Meta’s AI Agent is Better Than OpenClaw (Manus AI Demo)

Okay, the world is going crazy about Open Claw, the most popular personal autonomous agent on the planet, rumored

to be acquired by OpenAI for $1 billion.

Not a bad outcome for a side project.

But while all of those people are focused in OpenClaw, silently, Manis under the Meta Brand have launched a

personal autonomous agent of their own that we think is pretty incredible.

[music] Kip, I know we've been playing around with it and we've been WhatsAppapping each other to death about this agent.

Why don't you do the setup here and tell us why you love this agent?

>> So, yes, there's a ton of buzz around openclaw because you can basically message it in your messaging platform like Telegram and have it go and do stuff for you and connect it to your services. The challenge is it's like

services. The challenge is it's like it's kind of complex to set up. There's

a bunch of security risks. Like you need to be a pretty technical person to use that and I just quite frankly don't have the time to do a deep setup. However,

there's this company man. and they

recently a couple months ago got acquired by Meta and they essentially have released their own version of openclaw that I've been using that for me does everything I want without the

hassle setup and with the same intelligence and autonomy and really it's like you can go into telegram you leave a voice note or a text message of

what you want it to do and it will go and do it and it'll come back to you and the really cool thing Kieran it has like memory and and can do cron jobs and different things. So, like for example,

different things. So, like for example, I gave it a list of podcasts that I don't have time to listen to, but I think are really good, and it will listen to them every week, and every Monday send me a specific custom summary

to me of all of the episodes from those podcast that week. That's very different than using like Claude or Chat GPT to just kind of type in and get something right straight away. So, um you can kind

of lay it up there. Manus was launched in March 2025, not that long ago. no

>> beat GPT4 on real world benchmarks when it came out. It was called the second deepseek moment got acquired by Meta for 2 billion in December 2025. And now

after all of the kind of openclaw noise, they silently launched this personal agent to everyone. I think it was on February the 6th. And your point is, hey, this agent can do most of what

OpenClaw is. There's a whole bunch of

OpenClaw is. There's a whole bunch of skills that you can use with it, and we're going to show you those. You can

access it through Telegram. You can

access it through your email. And you do not have to worry about your computer getting hacked. There was a lot of great

getting hacked. There was a lot of great studies that showed there's a ton of security flaws and issues with Open Claw at large. And so it feels like if you

at large. And so it feels like if you were thinking, hey, I want to get set up with a personal autonomous agent to see what all the fanfare is about. This is

like a first great step. Go to Manis.

We're going to show you exactly how to use it.

>> Couldn't agree more. And I just want to say shout out to our boy Mark Zuckerberg. Zuck has one core skill and

Zuckerberg. Zuck has one core skill and his core skill is to see what somebody has figured out that's really good and people love and to copy it and that's

what's happening here is like manis took what people and especially the technical community loves about openclaw but delivered it in a web easy to use you

don't need to write code you don't need a Mac mini or AWS or any of these things to run it >> right and so when you come into Manis you'll have this agent tab. When you go

in here, it will ask you to connect your agent and set up your agent. So, you'll

get a QR code for Telegram. And so, you can now run it off your Telegram, right?

>> And other channels are coming soon.

Right now, it's Telegram, but they're going to have WhatsApp and Slack and other channels soon.

>> I'm going to show you a quick use case and then we're actually going to allow you to be able to copy it. If you want more on automating your work with AI agents, our team put together the

ultimate AI agents playbook. [music] It

gives you the exact frameworks to start deploying agents that will automate and do the work for you. Get it right now.

Scan the QR code or click the link in the description. Now, let's get back to

the description. Now, let's get back to the show. One of the things I was doing

the show. One of the things I was doing is I have it set up on my phone and I heard about the OpenAI and acquisition of OpenClaw, rumored to be 1 billion.

That guy, not only is he cashing out, I don't know if you've seen all the photos, that guy is also built like a tank. So, you can like go to the gym and

tank. So, you can like go to the gym and do work. that's what he managed to do.

do work. that's what he managed to do.

And so that acquisition went through and I was like, "Okay, that probably will make a bunch of great content." And so what I did was I basically voice noted my Manis agent. I basically said, "Hey,

like go retrieve all of the best threads on Reddit, all of the best articles that have been written about this acquisition, stack rank them by those, getting the most engagement, and like turn it into talking points for me." And

I did that all through the phone. Just

left a voice message. I was actually out at the time walking my dog. Just send it a voice message. That is the kind of unlock with autonomous agents, right?

All I did was send in a voice note. And

so I come back after walk my French bulldog who's an absolute nightmare to walk and I come to my desktop. And so

you can see here, this was done through telegram on my phone. I now have this thread and I can go in here. This is a skill. We are going to show you how you

skill. We are going to show you how you can build these skills for yourself. in

that skill. It basically goes and it extracts all of that research and turns it into like quoteunquote talking points, things that would make great content and actually turns it in to a little interactive website to make it

easier for me to see what talking points it is extracted. And you can see here it understands what my audience is, right?

CMOs, founders, marketing executives.

And then it pulls in these kind of talking points again looking just from the last couple of days of what's going on around this acquisition and turns it into these categories. It says like here are some educational talking points you

can turn into great content. Here are

some spicy takes, right? And Tropics

legal team, this is a really good one actually, uh, just handed OpenAI a 300 billion gift. And so why was that?

billion gift. And so why was that?

Because Claude was the front runner of OpenClaw. People loved using it with

OpenClaw. People loved using it with Claude. And then what was their

Claude. And then what was their response? They tried to shut it down

response? They tried to shut it down because it used a similar name. Open

Claw used to be called Claude. And so

the legal team got involved and that's rumored to be why he ended up selling to OpenAI. So, this is a great little uh

OpenAI. So, this is a great little uh talking point I could turn into content.

But just look how good this is, right? I

wasn't doing anything from a voice note on your phone and you would set up a skill. We're going to talk about skills

skill. We're going to talk about skills in a minute because they're really important part of building with these agents but >> you can now take this and build content or automate any type of like thought

leadership or viral content you want to build here, right? You can get a great first pass when you are on your phone doing any other thing and you can just tell your agent to create it all and you come back to all of this work being done. As you can see here, all of the

done. As you can see here, all of the research is done. It built the interactive page and I can uh soon because it's coming in Manis which is a team hand this off to another agent. So

I can say another agent should pick up the best talking point and create a first draft. So that is the you know the

first draft. So that is the you know the promise of these personal autonomous agents, right? You can just ask it to do

agents, right? You can just ask it to do something and it goes off and spends a bunch of time and when you come back all of this work is done like pretty pretty incredible.

>> Exactly. And the reason that the output you did Kieran is so good was something called skills. And I think Kieran skills

called skills. And I think Kieran skills are the most important aspect of learning AI this year. The reason for that is skills are the context for

agents to do work at the level of quality that you need them to do. So

skills is basically a system whereas a prompt is a conversation. What that

means is you teach the AI a course skill that is a system skill and it actually has that knowledge and can repeat that task time and time again. one of my favorite films of all time, maybe my favorite film of all time is that a

Matrix and there's this incredible scene and Trinity needs to be able to uh pilot a helicopter and she just calls back to the AI and says, "I need to be able to ride helicopters and it basically teaches her that skill and puts it into

her brain." Not too dissimilar from what

her brain." Not too dissimilar from what skills are. Skills, you can find them

skills are. Skills, you can find them here. You can manage your skills. You

here. You can manage your skills. You

can see I've installed my viral talking points. But if we go into let's say this

points. But if we go into let's say this um similar web analytics and you can see here a skill is basically an onboarding document to teach you how to do something in the

same way you would teach a human how to do a skill you kind of teach the AI exactly how to complete a task. It lives

in this skill.md file and anytime you ask it to do something that involves that skill, it will call back to this file. So you do not get inconsistent

file. So you do not get inconsistent results. You can continue to make that

results. You can continue to make that skill much much better. You can see here it actually comes with a skill creator.

So manice in of itself can help you create skills for itself. And the one thing I would just say is skills are universal. This is not like a manis

universal. This is not like a manis thing, a claw thing, an open AI thing.

If you create a skill.md

file, which is how I teach the AI how to do something, I can use it in all of those different systems. >> And that is why to your point, skill is the most important skill to learn for AI this year.

>> Yeah. And so you do need to go into settings if you're just getting started with manus and turn on skills. And then

some of these are pre-made skills and then you can make your own. But Kieran

to your point, you can make a skill in Claude or Chat GPT or Claude Code and then use it in Manis or vice versa because all they are is MD files or markdown files that have these

instructions that run with your agent when you're asking it to do something specific and to go and use that skill.

And the thing I will say, Karen, is creating skills is real work. In the

future, a lot of products are probably just going to be really awesome skills because getting skills right are so challenging. I mean, I could show you

challenging. I mean, I could show you Kieran probably for an hour, hour and a half on an ad optimizer skill. Uh, let

me walk through this with everybody.

Okay, you'll see how crazy long this thread is because I'm essentially starting off the process and telling it to make a skill. And this skill is to

make it really easy to analyze data for a product. Understand where there might

a product. Understand where there might be an arbitrage opportunity for advertising. Meaning where's there less

advertising. Meaning where's there less competition and I have the ability to buy ads cheaper and get customers cheaper because of that. And then

identify that placement opportunity and then design the creative to be placed in that opportunity. So this is a pretty

that opportunity. So this is a pretty advanced skill and something that would be really awesome to get done. And I

probably still have a couple hours of work before I'm really really happy with it. But what you'll see is I've given it

it. But what you'll see is I've given it different instructions around the platforms about looking at data on similar web. And every time I do that,

similar web. And every time I do that, Kieran, I basically install the skill and add the skill and I run it and test it and I find out all the things that

aren't good with the output. And you'll

see like look at how bad this output is.

Like it's bad. This is for HubSpot's customer agent product. And so I did this and eventually Kieran I had to go and do a bunch of research. I did one of my favorite advertising copywriters. I

built a detailed style guide for HubSpot. So I've built this so that it

HubSpot. So I've built this so that it does this for HubSpot products. I'm not

going to bore everybody with the amount I have gone back and forth. But I have made a very in-depth skill. But now that I have that skill, I can do some very

cool things. I want to show you some of

cool things. I want to show you some of the ads that I have made with my most recent version. you can tell me if

recent version. you can tell me if they're good or not.

>> And so how did it know to run this specific skill? Like you basically tell

specific skill? Like you basically tell it to do a certain task and it goes to look in your skills library and says, "So I have a skill to do this task, right?"

right?" >> Yeah. Yeah. Well, it will look for those

>> Yeah. Yeah. Well, it will look for those skills, but then you name your skills and this skill is named ad oper. So you

know, if you're on Telegram like you were just showing at the beginning, you can say, "Hey, using the ad optimizer skill, do the following." Right? And so

you can call the skill. So, these are some of my most recent ads, Kieran. So,

like we're starting to get the illustration style, right?

>> Pretty cool.

>> This copywriting's getting good. Like,

this is really good classic basic copywriting and design. They're

still not fully there, but like these are pretty close now.

>> Yeah, good first drafts.

>> They're at minimum good first drafts. I

I think if I keep hammering on the skill, I can get to like creative that you would at least run and test, >> right?

>> Why does it see these as the arbitrage over?

>> So, I keep working through the math. It

did a full report for me, but it's basically running math saying we should spend $10,000 on meta on Facebook and Instagram

because our competitors for this product don't allocate that much social media.

that meta ads are cheaper than LinkedIn and cheaper than Google. And so it's identifying that for this product relative to the competitors this and I don't believe the math here, right? Like

the the >> Well, here's one thing. One context is so important. The one thing I haven't

so important. The one thing I haven't done >> is I haven't uploaded any of our historical advertising data and real results. And so if I did that,

results. And so if I did that, >> that's the next part of this that would be really really great. I just haven't had time to do that yet. And so in absence of that, it's going to make up some numbers, but if you got any real ad data to add into something like this,

it's going to be much better, much effective. But the the logic is sound. I

effective. But the the logic is sound. I

like that it's looking at channels relative to the competition for this product right?

>> Yeah. And because they're owned by Meta now, you can just quickly integrate it with your Meta ad platform. And so I suspect this is going to be deeply intertwined with your Meta account. And

so you're going to be able to do a ton of cool paid optimization and things like that through this agent. And if

you're new to paid, something like this is a real cheat sheet because one of the things I've added, Kieran is that it has to give me a detailed targeting strategy for each recommendation.

So, it's telling me to build lookike audiences, interest targeting, what those interests are, custom audiences, uh, like web retargeting and current customers. So like this is something

customers. So like this is something like if you haven't done a ton of advertising like these are things you probably necessarily know to do and it's giving you a great first place to start right.

>> Yeah. Yeah. So basically you can kind of ask it to do it on the desktop app but one of the cool things is you can voice note it. You can talk you can ask it to

note it. You can talk you can ask it to do things in telegram. Walk away. Come

back.

>> I did this like just a minute ago. I

found an Instagram ad of this pillow.

It's from this pillow company my wife loves Ferbber Studios. Shout out to Kimberly and her pillows. Um, and I just left an 11second voice memo saying like, "Hey, I like this ad. Can you do an ad

for customer agent like this ad?" So, I literally screenshotted and left a voice memo and sent it in Telegram. Right.

That's what I did. And it's showing up here in in the web so it's easier to show you all on the show. And it came back with a couple different creative ideas.

>> Very cool. This is why people are going crazy about autonomous agents because you can just interact with them through many different channels. I want to actually show you one of the things that we really were excited by in terms of

mount because we talked about telegram and voice notes, but actually one of the killer things you can do here >> is you can email your agent.

>> I would just say what you're about to show is the most underrated AI feature >> that nobody's talking about.

>> Yeah. And inbox.

>> So basically you have this email address and you can just email your manis bot.

Now one of the other things you can do is you can set up workflows. So if I go here, one of the ones I'm going to set up is deal flow because I get a lot of uh companies looking for investment. And

whenever deal flow comes in, I'm going to say research the company looking for investment and give me the uh bull and

bear case. So whenever an investment

bear case. So whenever an investment comes in, all I'm going to do is forward it to this bot and then you'll see that bot will do that work and it will be in your inbox. So yesterday I was on email.

your inbox. So yesterday I was on email.

I saw some news in a newsletter uh about Bitcoin and I was like actually forwarded at the newsletter and say can you do some analysis on what they're saying about Bitcoin here and that is pretty incredible right that you can

actually just forward emails to different inboxes and have workflows set up to do things with those emails I have a couple set up here one's like to draft

responses to people another is any email newsletter I get that I don't have time to read I can forward it and it'll give me a personalized summary of that email newsletter for any action items or things that are relevant to me. It's

pretty cool.

>> Yeah, this is super cool. So again,

you're seeing the evolution here of a personal autonomous agent can do tasks all the way to their outcome and it's deeply intertwined to all your workflows. Email messaging. So what's

workflows. Email messaging. So what's

common here is not going to stop at Telegram. They're obviously going to

Telegram. They're obviously going to start to launch other ways to communicate with your personal agent. I

think the most obvious being WhatsApp.

That's would be my messaging platform of choice. I think Slack is coming and so

choice. I think Slack is coming and so it's going to be like intertwined across all of your channels over the next 30 days. And the other thing is what you

days. And the other thing is what you actually have right now is you have one personal agent that you talk to. You're

going to be able to create different agents who are specialized in different areas. And so if I want a content agent,

areas. And so if I want a content agent, I can just talk to them and a different chat than I would talk to my investor agent. So you can actually start to

agent. So you can actually start to build teams of autonomous agents that can all talk to each other and you can talk to those and that I think is going to be insane. Kieran, and one other

thing on the email that you may not have seen, but you can also CC anybody on that email and like you could forward that to Manis and CC me and it will also put me on that thread with you.

>> Oh yeah.

>> And we'll both be able to see the Bitcoin. uh result that you analysis

Bitcoin. uh result that you analysis that you uh did and we could collaborate on it together if we wanted which is >> that's a great growth hack for Manis that's a great way itself to other folks

I think all in all we are impressed by Manis you can go get set up now there's not a lot of like technical capabilities you need to be able to do that just connect to telegrams soon

you'll be able to connect to Slacks go email it set up a couple of skills if you want to set up a skill really quickly just tell Manis that I want a skill that can do this thing and it will create the skill for you. You install it in the way that you've installed skills

anywhere else and then you can join the party. Then you have a autonomous

party. Then you have a autonomous personal AI assistant. You've joined the party and you can start to see what all of the hype is about.

>> It's totally true. The the only caveat I'd give Kieran is you will you will probably spend some money if you really start using this heavily especially if you're making any

images or if you are accessing data from similar web anything that Manis has that's really valuable but also has like a real cost associated to it. I burnt

through probably 20,000 credits just iterating through my ad optimizer trying to get that right. And so you can get started doing everything we're doing for

like very cheap to free. I forget if this is all in the free version or if you have to get the base paid version, but if you start using it a lot, you're going to spend some money. And I would just warn everybody >> that's the one downside of Manus is it

is pretty token hungry.

>> Yes.

>> And so you do want to use it for you know strategic worthy things like not just send in every email and say what do what do you think my mom really means by this email? It's probably not going to be the best use of it.

>> And the reason for that though that is good is that an Mannis is a super agent.

uses all the different AI models out there. So it like uses nano banana to

there. So it like uses nano banana to make ads and there's just real costs associated with the using all those models. What's great is you don't have

models. What's great is you don't have to go and go to all those individual models and do all this manually but there is a cost for that convenience.

>> My usage moves very fast to whatever the best thing is. I used to use chat all the time. I couldn't tell you the last

the time. I couldn't tell you the last time I logged into chat.

>> Same, >> right? Like I literally could not tell

>> right? Like I literally could not tell you >> and I would just say all of my usage now is Claude and Manis and that I think tells you a lot of where things are going and part of that's because Manis

has a better user experience to even access like the Gemini models and like Nano Banana. It's like I'd much rather

Nano Banana. It's like I'd much rather be able to just Telegram a screenshot and a quick idea and have it draft something up for me then go into Gemini or Google AI Studio.

>> It also tells you that memory wasn't as sticky as we maybe thought it was. I

thought it was going to be pretty sticky. But it was pretty easy for me to

sticky. But it was pretty easy for me to just start using Claude again. And

Claude does have memory, but Chach has a lot of memory on me. But I think if you actually had to like look at those two things, I just think core value of the output is going to outweigh the stickiness of memory. And they all do have memory. So you'll have some

have memory. So you'll have some semblance of memory across all of them.

But I agree. I don't think uh AI users are loyal in any way. I think we will just gravitate towards whatever the best model to use is.

>> Yeah. And if you look at what we said on the show, one of the things I really encourage everybody to take away from today's show beyond some of the tactical agents you can go and set up that we showed you is that skills and building

skill files are going to be so important to getting better quality output from any of the AI work you're doing. And the

great thing about skills is they're totally portable. As we said, you can

totally portable. As we said, you can use skills everywhere. And so given that that's going to be such a big part of AI, you'll see everybody moving from product to product because they can take

their skills with them. The ecosystem

around skills is just becoming huge.

This is one of many, many, many websites. But this one here alone,

websites. But this one here alone, right, 47,000 skills. This is really intelligence for free.

>> Yeah. So all of these things that people used to have expertise in or they were core skills, you can literally just go down and start to install them. And so

the challenge is not how do I build skills or get them. It's going to be how do I choose the right ones and integrate them into my workflow and kind of adapt them for my needs. And that's why the best people are actually going to be at

a premium because the best people are going to be able to make super unique skills and scale their work. You know,

if you are 10x better than everybody else at something, you can now make a skill around that and get 10x more output at a 10x quality, which is really, I think, the magic of skills and something that I'm sure we'll be talking

a lot about in future episodes.

>> Right. So, that's the episode. Uh, set

it up in Telegram, get some skills, use it via voice notes, use it via the desktop app, use it via email, whatever works for you. And start up your personal autonomous assistant to start

doing some task today. Drop some

comments around the skills you're building or any cool agents you're building. Hit like, subscribe, and we'll

building. Hit like, subscribe, and we'll see you again real soon on Marketing Against the Grain.

>> This data is wrong every freaking time.

>> Have you heard of HubSpot?

>> HubSpot is a CRM [music] platform where everything is fully integrated.

>> Well, I can see the client's whole history. Calls, support tickets, emails,

history. Calls, support tickets, emails, and here's a task from [music] 3 days ago I totally missed.

>> HubSpot grow better.