40:24The Next New Thing · Reaction
GrokBot: Make Money, Grow Businesses, Save Cash
Two hosts react to eleven creators' real GrokBot setups, timing which automations are worth the tokens and which ones need a webhook instead of a timer.
October 3rdThe argument in one line.
Meta's Muse agent is free and already handles narrow real tasks like negotiating a bill by phone or summarizing a PDF, but it still stops to ask permission before nearly every action, so it behaves more like an assistant than a true autonomous agent.
Read if. Skip if.
READ IF YOU ARE…
- You're weighing a free AI agent against a paid one like ChatGPT or Claude before committing another $20 a month.
- You already run one or more AI agents and want a clear read on which specific tasks Muse handles well.
- You want to see real demo footage of an agent making a phone call, buying on Marketplace, or reading a PDF, not just a feature list.
SKIP IF…
- You've already built an equivalent workflow with Zapier, Grok, or Claude and don't need a Meta-specific option.
- You're looking for a technical explainer of how Muse's model works under the hood, not hands-on use case commentary.
The full version, fast.
Meta's Muse is a free AI agent built into Instagram, Messenger, Facebook, and WhatsApp, and this roundup reacts to the best public demos of it in action. Muse can call a company to negotiate a bill, build a self-updating price tracker, pay with a one-time virtual card, buy and sell on Marketplace, read a long PDF and answer questions about it later, turn a vague goal into a finished deliverable, control files on a remote computer, and draft a well-designed morning brief from your email. The catch repeated across nearly every clip is that Muse still stops to ask for approval before it acts, which the host argues defeats the point of delegating work to an agent in the first place.
Free for members
Chat with this breakdown — free.
Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.
Create a free account →Where the time goes.
00:00 – 00:29
01 · Meet Meta Muse
Opens with Meta's own Connect keynote clip claiming Muse will make viewers money, then the host frames the video as a reaction roundup built from the best public Muse demos he's found.
00:29 – 03:13
02 · Negotiating Bills
Peter Yang has Muse call Xfinity to negotiate his internet bill. The agent gets a $60-a-month promo rate with autopay and hands back a full transcript of the call it made.
03:13 – 05:26
03 · AI Phone Calls
Paul Lipsky has Muse call his own company to ask its business hours. The agent openly identifies itself as an AI, which the host says splits viewers into people who find it magical and people who hang up on bots on principle.
05:26 – 07:10
04 · Price Tracking
Another demo has Muse build a self-updating price tracker that watches deals, stock prices, and spending limits, then proactively messages the user in chat the moment something changes.
07:10 – 09:01
05 · Shopping and Payments
Meta's Link feature issues a one-time virtual card number scoped to a single purchase so Muse can check out without ever handling the user's real card, though it still pauses to ask for approval before paying.
09:01 – 09:53
06 · Zapier MCP
A sponsored segment for Zapier MCP, which connects an agent to thousands of other apps while letting the owner restrict exactly which actions it's allowed to take.
09:53 – 11:56
07 · Facebook Marketplace
Peter Yang uses Muse to browse Marketplace listings, list his own items for sale, and message sellers to negotiate price through Messenger. The host says its taste is still unreliable and the constant confirmations defeat the point of delegating.
11:56 – 13:20
08 · School PDFs
Uploading a 50-page school handbook once lets the host ask Muse new questions about it later without resending the file, though he'd rather the extracted facts live in Notion than stay locked inside the chat.
13:20 – 14:47
09 · Goals
Picking a goal in Muse triggers a short set of multiple-choice questions that work like a guided brain dump, then the agent reverse-prompts a finished deliverable, in this demo a LinkedIn carousel, from the answers.
14:47 – 17:41
10 · Local Files
Muse reads and edits files on a second Mac left running around the clock, in fewer steps than the equivalent flow in the ChatGPT desktop app, though it's still more limited than a full computer-use agent.
17:41 – 19:59
11 · Family Morning Brief
Asked for a single-page, print-ready morning brief covering each kid's schedule, email call-outs, and local news, Muse pulls real context from the user's inbox and produces a well-designed PDF on the first attempt.
19:59 – 22:24
12 · Browser Research
Edwin Chen has Muse open its own virtual browser to comparison-shop AI receptionist vendors and carry a purchase through to checkout, stopping only when it hits a signup form and an 'I'm not a robot' check.
22:24 – 24:51
13 · Messaging Integrations
Muse connects natively to Instagram, Threads, Facebook, and Messenger and can monitor them for anything worth a reply, but it only reaches WhatsApp through a side channel and can't manage WhatsApp contacts or messages.
24:51 – 26:19
14 · Andrew's Verdict
The host weighs Muse's free price, and Meta's admitted plan to profit from purchases it influences, against the fact that it still checks in before nearly every action, closing with his usage recommendation for both agent beginners and people who already run other agents.
Lines worth screenshotting.
- Meta is giving Muse away free while admitting on stage that its real business model is taking a cut of whatever Muse convinces you to buy.
- A Muse-placed phone call got Xfinity to drop a bill to 60 dollars a month with autopay, a discount a human rep rarely volunteers without being asked.
- Muse always identifies itself as an AI calling on your behalf rather than pretending to be you, which some people find reassuring and others hang up on immediately.
- Some providers legally require the account holder's own voice on a call, so even a working phone-call agent can still end in a chore it can't finish.
- Link, Meta's payment layer, issues a one-time card number scoped to one merchant and one amount so neither Muse nor the store ever sees your real card.
- Letting an agent buy things for you is less about the technology and more about trust. Some users would rather hand over a virtual card and say go than approve every purchase.
- A 50-page school PDF only has to be uploaded once. After that, Muse can answer new questions about it without the file being reattached.
- The Goals feature asks multiple-choice questions that function as a structured brain dump, then reverse-prompts a finished deliverable like a LinkedIn carousel from the answers.
- Muse can read and edit files on a second computer left running 24 hours a day in fewer steps than the ChatGPT desktop app's remote-control flow.
- Given a loose request and read access to email, Muse produced a one-page, print-ready family morning brief that worked on the first attempt, beating prior output from Claude and Codex on design.
- Muse can open its own virtual browser, comparison shop across vendor sites, and carry a purchase all the way to checkout before stopping to ask for a final yes.
- Muse connects natively to Instagram, Threads, Facebook, and Messenger, but only reaches WhatsApp through a side channel and cannot manage WhatsApp contacts or messages at all.
- The complaint that shows up across almost every demo is the same one: an agent that checks in before every action has quietly become the user's boss instead of their employee.
What a free AI agent can, and can't, do for you yet
WHAT TO LEARN
Muse proves a free agent can already handle narrow real tasks like negotiating a bill or summarizing a PDF, but the same constant approval requests show up across nearly every use case, which means it's still an assistant and not yet an autonomous employee.
02Negotiating Bills
- An agent that can place a real phone call and negotiate a bill removes the single most annoying step of customer service: waiting on hold and asking for a human.
- Getting a transcript of every call an agent makes on your behalf is what makes trusting it with a phone call possible in the first place.
03AI Phone Calls
- An AI caller that honestly identifies itself as an AI avoids one failure mode, deception, while running straight into another: people who refuse to talk to a bot on principle.
- Treat an AI-calling feature as reliable for simple, factual asks like business hours, not yet for anything that requires persuasion or judgment from the person on the other end.
04Price Tracking
- A tracker that proactively messages you when a price drops removes the actual labor most price-tracking tools skip: remembering to go check the dashboard yourself.
- The same watch-a-number-alert-on-a-threshold pattern works for stock prices and spending limits, not just retail deals, once it's set up once.
05Shopping and Payments
- A one-time virtual card scoped to a specific merchant and amount is what makes handing a purchase to an agent safe, not the agent's own judgment.
- Whether you want an agent to confirm every purchase or just go is a personal risk-tolerance decision to make deliberately, not a default to accept.
07Facebook Marketplace
- An agent that still asks you to confirm every single message isn't delegation. It's doing the task yourself with extra steps in between.
- A shopping agent is only as useful as how well it has learned your taste. One that hasn't built context on you yet will keep surfacing options you'd never pick.
08School PDFs
- Uploading a long reference document once and querying it later beats re-searching the same PDF every time you need one fact from it.
- Pushing extracted information into a shared tool like Notion keeps it usable by every agent you run instead of trapping it inside one chat history.
09Goals
- A short set of multiple-choice questions gets more usable output from a vague goal than an open prompt box, because it does the brain-dumping for you.
- Reverse-prompting, where the agent proposes ideas back to you instead of waiting for a fully formed request, turns a feature most people ignore into one worth using.
10Local Files
- Fewer clicks between you and a remote file matters more day to day than raw capability. A simpler flow that takes 3 steps beats a more powerful one that takes 7.
- Keeping a secondary, always-on machine dedicated to agent automations is a workable pattern once you're already running more than one AI tool.
11Family Morning Brief
- Giving an agent read access to your email lets it infer real context, names, schedules, routines, that would otherwise have to be typed out by hand every time.
- Visual design quality is still a real gap between agents. The same request produced a noticeably better-looking one-page PDF here than the same prompt gets from other tools.
12Browser Research
- Letting an agent run its own virtual browser to comparison-shop saves the time cost of opening five tabs yourself, even when it still stops before paying.
- CAPTCHAs and 'I'm not a robot' checks remain the hard wall every shopping agent hits at the final step, regardless of how capable it is up to that point.
13Messaging Integrations
- Native integration with a platform's own apps is faster to set up than a browser-controlled agent, but it inherits whatever that platform chooses not to expose, like WhatsApp contact management here.
- Setting an agent to monitor multiple message inboxes and summarize what's worth a reply is a useful pattern on its own, independent of which specific agent runs it.
14Andrew's Verdict
- A product given away for free is rarely free of a business model. Here the plan is to profit from a cut of whatever it convinces the user to buy, worth knowing before trusting its recommendations.
- The complaint that cuts across almost every use case in this video is the same one: an agent that must check in before every action functions like the user's boss, not their employee.
Terms worth knowing.
- Muse
- Meta's free AI agent built into Instagram, Messenger, Facebook, and WhatsApp that can browse, message, and take actions on a user's behalf.
- Link
- Meta's payment feature that generates a one-time virtual card number for an agent purchase so the user's real card number is never shared with the merchant or the agent.
- Zapier MCP
- A Zapier tool that connects an AI agent to thousands of other apps while letting the owner limit exactly which actions the agent is allowed to take.
- Goals
- A Muse feature that asks a user multiple-choice questions about a stated goal, then uses the answers to reverse-prompt and build a finished deliverable.
- MCP
- Model Context Protocol, a standard that lets an AI agent connect to external tools and apps, used here as the basis for Zapier's agent integration.
Things they pointed at.
00:29channelPeter Yang
03:23channelPaul Lipsky
16:34productRamp virtual cards
20:01channelEdwin Chen
20:30toolVAPI
Lines you could clip.
00:29
“We believe that Muse will make you money.”
11:06
“I don't want that pain. The whole idea of an agent is you do the work, not come back to me every time and check in.”
24:18
“If I'm asking you to do something and every time you want to do it, you come back to me, you're no longer my employee. Now you're my boss.”
Word for word.
Read-along
Don't just watch it. Burn it in.
See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.
metaphor
We believe that Muse will make you money. And we are standing behind this by making Muse free for a huge number of tokens. Everyone's talking about how great Meta is at creating this Muse agent and how much they love the Muse agent.
It can make money. It could be useful for day -to -day tasks. I've watched a few videos to try to figure out what I can do with it since I've been using it since the beginning and what you can do with it.
I'm going to watch it with you right now, show you some of the best ones, and let's talk about whether it's useful or not. Here's my favorite use case from Peter Yang. It's one of the real killer features of Muse that I think the other AI citizens are going to copy, which is actually making phone calls for you.
So I told it to basically help me negotiate my Xfinity Comcast bill. And first it asked me, what kind of voice do I want? Do I want a female or male voice?
And then it asked me for my phone name and my phone number so that it can make the call for me. Then the magical part is, while I was eating breakfast, it actually called Xfinity for me and found that there's actually a way to reduce my bill, just like with AT &T earlier. So basically, I can get a promotional rate of $60 per month.
If I turn on auto pay, I can get another $10 off. All right. And I love that he actually got a transcript of the call afterwards.
Here's the actual transcript of the conversation, which is quite an interesting to see. All right, so here is Muse talking to an actual human. So Muse asked to lower my phone bill.
And it says, just a moment. And Muse said, hey, this is Haley calling on Peter's behalf. I'll transcribe his call for notes.
I'm calling to see if Peter can keep the same internet plan but lower the $84 monthly bill. Can you help me with that? Then the human customer support asked me to confirm the first and last name.
I'm not really sure why it said Chow Min here. That seems kind of weird. But I confirmed it.
And yes, it went back and forth with this conversation. And eventually it got the human support agent to tell me that if I turn on autoplay discount, I can get $60 per month, right? Okay, real savings.
And actually Peter in his video showed multiple savings like that. I will say that it did have a bit of a catch here. Let's play that clip.
So one thing to call here is that Muse is representing itself as my personal agent, not pretending to be Peter. And based on this transcript, the Xfinity agent actually wants me to call myself to get the discount. They want to talk to me.
But the fact that I'm using all this research and found the discount still saves me a lot of time. We've all experienced how annoying it is to call these companies and have to dial one or say two to try to get to a human to talk to and also wait on a line for a human to be put on hold and all that kind of stuff. And now we have Muse to basically do this stuff for us, which is just a massive time saver.
But you still do need to make the phone call, which is such a disappointment. But it's not because of Muse. It's because these companies insist that you yourself call.
In fact, I can't even call if my wife is the one who's registered for the service. It has to be her on the call. All right.
This, though, is not as easy as it seems. Before I go to the next use case, I will say a couple of things about calls. Number one, it's not rolled out for everyone yet.
You have to ask your agent to give it to you and then you get put on a wait list and then you do get to make a call. But that is magical and he is right that we do need to use it. And then a lot of people who are getting calls like this from agents are saying.
get out of here. I don't want to talk to an agent because there's a lot of hate for AI. And so Paul Lipsky actually said, let me hear what it's like when an agent calls.
And he had the agent call his company. Here's what that's like. This sounds like from the business's side.
Hi, I'm calling on Paul Lipsky's behalf. I'm transcribing our call for notes. What hours does the uncarved block LLC operate?
Oh, who is this? My name's Brett. I'm just calling to ask your business hours.
By the way, Brett, as the Muse agent, he got to pick whether it's a male or female voice, and he got to give it some information to prep it. It sounds decent. It's not the most human -like voice there is, but it's decent, and it's representing itself not as Paul, but as someone else.
I think I like it. All right, let's see again how his agent makes the call. Are you an AI?
Yes, I'm an AI assistant. What hours does the Uncarved Block, LLC, operate? and and who who told you to make this phone call paul litsky asked me to call what are your operating hours oh he just that's the only thing he wants to know yes just the business hours for the uncarved block llc it's 24 7.
got it thank you that's all i needed i don't know about you but i thought that sounded pretty good you could definitely tell it wasn't a human which i actually think is a good thing for now But I think there are just a lot of people out there who are so anti -AI, as soon as they even suspect that it's an AI, they get angry.
And he had made a few or had his agent make a few calls, and he did see a lot of anger. But really good use case. If you're calling into a business to get information like, what are the hours?
Can I make a reservation? Super. It will not call other human beings, just businesses.
And it will not let you do cold calling this way. All right, let's go on to another use case. Here is one that I really like.
Stick with me. She's going to show you. how it tracks prices for her, but then she's gonna show you how it relates to more financial related topics.
Let's get into it.
Flag anything that reaches a really good deal and give me one place where I can keep checking all of them. Muse immediately builds a product tracker dashboard for me, which ChatGPT can do as well. But more importantly, Muse will keep looking for deals, price cuts, promotional events across the internet and keep updating the dashboard.
I can simply go to artifacts tab and look at the dashboard for how the prices have been trending. But here's the even better part. I don't have to keep looking at the dashboard every day myself.
Muse will do it for me as well. As soon as something goes on sale, it'll proactively notify me in the chat that an item price has reduced. Since the interactive is very similar to a message app, it'll be like someone messaging me once it goes on sale, like a real person.
In fact, I've built multiple trackers like this. A stock market tracker, which will tell me if a stock that I like drops in value for me to buy at a good price. A financial tracker, which will message me if my monthly dining out expense have crossed a certain limit.
I really, really like this. I love the chat experience, but sometimes I just need a visual that's a tracker and the combination of the two is fantastic. Storing it on artifacts is fantastic.
Let's go on to credit card use.
And when Muse needs to make a purchase for you, it doesn't just hand your actual credit card number over to the website. Instead, Link generates a one -time card number specifically for that purchase. So your real card details are hidden from both Muse and the merchant.
For example, I can ask Muse, find this product for the best price and buy it for me. Muse can browse different websites, find the product, go all the way through checkout, and then when it's actually time to spend my money, it comes back and asks me to approve the purchase and after i approve it the payment can go through using the temporary credit card number what i really like here is that the temporary payment credential is tied to the particular purchase including things like the merchant and amount and it only works for a limited period of time Okay, Muse is really big on telling you that the Muse agent can actually go and shop for you.
I have two thoughts on it. Number one, I specifically gave this example to my Grok bot and it said, no problem, we can handle it. This is much more of a Stripe thing than it is this credit card option that I think Muse is giving it.
Number two. It still needs some kind of okay, which drives me freaking crazy. Maybe it's just the way that I make decisions, but I'm mostly buying from places where I could get a refund, and a few dollars here and a few dollars there don't matter as much as speed does to me, and so I'm much more comfortable giving a credit card number to the agents and saying, go, go do it, and Ramp and other products will let me create these virtual credit cards that I can give my agent.
Anyway, that's my approach, but I do like that they're aiming to give the agent the ability to buy. All right. By the way, if you do have an agent and you want to protect it from doing certain things, that's where Zapier comes in.
If you want, in fact, it's not just even about you. Let's be honest. You know how to control your agent to the millionth degree.
That's why you're watching this. But if you have someone on your team, someone in your family who needs to manage an agent and you don't want to let them go crazy by giving the agent access to every app that they have and letting it do everything that the agent can with the apps, like deleting, which happened to me with OpenClaw.
Zapier MCP lets you connect over 8 ,000, over 9 ,000 different apps right now, and then give the controls that you want. So if you have someone on your team that needs to give access to your team's apps, but you want to restrict what the agent does, Zapier MCP lets you do it. And if you're connecting something or trying to connect something to an agent where there isn't a connection, Zapier MCP will make that for you.
Go to zapier .com slash MCP and tell them Andrew sent you. All right, next. Peter Yang again, selling on Marketplace.
I tried this or I tried the version of this I want to talk to you about. Let's play the clip first. So since you can connect to Facebook, it's also really great for buying and listing things on Facebook Marketplace.
So for example, here I asked it to look up single golf clubs near me to buy on Marketplace. And it gave me a list of these three different options. And I can click on each one to check out the full listing.
And if I want to, I can also list my own stuff. So if I want to list my old iPhone 13 Pro on Facebook Marketplace, it just asks me for the photos, the price, and everything else, and it will list it for me.
And because Muse is also connected to Facebook Messenger, it can also message sellers and negotiate on price. So for example, here, I'm asking it to send a message to the person who listed this golf club to ask about availability. And after confirming, it can send the message and negotiate with the seller.
to buy the golf club at the right price? No. Two thoughts on this one too.
Number one, I am trying to have Muse buy a bed frame for a house that we're staying in temporarily. It is just not bright enough. Now that could be me.
It could be that I'm not using it as a daily driver and so it hasn't gotten to know my taste, but it's constantly being stupid. It's saying, okay, how about these four other options that you have? Or these four options are available.
That's one thought. The second one is notice how he still has to confirm every freaking message. I don't want that pain.
The whole idea of an agent is you do the work, not come back to me every time and check in. My better solution for this is give Grokbot or another agent access to your Facebook account. If you're me, that's what I do.
And then it can negotiate and go back and forth. I did that before we left Austin, Texas, and Grokbot was able to go back and forth. There are too many restrictions here for my taste.
I do not like it. It has not been working for me. But I guess it's a direct connection.
And for people who are looking for a free agent that will do it. Good luck. There you go.
You got it. Let's go on to the next one. I'll keep giving my opinions as we go on here.
Ooh, PDF for events. I have this 50 -page PDF document that my daughter's school gave us, which contains all the information about her school. I can't remember everything and I can't keep searching the PDF every time I need to know when is the next holiday or when is the next parent -teacher meeting.
Well, that's why we have an assistant now. I simply attached the PDF to Muse and said, manage my daughter's school information for me. That's it.
It read the whole document, retrieved every information and stored it for future use. So maybe after a few months, I can simply ask, Anya will be on vacation for 20 days. What should I do to let the school know?
Muse will know who my daughter is. What are the school vacation policies? If we have to email someone, then pull the right email and then maybe even draft a nice one.
I don't have to attach the PDF again. I don't have to specify all the details again. Muse will know it.
Okay, my preference for doing something like this is to not just give it to Muse and say you hold on to it. I like to say, hey, Muse, here it is. Put it in a Notion doc because a month from now, a week from now, honestly, a day from now, because it looks like OpenAI is about to release something, you're going to want another agent to have access to the same data.
And by putting it in Notion or something, you pick whatever your version of it is, all the agents will have access to it. And you should never keep going back to PDFs like this or contracts. You should have the agent go back and find you the relevant information.
Good use case, though. Let's go on. This might actually be my favorite feature of them all in Metamuse, which is goals, which is basically the brain dump to reverse prompts I talk about all the time, but built into the AI agent.
So you can click a goal. It asks you a bunch of questions, right? So I clicked on the goal to grow my creator business.
And what did it do? It went in and started asking me a bunch of multi -choice questions on that goal, kind of coaxing me to brain dump in it. And then from there, it reverse prompted.
So basically said, here's everything I can do based on that goal. One of my goals is I want to grow on LinkedIn. I came in and came up with a bunch of ideas on how it can help me grow on LinkedIn.
It actually ended up building me a carousel for LinkedIn. It's not something I ever thought of doing with this. You know what?
If you're on LinkedIn, you know to do that kind of thing. But I like now understanding what goals is. If you use meta the way that I do, it constantly tells you, how about doing this?
How about doing that? It's like a little too pushy with all the different things that it can do. And because of that, I tuned out all the things that it can do until I saw Alex here say that goals.
was a feature that I should pay attention to, and now I understand what it is. This is one example. I probably could come up with a deeper, more personal example that would put that smile on my face like he has a smile on his face when he talks about this.
By the way, if you like this approach, every time something new launches, I go out and I pick the best videos. Subscribe so you can get that. Let's go on to the next one, local files.
Very few people are talking about it, though I think Mark Zuckerberg recently did with Joanna Stern about how it's also able to control your computer. My other computer, my Mac Mini. That's a little computer that I leave on 24 -7.
I don't even have a monitor plugged into it. So what I'm doing right now is screen sharing that into my MacBook. So that's what we're looking at right now is the screen sharing software, and within that is the desktop of my Mac Mini.
And on that computer, I have Muse open 24 -7. That's because I found that surprisingly, Muse is very good at accessing files on that computer. So I have that computer set up for months now because I have ChatGPT and Cloud running automations.
on that device 24 -7. And another thing I could do is use the ChatGPT desktop app to access files on that computer. And the experience works fine.
I use it all the time, but it's like a little bit clunky. You have to open the side menu and then click on remote. And then at the top, you have to choose the computer you want to get into, which is the Mac Mini in this case.
Let me pause. When this came out, we were all super enthused about it. It is amazing that what he's expressing right now It's true.
I also find it. It's not that many steps to go from your mobile phone to controlling your chat GPT on the computer, which then controls the computer. But the same thing that Paul Lipsky is saying right here, I experience the same things like now I got to go to here and then I got to go click there.
Why can't they make it even easier? It's funny that one day we're all amazed by this. And then a few days later, we come back and go, I don't want to do the drudgery of clicking around.
All right, Paul's right, though. Let's go on. And then from there, you can start a new chat.
So it definitely works. It just requires a few steps. With Muse, on the other hand, the experience is a lot more straightforward from within any chat.
I can tell it's something like, hey, I left a document on my desktop on my Mac mini. Can you just look at that and give me a quick summary of it? So now it is looking at this file right here.
And then it just gave me a summary of it. And so now I'm going to ask it to make some changes to that file. Hey, can you just take that summary and just add it to the top of that document in a summary section?
And it says that it's done. So let's take a quick look. I will open up this file.
And right here at the top, you can see the summary, the thing it just added, and then the rest of it. And it's that easy. So like I said, other tools like ChatGPT and Cloud are a lot more powerful.
They can do a lot more things on your computer. But if you just need access to local files, reading them, editing them, Muse is so straightforward and so easy to set up and use for that. Now.
Fair. I like this. Let's go on to the next thing that I think is so cool.
Every morning we print out a morning brief with the schedule for the kids, what the kids need to know, what we need to know, and then news for the day. And so I prompted it. I said.
What I really want is a PDF agenda for tomorrow morning, kind of like a newsletter, one page style that I can print out and put on the kitchen table. I wanted to have not only our schedule for a week, but highlights of each kid's call outs for the parents. You might need some information from my email about what's going on.
And then recent news, recent events, things in San Francisco, the weather. We can make this kind of a fun read for the whole family that we can talk about, not just the logistics, but also things that are going around in the world, in our city, in our lives. And I.
Y 'all, I really like this. So it did have me connect my email address. Then it ingested some things that it knows about me and my family from my email.
And again, this was a really nice experience. It found out my kids' names. It found out my husband's name.
It found the kids' activities, like they do jujitsu and basketball and piano and all this kind of stuff. It inferred that we're civically engaged in San Francisco. Interesting.
Okay. I like, let's see the result. open clock email this is an idea that my husband had and we have been printing these out and this is the first one that has been one shot in a design i actually like so this has made me want to test some of the facebook models because i have hated the front end design of a lot of these pdfs out of clod and out of codex they've just been like kind of boring and this is so easy to read so beautifully laid out and you know what I think everything that she just said here you can do with every other agent, but she is right to call out the design.
There's something about good visual design that makes me see in a – like I could just quickly scan and understand what's there. I do not like how Rockbot and how – I guess other models are really bad at giving me the design that I want. Even with Claude, I have to really prompt it to help it understand what I'm looking for and get away from that cream Claude design style.
Very good. I like this use case. I also like actually the idea of just printing out for the family what's going on every day.
It keeps me from looking at my phone and telling the kids what's up. Okay, let's get back to a business use case here from Edwin Chen. I actually do research on his own computer.
So I just tell him, find me the best price on voice AI products in London, compare at least three suppliers, find the best one in the basket and stop before paying. Take your time. I'm going to go and do something else.
So it asked me kind of what product, and I mean AR receptionist. So now it actually opened its own virtual browser and actually doing research. So if you go to the browser, if I click on open browser, this will be the browser, the virtual machine that Muse is actually operating on.
So now it's actually checking for like VAPI, the documentation of VAPI, just to see how much it is. So it's actually doing this live on the run. So technically, what I can do is take control of the browser and just do it on the machine, but I don't want to do it myself.
So now it's actually looking up, for example, the phone number pricing, UK cost per month, and just to see how much this actually is.
Okay, let's keep going. There's another clip I want you to see. By the way, what is with all these people holding on to their little microphones that's supposed to be pinned down like a lapel?
Okay.
I don't know why I notice these things. basically finished researching basically thinking retail ai has to pick for me and it's now doing the retail sign up and will stop before any payments so basically trying to create this own account with his own email which will be interesting to see what emails they decided to use for i'll actually try to log in with my email interesting Yeah, so you try to use my email, but obviously there's another password.
And then you say, I'm not a robot checkbox, but obviously it's a robot. So I can take over and do it myself or I guess take it for this sign up. Like, I'm not going to risk it.
I'm not actually, I already have a retail account. I'm not going to buy it. But this is just a showcase for you, like the ability that you can do.
The I'm not a robot thing drives me crazy. It's, I'm sending an agent to go pay you money. Like, why?
Help me figure out a way to do this. Okay, finally, I want you to see something that is unique to Muse from Peter Yang. I think one of Muse's biggest differentiators is that it can connect natively to all of Meta's apps, like Instagram, Threads, Facebook, and Messenger.
Notice that he didn't say WhatsApp. WhatsApp is... probably bigger than all of those as far as usage, it will connect to WhatsApp in only one way.
It's nice. I can go into WhatsApp and I can see my Muse agent and I can chat to it from WhatsApp instead of using the Muse app itself. And it's a little bit restricted because I can't see the browser use, for example, but I like that I have that.
What it will not do is manage my WhatsApp messages, which I'd love for it to do. Do little things like... add names for people who I've started chatting with, but never added their names into my phone books.
I don't know who they are two weeks or two months or two days later. It can't do that. And that's kind of frustrating.
Other agents can when they're controlling it through the browser. So he didn't say that. It's very important to include it.
And real quick, I actually set up a messages thread and basically asked it to monitor my messaging apps, not including email, and let me know if there's something that I need to follow up on. And then it checked iMessages and asked me to connect to all these other things. And it turns out that all I'm getting through Instagram and threads DMs is a bunch of spam.
So nothing super important, but you can see how useful this is. I could see it. I could see it.
There's an issue. Let's go on to the next clip. Messenger, WhatsApp, and iMessages.
It's just hard to keep track of everything. So be able to set up, for example, a scheduled task for my muse to check unread messages every day and let me know what's important. It's just incredibly valuable.
Okay, one thing. Before it sends a message out, it has to check in with me. Again, if I'm asking you to do something and every time you want to do it, you come back to me, .
Now you're my boss. Andrew, I want to send a message out through Messenger. Can I send it?
Yes, go do it. Andrew, I want to do this. Go do it.
It's just too much. Then, so all the other agents don't have to do it. They just control Facebook, Messenger, and all that through the browser.
So yes, it's nice that it's a direct connection, but it's a crippled direction connection. I don't love it. But there's some word that he said in there.
He said iMessage that I didn't know before. So I went into my muse and I said, can I actually have you control iMessage and tell me what's in there? And it said, yes, it could through the phone.
And I thought, okay, I've learned something. Okay, so with that in mind, let me give you my overall opinion here about Muse. Number one, the fact that it is free is really exciting.
Now, we know Mark Zuckerberg, the way that he works. If you continue watching that presentation that I showed at the very beginning, you'll see that he says, we are planning to make money by this presentation right here. We are planning to make money.
By taking a cut of sales. And obviously people are going to be spending a lot of money on this thing. And this thing is going to keep promoting things.
Beyond the features that it has. It'll promote you to buy stuff. So that's how they're going to make money.
For now it is free. And you can use it. And if you know somebody who is interested in agents.
But doesn't want to spend the 20 bucks a month. Or create a computer for Hermes agent. Or go even more.
Like you really. 20 bucks a month is not going to get you enough. This is a good opportunity for them to try it.
And if they're in the Facebook ecosystem. Phenomenal. I think they're going to like it.
If you're experienced. You might use it as just. extra free usage for stuff that you might want them might want to offload over to meta and you can have your agents uh that you actually like guide meta and send things their way i do that with my grokbot agent it sends things over to chat gpt it sends things over to claude etc so now meta is another agent for it to send uh work to So if you're experienced, I think that's the way for you to use it.
So if you like all this and you want more, I've got another video for you to watch. And if you don't like this, let me know in the comments why not. But this is my goal.
I want to study all these things by personally using it and watching the best videos. And if that's where you are, see you on the next one.
The Hook
The bait, then the rug-pull.
Meta's own pitch for Muse is blunt: it exists to make you money. This video puts that claim through the paces, pulling the best public demos of the agent calling companies, shopping, and reading files, then grading each one on whether it actually delivers or just asks for your approval one more time.
Named ideas worth stealing.
13:20model
Muse Goals reverse-prompt flow
- Pick a goal
- Answer a short set of multiple-choice questions
- Muse treats the answers as a structured brain dump
- Muse reverse-prompts and builds a finished deliverable from them
Instead of a blank chat box, Muse's Goals feature interviews the user first, then turns the interview into a finished piece of work like a LinkedIn carousel.
Steal forany onboarding flow that needs to turn a vague goal into a specific output
How they asked for the click.
VERBAL ASK
14:25subscribe
“Subscribe so you can get that.”
A single soft line dropped mid-video right after praising the Goals feature, not saved for a dedicated end screen.
MENTIONED ON CAMERA
FROM THE DESCRIPTION
PRIMARY CTAWhere the creator wants you to go next.
Visual structure at a glance.
