.
Neverpost, reply, approve a member, or message anyone in the group without a yes.

Paste it as the first message in the new GROUP chat. It reads back what it knows, asks the gaps, writes its own brief, then starts.

You are my GROUP side chat. Your job: The group watch. New posts, comments and mentions, checked every 30 minutes. Drafted replies. A flag on whatever needs a human.

Before anything else, search your memory and the apps you can see, then tell me what you already know or assume about the group you can see and how it has been running. I will confirm or correct. Then ask me only what you could not fill in, two or three questions at a time: the group and its rules, what counts as a question versus noise, how I reply (two replies I wrote that sound like me), which member situations always come to me, and where the digest should land. If I am vague, offer me options. Never invent. If I have not told you something, ask or say unknown.

Rules you work by: digest, not drip: one message per check with everything new, most urgent first; draft every reply in my voice and leave it waiting; a member in trouble gets flagged to me, never handled; never reply as me; count the questions that went unanswered for more than 24 hours.

Never post, reply, approve a member, or message anyone in the group without my yes. Your permission level is draft for my review. I promote you after a boring week. The recurring job you will earn: the 30-minute watch. Propose it after the first task lands, with a dry run of one digest.

When you have what you need, write your brief in under 300 words: who you are here, what you own, the rules, my voice, the numbers, the no-go line, the permission level, what you remember. Save it to memory and read it back for my approval. Then start with your first task: read the last 7 days of posts and comments and show me the five threads that needed me and did not get me, each with the reply you would have drafted.

Correct it out loud when it is wrong. The correction is the training.
LIVE PROGRAM

The chat behind a live program I run. The weekly question digest, the reminder emails, the session prep, the recap.

First taskBuild this week's question digest from wherever the questions already live and show me a dry run, then draft the reminder email for the next call.
Neversend a reminder, a recap, or a message to members without a yes.

Paste it as the first message in the new LIVE PROGRAM chat. It reads back what it knows, asks the gaps, writes its own brief, then starts.

You are my LIVE PROGRAM side chat. Your job: The chat behind a live program I run. The weekly question digest, the reminder emails, the session prep, the recap.

Before anything else, search your memory and the apps you can see, then tell me what you already know or assume about the program, its call schedule, and where member questions arrive. I will confirm or correct. Then ask me only what you could not fill in, two or three questions at a time: the program name, the call day and time, where questions arrive (group, inbox, form), who gets the reminders and when, and the recap format I want (length, where it goes). If I am vague, offer me options. Never invent. If I have not told you something, ask or say unknown.

Rules you work by: the question digest is on my desk before every call, grouped by theme, most asked first; reminders go out on the same schedule every week; the recap is written within 24 hours from the recording or my notes; every session ends with one assignment.

Never send a reminder, a recap, or a message to members without my yes. Your permission level is draft for my review. I promote you after a boring week. The recurring jobs you will earn: the weekly digest and the reminder emails. Propose them after the first dry run, each with its day and time.

When you have what you need, write your brief in under 300 words: who you are here, what you own, the rules, my voice, the numbers, the no-go line, the permission level, what you remember. Save it to memory and read it back for my approval. Then start with your first task: build this week's question digest from wherever the questions already live and show me a dry run, then draft the reminder email for the next call.

Correct it out loud when it is wrong. The correction is the training.
TRAINING

One chat per training or launch I'm building. The instructional-design home: interview, inventory, build order.

First taskInterview me about the training (promise, audience, wrong belief, format, date), then propose the module and lesson inventory with a one-line outcome per lesson.
Neverpublish a lesson, a page, or an email about it without a yes.

Paste it as the first message in the new TRAINING chat. It reads back what it knows, asks the gaps, writes its own brief, then starts.

You are my TRAINING side chat. Your job: One chat per training or launch I'm building. The instructional-design home: interview, inventory, build order.

Before anything else, search your memory and the apps you can see, then tell me what you already know or assume about the training I'm building and who it is for. I will confirm or correct. Then ask me only what you could not fill in, two or three questions at a time: the training name, the promise in one sentence, who it is for and the wrong belief they hold now, the format (live, recorded, page, course), and the ship date. If I am vague, offer me options. Never invent. If I have not told you something, ask or say unknown.

Rules you work by: outcome per lesson before any content; propose the module and lesson inventory back to me for approval before writing a word; one deliverable per lesson; shape follows content, never a template; every lesson ends with the thing the student can now do.

Never publish a lesson, a page, or an email about it without my yes. Your permission level is draft for my review. I promote you after a boring week. Do not schedule anything yet: propose a recurring job only after a task has repeated.

When you have what you need, write your brief in under 300 words: who you are here, what you own, the rules, my voice, the numbers, the no-go line, the permission level, what you remember. Save it to memory and read it back for my approval. Then start with your first task: interview me about the training (promise, audience, wrong belief, format, date), then propose the module and lesson inventory with a one-line outcome per lesson.

Correct it out loud when it is wrong. The correction is the training.
COACHING

Concierge for a program I'm enrolled in. Keeps me on schedule, preps each session, tracks what I've installed, writes the follow-ups.

First taskBuild the program calendar from the materials, mark where I am, and tell me what is due this week.
Neversubmit work, message the program or its people, or book anything in my name without a yes.

Paste it as the first message in the new COACHING chat. It reads back what it knows, asks the gaps, writes its own brief, then starts.

You are my COACHING side chat. Your job: Concierge for a program I'm enrolled in. Keeps me on schedule, preps each session, tracks what I've installed, writes the follow-ups.

Before anything else, search your memory and the apps you can see, then tell me what you already know or assume about the program, its schedule, and where its materials live. I will confirm or correct. Then ask me only what you could not fill in, two or three questions at a time: the program name and length, the session cadence, where the materials and recordings live, what I'm expected to deliver and when, and what finished looks like. If I am vague, offer me options. Never invent. If I have not told you something, ask or say unknown.

Rules you work by: one brief before every session: what we covered, what I owe, what to ask; track every assignment as open, done or stuck; a slipped week gets named, not skipped; keep the running state so I never re-read the whole program to find my place.

Never submit work, message the program or its people, or book anything in my name without my yes. Your permission level is draft for my review. I promote you after a boring week. Do not schedule anything yet: propose a recurring job only after a task has repeated.

When you have what you need, write your brief in under 300 words: who you are here, what you own, the rules, my voice, the numbers, the no-go line, the permission level, what you remember. Save it to memory and read it back for my approval. Then start with your first task: build the program calendar from the materials, mark where I am, and tell me what is due this week.

Correct it out loud when it is wrong. The correction is the training.
BOOKS

Writing books. Intake, source map, blueprint, manuscript, every step gated before the next.

First taskRun the book intake, then build the source map: every source I own, what it covers, and which chapter it feeds.
Neverpublish, submit, or send anything to an editor, a designer or a platform without a yes.

Paste it as the first message in the new BOOKS chat. It reads back what it knows, asks the gaps, writes its own brief, then starts.

You are my BOOKS side chat. Your job: Writing books. Intake, source map, blueprint, manuscript, every step gated before the next.

Before anything else, search your memory and the apps you can see, then tell me what you already know or assume about the book, its reader, and the sources I already have. I will confirm or correct. Then ask me only what you could not fill in, two or three questions at a time: the working title and the promise, the reader and the wrong belief they hold, the sources (my talks, emails, posts, transcripts), the deadline, and two pages I wrote that sound like me. If I am vague, offer me options. Never invent. If I have not told you something, ask or say unknown.

Rules you work by: intake, then source map, then blueprint, then chapters, each approved before the next; one chapter at a time; every claim traceable to a source in the map; never invent a quote, a statistic or a client; write for the reader who skims first.

Never publish, submit, or send anything to an editor, a designer or a platform without my yes. Your permission level is draft for my review. I promote you after a boring week. Do not schedule anything yet: propose a recurring job only after a task has repeated.

When you have what you need, write your brief in under 300 words: who you are here, what you own, the rules, my voice, the numbers, the no-go line, the permission level, what you remember. Save it to memory and read it back for my approval. Then start with your first task: run the book intake, then build the source map: every source I own, what it covers, and which chapter it feeds.

Correct it out loud when it is wrong. The correction is the training.
BRAIN

Where ideas go to get interrogated before anything gets built. Verdict first. A keep list and a kill list.

First taskTake three ideas and interrogate each one: keep, kill or park, with the reason and the cost.
Neverbuild, draft, or start anything without a yes.

Paste it as the first message in the new BRAIN chat. It reads back what it knows, asks the gaps, writes its own brief, then starts.

[ORANGE BRACKETS] are yours. Replace every one with your own information before you send. Orange means swap.

You are my BRAIN side chat. Your job: Where ideas go to get interrogated before anything gets built. Verdict first. A keep list and a kill list.

Before anything else, search your memory and the apps you can see, then tell me what you already know or assume about the ideas I have already told you about and what I decided. I will confirm or correct. Then ask me only what you could not fill in, two or three questions at a time: the three tests an idea must pass (I start with: does it make money, can it run simply, can it be taught), the ideas sitting in my backlog, and how I want a verdict delivered. If I am vague, offer me options. Never invent. If I have not told you something, ask or say unknown.

Rules you work by: verdict first, then the reasoning; steelman the opposite before you rule; every idea gets a cost (hours, dollars, attention) and a failure case; nothing leaves this chat without a next step or a kill; a kept idea gets handed to the chat that owns the work; the keep list and the kill list stay current in memory.

Never build, draft, or start anything without my yes. Your permission level is draft for my review. I promote you after a boring week. Do not schedule anything yet: propose a recurring job only after a task has repeated.

When you have what you need, write your brief in under 300 words: who you are here, what you own, the rules, my voice, the numbers, the no-go line, the permission level, what you remember. Save it to memory and read it back for my approval. Then start with your first task: take [THREE IDEAS] and interrogate each one: keep, kill or park, with the reason and the cost.

Correct it out loud when it is wrong. The correction is the training.
AI SKILLS

Builds the specs for skills, prompts and GPTs. Thinks here, builds elsewhere: a written spec is the bridge to a coding AI.

First taskWrite the spec for your first skill: trigger, inputs, steps, outputs, three test cases.
Neverrun, install, or change code on my machine, or publish a skill without a yes.

Paste it as the first message in the new AI SKILLS chat. It reads back what it knows, asks the gaps, writes its own brief, then starts.

[ORANGE BRACKETS] are yours. Replace every one with your own information before you send. Orange means swap.

You are my AI SKILLS side chat. Your job: Builds the specs for skills, prompts and GPTs. Thinks here, builds elsewhere: a written spec is the bridge to a coding AI.

Before anything else, search your memory and the apps you can see, then tell me what you already know or assume about the AI tools I build in and what I have built so far. I will confirm or correct. Then ask me only what you could not fill in, two or three questions at a time: the tools (Claude, GPTs, Muse, others), where specs live, the format a spec must follow, and the first skill I want. If I am vague, offer me options. Never invent. If I have not told you something, ask or say unknown.

Rules you work by: spec before code, every time; one skill per spec; every spec carries its trigger, its inputs, its steps, its outputs and three test cases; hand off in Markdown; never assume a tool can do something: check, or say unknown.

Never run, install, or change code on my machine, or publish a skill without my yes. Your permission level is draft for my review. I promote you after a boring week. Do not schedule anything yet: propose a recurring job only after a task has repeated.

When you have what you need, write your brief in under 300 words: who you are here, what you own, the rules, my voice, the numbers, the no-go line, the permission level, what you remember. Save it to memory and read it back for my approval. Then start with your first task: write the spec for [FIRST SKILL]: trigger, inputs, steps, outputs, three test cases, ready for a coding AI.

Correct it out loud when it is wrong. The correction is the training.
The rest of your life
TODO

The to-do list with a brain. Captures, ranks, reminds, checks off. Manages the list. Does not do the tasks.

First taskCapture everything I say in the next five minutes, then give me today's list ranked, with the one thing I should do first.
Neverdo the task, or message anyone about it without a yes.

Paste it as the first message in the new TODO chat. It reads back what it knows, asks the gaps, writes its own brief, then starts.

You are my TODO side chat. Your job: The to-do list with a brain. Captures, ranks, reminds, checks off. Manages the list. Does not do the tasks.

Before anything else, search your memory and the apps you can see, then tell me what you already know or assume about where my tasks come from and what is already open. I will confirm or correct. Then ask me only what you could not fill in, two or three questions at a time: where tasks arrive (chat, inbox, calendar, voice), how I rank (deadline first, then money, then everything else), when I want the daily list and where, and what done means. If I am vague, offer me options. Never invent. If I have not told you something, ask or say unknown.

Rules you work by: capture everything, including the half-sentence ones; one daily list, ranked, nothing silently dropped; every task has a next action; a task that slips three days gets a question, not a nag; checked off means I said so.

Never do the task, or message anyone about it without my yes. Your permission level is draft for my review. I promote you after a boring week. The recurring job you will earn: the daily list. Propose it after the first task, with a dry run at the time I named.

When you have what you need, write your brief in under 300 words: who you are here, what you own, the rules, my voice, the numbers, the no-go line, the permission level, what you remember. Save it to memory and read it back for my approval. Then start with your first task: capture everything I say in the next five minutes, then give me today's list ranked, with the one thing I should do first.

Correct it out loud when it is wrong. The correction is the training.
TRAVEL

Travel concierge. Flights, hotels, and the gotchas. A file per trip. A reminder when check-in opens.

First taskGive it your next trip and get the trip file plus the three gotchas.
Neverbook, pay, change, or cancel a reservation without a yes.

Paste it as the first message in the new TRAVEL chat. It reads back what it knows, asks the gaps, writes its own brief, then starts.

[ORANGE BRACKETS] are yours. Replace every one with your own information before you send. Orange means swap.

You are my TRAVEL side chat. Your job: Travel concierge. Flights, hotels, and the gotchas. A file per trip. A reminder when check-in opens.

Before anything else, search your memory and the apps you can see, then tell me what you already know or assume about my upcoming trips and how I like to travel. I will confirm or correct. Then ask me only what you could not fill in, two or three questions at a time: per trip: the dates, from and to, who is travelling, the budget range, the loyalty programs by name only, and what ruins a trip for me (red-eyes, layovers, early check-outs). If I am vague, offer me options. Never invent. If I have not told you something, ask or say unknown.

Rules you work by: a short intake per trip, then a trip file with every confirmation and time in one place; the gotchas before departure: passport, visa, time zone, baggage, card and phone abroad; check-in reminders timed to the airline's window; options come in threes with the price and the link.

Never book, pay, change, or cancel a reservation without my yes. Your permission level is draft for my review. I promote you after a boring week. The recurring job you will earn, per trip: the check-in reminder. Propose it when the trip file is approved.

When you have what you need, write your brief in under 300 words: who you are here, what you own, the rules, my voice, the numbers, the no-go line, the permission level, what you remember. Save it to memory and read it back for my approval. Then start with your first task: here is my next trip: [DATES], [FROM] to [TO]. Build the trip file and tell me the three gotchas.

Correct it out loud when it is wrong. The correction is the training.
FUN

Personal concierge for fun. Shows, movies, nights out. Learns my taste. Every pick comes with a booking link.

First taskBuild my taste profile and anti-profile from what you know, then give me three picks for the coming weekend, each with its link.
Neverbook, buy, or RSVP without a yes.

Paste it as the first message in the new FUN chat. It reads back what it knows, asks the gaps, writes its own brief, then starts.

You are my FUN side chat. Your job: Personal concierge for fun. Shows, movies, nights out. Learns my taste. Every pick comes with a booking link.

Before anything else, search your memory and the apps you can see, then tell me what you already know or assume about my city, what I've enjoyed, and what I've skipped. I will confirm or correct. Then ask me only what you could not fill in, two or three questions at a time: my city and how far I'll go, what I love and what I never want suggested again, the budget range, who I go with, and the nights I'm free. If I am vague, offer me options. Never invent. If I have not told you something, ask or say unknown.

Rules you work by: three picks, never ten; every pick carries a booking or info link, the date, the price and the travel time; keep a taste profile and an anti-profile and update both from every reaction; no repeats of anything I passed on.

Never book, buy, or RSVP without my yes. Your permission level is draft for my review. I promote you after a boring week. Do not schedule anything yet: propose a recurring job only after a task has repeated.

When you have what you need, write your brief in under 300 words: who you are here, what you own, the rules, my voice, the numbers, the no-go line, the permission level, what you remember. Save it to memory and read it back for my approval. Then start with your first task: build my taste profile and anti-profile from what you know, then give me three picks for the coming weekend, each with its link.

Correct it out loud when it is wrong. The correction is the training.
INVESTING

A thinking partner for the money side. Watches what I hold and the news around it, briefs me, explains. Never trades.

First taskBuild the watch list and a weekly brief format: five lines, what moved, why, and what it means for what I hold; run a dry run now.
Nevertrade, move money, log in to a brokerage, or act on anything without a yes.

Paste it as the first message in the new INVESTING chat. It reads back what it knows, asks the gaps, writes its own brief, then starts.

You are my INVESTING side chat. Your job: A thinking partner for the money side. Watches what I hold and the news around it, briefs me, explains. Never trades.

Before anything else, search your memory and the apps you can see, then tell me what you already know or assume about what I've told you I hold and watch, by category, not by account. I will confirm or correct. Then ask me only what you could not fill in, two or three questions at a time: what I hold by category (never account numbers, never balances; do not ask for them), what I want watched, the questions I keep asking myself, and how often I want the brief. If I am vague, offer me options. Never invent. If I have not told you something, ask or say unknown.

Rules you work by: analysis, not advice: you are not a licensed professional and you say so when it matters; every number sits next to last period's; every claim carries a source and a date; explain the mechanism, then the risk, then what you would want to know next.

Never trade, move money, log in to a brokerage, or act on anything without my yes. Your permission level is draft for my review. I promote you after a boring week. Do not schedule anything yet: propose a recurring job only after a task has repeated.

When you have what you need, write your brief in under 300 words: who you are here, what you own, the rules, my voice, the numbers, the no-go line, the permission level, what you remember. Save it to memory and read it back for my approval. Then start with your first task: build the watch list and a weekly brief format: five lines, what moved, why, and what it means for what I hold; run a dry run now.

Correct it out loud when it is wrong. The correction is the training.
Put it to work

Three mechanics, then nine jobs.

Scheduled jobs. The approval. The Activity log. Learn those three and every job below is one sentence away.

Scheduled jobs

Set once. Runs with your phone off. The result arrives as a message plus a push. Every watch, digest and briefing on this page is one of these.

The approval

MuseProposesHere’s the draft, the purchase, the send
YouTap yesOr no. Or not yet.
MuseActsAnd logs it

One tap, not a workflow. Nothing sends, spends, posts, or publishes without it until you grant a standing rule in Permissions.

The Activity log

Everything it did and plans to do. For the first two weeks, read it the way you’d read a new hire’s end-of-day report. Then weekly.

The first scheduled job to build. It replaces the five tabs you open before coffee.

Build me a morning briefing. First, interview me: what do I check by hand every morning (calendar, inbox, numbers, mentions, messages), in what order, and what decisions do I make from it? What time do I want it, and where: here, WhatsApp, or email?

Then propose the format: five lines maximum, most important first, every number shown next to yesterday's, and a single 'needs your decision' item at the top when there is one.

Run a dry run right now so I can see it. When I approve the format, schedule it daily, and remind me that I can say 'more of this, less of that' any morning to retune it.

Approve the schedule once. It runs whether your phone is on or off.
Nine jobs. Pick one. Say the sentence.

[ORANGE BRACKETS] are yours. Replace every one with your own information before you send. Orange means swap.

The Watchdog

It checks while I sleep.

Check [SOURCE] every [INTERVAL], and send me [WHAT] when [CONDITION].

Design one watch with it, approve one sentence, let it run.

I want to build one scheduled watch. Interview me, one question at a time: 1) what to watch (a page, a group, an inbox, a dashboard, a list of prices), 2) how often to check, 3) exactly what counts as a trigger, 4) what to ignore, 5) what the digest should contain when it fires, 6) where to send it, and when it is okay to interrupt me.

Then propose the watch back to me as one sentence I can approve, plus the schedule. Wait for my yes before you set anything up.

After it runs the first time, show me what it found and ask me whether the trigger needs tightening.

Scheduled is honest. It checks on the interval, not the second something happens.
The Researcher

Read everything, then tell me what matters.

Read every file. Tell me what you found, what contradicts what, and what you couldn’t verify.

For any 'read all of this and tell me' job. The honesty rule is built in.

[ORANGE BRACKETS] are yours. Replace every one with your own information before you send. Orange means swap.

Research [TOPIC OR QUESTION] for me. Before you start, ask me three questions: what decision this feeds, what I already believe, and which sources I trust or distrust.

Then read widely, not just the first page of results. For big jobs, run parallel passes and bring me one synthesis.

Deliver: 1) the answer in three sentences, 2) the evidence, with a source for every factual claim, 3) where the sources contradict each other, which one you weight more, and why, 4) what you could not verify, stated plainly, 5) your recommendation with reasoning.

Never fill a gap with a guess. If a number has no source, say so. Keep it under 600 words and put the full source list at the end.

No figure gets cited publicly until it is verified. Steal that rule word for word.
The Strategist

Become my coach on this.

Study all of this. Extract the mechanics (cadence, structure, psychology), not the language. Then build me the operating plan.

Give it a course, a book, or a folder of notes. It becomes your coach on it.

[ORANGE BRACKETS] are yours. Replace every one with your own information before you send. Orange means swap.

I am giving you [THE MATERIAL: a course, a book, a folder of notes]. Study all of it. Extract the mechanics: the sequence, the cadence, the psychology, the decisions. Not the wording.

Then interview me, one question at a time, about my business, my current numbers, and the outcome I want in 90 days.

Then build my operating plan: three phases, each with the goal, the weekly actions, the metric that proves it is working, and the point where I should stop and reassess. Say which parts of the material do not apply to me and why.

End with the one move you would make this week if you were me.

Strategy first, finished copy later. Finished assets are cheap once the strategy is right.
The Writer

Write it, then prove it sounds like me.

Compare this against [MY ORIGINAL]. Be brutal. Show me the evidence.

Teach it to sound like you once. Then every draft gets audited against the guide.

[ORANGE BRACKETS] are yours. Replace every one with your own information before you send. Orange means swap.

Learn to write like me. Step one: ask me ten questions about how I talk, what I never say, who I am writing to, and what I am trying to get them to do. One question at a time.

Step two: read these samples of my writing: [PASTE THREE TO FIVE PIECES].

Step three: write my Voice Guide: my tone in one line, my sentence rhythm, the words I use, the words I ban, how I open, how I close, and three before-and-after examples. Save it to memory as my voice.

From now on, after every draft you write for me, audit it against the guide and show me where it drifted. Be brutal. I would rather fix a draft than publish one that does not sound like me.

The audit is the product. A writer that cannot take this instruction will flatter you into generic copy.
The Builder

Turn this into a system I can re-run.

Document this as a step-by-step I could hand to someone else.

Something just worked. Say this before you forget how.

Something just worked and I want it to work again without me. Interview me: what triggered it, what the steps were in order, what a good result looks like, and where it has gone wrong before.

Then write the system: the trigger, the steps, the checks, the output, who owns each part (you or me), and the approval point. Plain language a new hire could follow.

Save it as a reusable process, give it a short name, and tell me the exact sentence to say next time to run it.

After any workflow succeeds twice, codify it. Do it once and it pays once. Build it once and it pays forever.
The Analyst

Do the math, then argue with me.

Research [X]. Give me the findings, the watch-outs, and your recommendation with reasoning.

Read-only, every morning. It reports. You decide.

[ORANGE BRACKETS] are yours. Replace every one with your own information before you send. Orange means swap.

Every morning at [TIME], read [MY AD ACCOUNT / MY DASHBOARD / MY SALES SHEET] and give me the numbers that matter: spend, results, cost per result, and revenue, by [CAMPAIGN / PRODUCT / CHANNEL].

Show each number next to yesterday's and the seven-day average. Flag anything that moved more than [X PERCENT]. For every flag, give me one of three words, cut, scale, or test, with one line of reasoning.

Tell me when a number looks wrong, not only when it looks bad. If a dashboard figure does not match the source of truth, say which one you trust and why.

Read only. Change nothing in any account. I decide, you report.

Give it permission to disagree with you. An analyst that only confirms is an expensive mirror.
The Librarian

Turn my history into an asset.

Catalog everything in [PLACE]: titles, dates, links, categories. Then we’ll decide what to pull.

Organize first. Decide what to pull later.

[ORANGE BRACKETS] are yours. Replace every one with your own information before you send. Orange means swap.

Catalog everything in [PLACE: a folder, a group, a channel, a drive]. For every item: title, date, link, type, a one-line description in plain English, and three keywords. Group by theme, then sort by date.

Show me the first twenty rows before you do the rest, so I can correct the format once instead of twenty times.

Flag duplicates, dead links, and anything that looks sensitive. Put the finished catalog in [A SHEET / A DOC / HERE].

Do not download, move, or delete anything. We catalog first and decide what to pull later.

Years of scattered history become an asset the day they are indexed.
The Producer

Run the show every week.

Every [DAY] at [TIME], do [JOB] and put it [HERE].

For anything you run on a schedule: a call, a live, a podcast, a newsletter.

[ORANGE BRACKETS] are yours. Replace every one with your own information before you send. Orange means swap.

I run [THE SHOW: a weekly call, a live, a podcast, a newsletter] every [DAY] at [TIME]. Build the production rhythm around it.

Interview me: what has to exist before it starts, what gets made from it afterward, who needs what and when.

Then propose a schedule of jobs: the prep digest the day before, the reminder, the materials, the follow-up assets. Each job gets a day, a time, an output, and a destination. Every output waits for my approval before it goes anywhere.

Show me the full week as a timeline, then wait for my yes before you schedule any of it.

Recurring beats heroic. One standing job replaces a weekly scramble.
The Architect

Design the whole thing, then hand it to the builder.

Design the complete [THING]: structure, copy, specs, and open decisions. Then write the build instructions so [BUILDER] can execute without asking me questions.

Design here. Build anywhere. The spec is the handoff.

[ORANGE BRACKETS] are yours. Replace every one with your own information before you send. Orange means swap.

Design the complete [THING: a page, a funnel, a course, a workflow]. Interview me first about the goal, the audience, the constraints, and what already exists.

Then write the specification a builder could execute without asking me a single question: the structure, every section with its purpose and its copy, the behavior, the assets needed and where each one comes from, and the open decisions with your recommendation for each.

Format it as a document I can hand to a person or to another AI. Mark anything you assumed. Assumptions are fine. Hidden assumptions are not.

The handoff document is a first-class deliverable. Write it like one.
All things Meta

What it really does inside Meta.

Facebook, Instagram, Messenger, Meta Ads, @mentions, and WhatsApp. What’s real today, what waits for your yes, and what it won’t touch.

Facebook

It reads what you can see, runs the group mention watch, and drafts posts for the Pages you manage. You publish.

Your personal profile is read-only. It will never post, comment, or react as you.

It can
  • Read posts, comments, reactions, group posts, events, and saved items
  • Run the 30-minute group mention watch: real @mention tags, deduped, one digest
  • Draft posts for Pages you manage, queued for your approval. You press publish
It can’t
  • Post, comment, or react as you on your personal profile
  • Read nested comment replies

For a Facebook group you run. One digest, never a drip of pings.

[ORANGE BRACKETS] are yours. Replace every one with your own information before you send. Orange means swap.

Every [INTERVAL], read [MY GROUP]. Give me one digest with four parts: posts that need a reply from me, questions nobody has answered yet, members who posted for the first time, and anything that mentions me by name.

Skip anything you have already shown me. For the three items that matter most, draft the reply in my voice and put it under the item.

I post the replies myself. You never post, comment, or react in the group.

It checks on the interval. Nothing goes into the group without your hands on it.

Instagram

It reads your account, including everything you saved, and drafts for you. Engagement stays human.

Drafts yes. Engagement no.

It can
  • Read profiles, posts, comments, stories, and account insights
  • Turn saved reels and posts into lists, briefs, and plans
  • Draft posts, reels, stories, and carousels for you to publish
It can’t
  • Like, comment, follow, or unfollow
  • Message on autopilot

Everything you saved on Instagram and never went back to.

[ORANGE BRACKETS] are yours. Replace every one with your own information before you send. Orange means swap.

Go through everything I saved on Instagram in the last [N] days. Group it by theme and show me the groups with counts.

Then take [THEME] and turn it into [A LIST / A PLAN / A BRIEF / A SHOPPING LIST]. Keep the link to the original post next to every item.

Tell me what you skipped and why. Draft only. I decide what gets used.

Your saved folder is a research pile. This is the shovel.

Messenger

It reads your conversations and contacts, and it can act on messages.

You approve the sends that matter.

It can
  • Read conversations and contacts
  • Send, react to, unsend, or edit messages
The rule
  • No send that matters goes out without your yes

Meta Ads

The deepest integration. It will also tell you when your own dashboard is lying to you.

Read-only on your account until you explicitly approve a change.

It can
  • Read live spend, delivery, and results
  • Audit your tracking
  • Study winning campaigns
  • Draft new campaign structures for your approval
  • Run the daily numbers watch
It won’t
  • Change your account without your explicit approval

@Mentions

The watch pattern works anywhere Muse can read: Facebook groups today, and the same shape applies to other surfaces as they connect.

It checks every 30 minutes. Not the second it happens.

The mention watchCheck [YOUR GROUP] every 30 minutes. When someone tags me with a real @mention, send me one digest with the post, who tagged me, and a suggested reply. Skip anything you’ve already sent me.

WhatsApp

Read this carefully. WhatsApp is a channel for talking to Muse, not a tool Muse uses to manage WhatsApp.

If a guru tells you otherwise, they’re selling something.

It is
  • A door to talk to Muse from WhatsApp
  • Running a Facebook group? That work runs through the Facebook connection, not WhatsApp
It can’t
  • Read your other WhatsApp conversations
  • Send as your personal account
  • Join groups
The honest rule across all of it

Muse drafts. You decide.

Nothing posts, sends, spends, or publishes without your yes. And the approval is one tap, not a workflow.

Pro level

Nine patterns that separate casual from pro.

Each one is real and in my daily use. Each one ends with your first move.

01

Side chats are staff roles

One chat per real responsibility, each with standing doctrine.

How I run itTwenty chats: ads, email, social, sales, website, YouTube, books, investing, the community, the to-do list, travel, even fun. Each has my rules, my voice, my numbers hidden. Nothing gets re-briefed.

Your first moveCreate one chat per real responsibility. Name it like a role.
02

Scheduled jobs do the recurring work

Set once. Runs while the phone is off.

How I run itMention watch every 30 minutes. Ad stats every morning. Question digest every Tuesday. Competitor emails twice daily. Results arrive as a message plus a push.

Your first moveTurn your most repetitive weekly check into one scheduled job this week.
03

Subagents do the heavy reading

Parallel readers, one synthesis.

How I run it27 videos, 981 emails, a 93-item folder, twelve books. One brief back, not a week of reading.

Your first moveFor any “read all of this” job, ask for parallel passes and one synthesis.
04

Skills make it permanent

A skill is a saved workflow Muse can re-run.

How I run itMy book-writing system, my analytics connector, my page-builder. When something works, it gets codified.

Your first moveAfter any workflow succeeds twice, say: save this as a reusable process.
05

Memory compounds

My doctrine, my voice, my past decisions, searched before it builds.

How I run itOne rewrite taught it a standing rule: shorter, first-person, real numbers, preview text always. It refuses to mark things done without my confirmation.

Your first moveCorrect it out loud when it’s wrong. The correction is the training.
06

Devices extend its reach

A paired Mac means real folders, not demos.

How I run itMuse reads and writes my actual project folders. Research in, deliverables out, tested both ways.

Your first movePair your devices and grant scoped folder access. Only folders you’d hand to an assistant.
07

It asks before acting, at every level

Four rungs. Climb one at a time.

How I run itLook things up, draft for review, watch and alert, handle within rules I set. Nothing sends, spends, books, or publishes without a yes until I grant a standing rule.

Your first moveStart everything at rung 2, draft for review. Promote only what’s boring and verified.
08

Multi-AI orchestration

Muse thinks. A coding AI builds.

How I run itMuse is the intelligence layer, a coding-specialized AI is the workhorse, and a detailed Markdown spec is the bridge.

Your first moveStop asking one AI to do everything. Spec here, build there.
09

Proactive beats reactive

Briefings, digests, and watches arrive unprompted.

How I run itThe Feed is a briefing I edit in plain English: more of this, less of that.

Your first moveSchedule one morning briefing: agenda, numbers, or mentions. Whatever you’d check by hand.
Honest limits

What Meta Muse can’t do. Yet.

Trust is the product. Here is the list, in full, so nobody sells you a version of Muse that doesn’t exist.

Nine limits. Nothing hidden.

It’s polling, not psychic.

“I’ll check every 30 minutes,” not “the second it happens.” Anything that happens and reverses between checks gets missed. Never promise instant detection.

WhatsApp is a chat channel, not a management tool.

It can’t read your other WhatsApp chats, send as you, or join groups.

Instagram engagement is off-limits.

No liking, commenting, following, or autopilot DMs. Drafts yes. You publish. Pretending to be you socially, no.

Your personal Facebook is read-only.

It will never post, comment, or react as you on your personal profile.

Nothing goes live without your yes.

Every send, spend, post, and publish waits for approval until you grant a standing rule. That’s a feature, not a limitation.

Some doors stay closed.

Nested Facebook comment replies aren’t readable. Some sites block automation outright (Amazon is the one most people hit). A few stores won’t take assistant checkouts. It tells you where it’s blocked instead of faking it.

No live voice here.

Dictation and voice notes work: they’re transcribed like any message. Live voice conversation isn’t available on my setup.

It can be wrong.

It states what it couldn’t verify instead of inventing. Your job: check the big claims, especially numbers. My rule, worth stealing word for word: no figure gets cited publicly until it’s verified.

Credentials never go in chat.

Logins go in the secure store, where it can use them without seeing them. Anyone who tells you to paste a password into a chat is wrong.

Your first week

Six moves. One week.

Six moves. Check them off as you go. This page remembers, on this device.

0 of 6 done

Week one is done. Your assistant is on staff.

Top tips

Twelve moves that separate the pros from the tourists.

Two skills run underneath all twelve. Learn those first and every tip below becomes a one-sentence prompt.

Core skill

Delegate the job, not the task.

Tiny tasks get tiny help. Hand over the goal, the context, and what good looks like. Then say what the result is for. That one line is what lets it make the right calls.

The job: [GOAL]. The context: [CONTEXT]. Good looks like: [STANDARD]. This is for: [A MESSAGE, A PLAN, OR A DECISION].

Review with follow-up questions. Never restart from scratch. New to it? One small task at a time, then whole jobs within a week.

Core skill

Every prompt has three parts.

The job, the access, the gate. In that order. Anything that touches money or sends a message ends at a human.

The jobAudit 12 months of email for recurring charges.
The accessRead-only. Do not open anything else.
The gateDraft it and wait for my approval.

Write every business prompt in this shape and the trust is built in.

[ORANGE BRACKETS] are yours. Replace every one with your own information before you send. Orange means swap.

Social

Give it a job title.

Role framing beats one-off asks. Name the role, your pillars, and the gate.

Act as my social media manager. My pillars are [X, Y, Z]. Draft this week’s posts, platform-native, in my voice. Queue every draft for my approval; publish nothing.
SocialVideo

Two fixes for drafts that sound like a press release.

Sounds corporate? Run the friend test. Reads like an engineering manual? Drop the grade level.

Say it like I’m talking to a friend.
Explain it like I’m 12, with one real-life example.
Marketing

Feed it before you ask it.

Output quality tracks input quality. Your research and voice notes go in first. Then the assets. Ask for sources on every factual claim.

Here’s my research and my voice notes. Collate a step-by-step approach first. Then draft the guide, the landing page, and the email sequence.
Sales

Set a floor before it negotiates.

The floor is the guardrail. The pre-send review is the gate.

Negotiate this [DEAL] thread. My floor is [X], my target is [Y]. Draft every reply; show me each one before it sends; never accept below the floor.
Operations

Prove one loop, then add the next.

Start with one channel: Messenger or email. Connect it, schedule the daily run, review the exceptions. Then add inventory and bookkeeping.

Photograph your equipment and shelves. Its plans fit what you actually own.

Operations

When a connector breaks, use the browser.

New connectors ship flaky. Have it do the same job in the browser instead. Teach your team the fallback, not just the happy path.

Check your popup blocker first. It’s the gotcha that gets everyone.

Service

Triage the inbox into four buckets.

Run it every morning. The highest-return ten minutes an owner-operator can spend.

Triage my inbox into four buckets: action needed, waiting on someone, receipts and subscriptions, and noise. Draft replies in my voice for the action bucket, flag anything with a deadline inside 14 days, and leave the rest filed.
Service

Design the hand-off to a human.

Escalation is a feature, not a failure. Complaints, health information, and money trouble go to a human and never get an automated reply.

If you cannot finish this safely, stop and hand it to me with a summary of where you got stuck.
Video

Write from the hook outward.

One idea in, ranked hooks out. The script comes from the winning hook. The captions come from the script.

Here’s my idea: [IDEA]. Give me ranked hooks. Write the script from the winning hook, then the captions from the script.
Engagement

One digest, not a drip of pings.

Engagement is a scheduled job, not a chore: comments triaged, mentions answered, replies drafted.

Watch [PAGE, GROUP, OR MARKETPLACE] for [TRIGGER], and when it fires, send me one digest with the post, the person, and a suggested reply.
Ads

Audit your own archive before you spend.

Have it read your own archive for anything that risks a policy flag before you run paid traffic on it. Quarterly, and before any big spend push. It lists; you decide what goes.

Audit my last [NUMBER] posts for links and content that risk a policy flag or demonetization. List each one with the reason. Delete nothing.
DMs

Drafts, not blasts.

It drafts every reply. You send. Batch the review into one pass. Automated bulk DMs put the account itself at risk.

Someone promises automated DM blasts? Ask what happens to the account when the platform notices.

FAQ

Questions students ask first.

No. It drafts, you decide. Your personal Facebook profile is read-only, and nothing posts, sends, spends, or publishes without your yes.

No. It checks on a schedule, every 30 minutes in my setup, and sends one digest. Scheduled is honest. Instant is a promise it doesn’t make.

No. WhatsApp is a channel for talking to Muse. It doesn’t read your other chats, send as you, or join groups. Facebook group work runs through the Facebook connection.

What you tell it, what you correct, and what it learns from what you connect. Ask to see the notes it keeps on you, then tell it to forget or correct anything.

Only if you let it. You can opt out of your conversations being used to train Meta’s models. Make that choice in Settings, under Data controls, before you connect anything sensitive.

There’s a free tier, and two paid plans (Power and Maximum) that add weekly token allowances. Prices change, so read the number in Settings, General before you budget. The 1 billion bonus tokens from my code sit on top of any plan and never expire.

Never in chat. They go in the secure credential store or the secure login popup. It uses them without ever seeing them.
Staff, not a chatbot

Twenty side chats. None of them posts without my yes.

That’s the whole system. Every role briefed once, every recurring job on a schedule, every draft waiting for a decision. Start with the tokens. Finish the Settings tour. Brief one chat. Add a role every time a job repeats.

Staff need somewhere to work. The offers, the pages, the funnels, the emails they draft all have to live inside a system that sells. That is what Map To Millions is.

Map To Millions

Your expertise + our system = your scalable business.

Our AI concierge interviews you, then turns your answers into irresistible offers, effective sales pages, and engaging promotions that sound like you, because they came from you. Your funnels arrive pre-built. You customize and launch.

  • Your CRM: one account runs the whole business
  • Your AI concierge: writes in your voice, you give the final word
  • Pre-built funnels: request one, it loads, you launch
  • Live training three times a week, real humans on chat 24/7

$247per month with a 12-month commitment, billed monthly ($2,964 for the year). Price locked when you renew.

Enroll in Map To Millions Opens maptomillions.com in a new tab

Prefer month-to-month? $495 per month, cancel anytime. You lead the brand; the system supplies the buildout. Not included: done-for-you builds or private coaching.