This AI is FREE & Better Than GPT Operator (Manus AI Demo)
Okay, this is an incredible time to be alive. It is now possible to create an entire AI workforce who work for you around the clock and do anything you want to do. And Kip and I are going to showcase the tool that you can use today to do this for you, and give you some of the best use cases that we found of Manus AI, the very first ever autonomous agent. And, oh, wow, is it powerful. Okay, Kieran, the entire internet is talking about Manus, and Manus is another one of the growing AI startups out of China, and it is really the first real like you don't need to know how to code. You can just type in in your natural language and build a multi step agent. Is what we I think what we thought open AI's operator was going to be but it is actually able to deliver on that promise, right? It's actually accurate. It's pretty fast, actually, compared to
operating pretty fast. And it can do a lot of work, like think about it as you set a task, and if it has to do data analysis, if it has to write some lightweight code, whatever it has to do to complete that task, it can actually do it. And on this episode, we're going to break down how it does it, and also give you some real examples that we've done and use Manus for so you can actually go sign up for the wait list and start to use it today. This is the incredible thing. It's not a vaporware product. You can actually start to use this today. And some of the actual use cases that I've seen in some of the use
cases you run it for, are pretty mind blowing. I want to just go straight in and show everybody something I built in Manus shout out to the Manus team for getting me off the wait list and letting me go in and build here. So Kieran, you and I have talked a lot about in the world of AI, kind of targeting and personalization and conversion rate is kind of a solved problem. And one of the ways I wanted to test that out is I
wanted to say, hey, could I write a prompt that would basically allow Manus to be my own kind of BDR? And what I said is, please identify companies and contacts at those companies that might be a strong fit for HubSpot product offering. These companies should be growth minded, small, medium sized businesses in the United States. I would like you to do the following. I said, first research and identify prospective customers for HubSpot. In this research, please return company name, contact name, and any contact information and LinkedIn URL like this is pretty complicated, right, right? You're basically the setup here for listeners. Obviously we're showing
the actual screen is you're giving it all of the context needed to figure out how to identify the right ICP, which is your your ideal customer profile. Yeah, I love that. And so what we're really trying to do is give it some very precise, specific and like, step by step instructions. And so, like, I wanted it to evaluate each of these company sales and marketing strategy, their strengths and their weaknesses. That is something like, if you are an entry level salesperson, you go out and do you identify a company, you look at its website, you say, Hey, I think it's really good. And I say sales and marketing, because that's the tools HubSpot provides. We help sellers,
marketers, customer service, folks. And so I did sales and marketing. It identified what those sales and marketing strengths and weaknesses are by company, right? And then I was like, write a call script for each of those companies. If I was actually going to pick up the phone and call those companies, like, what would I tell them? And then I was like, Please give me all this information in a CSV file. So that's what I requested, right?
And I think what you're showcasing is, again, you still need the domain expertise to be able to guide the agent, and the better the prompt you give it, the better context you give it, the better you set it up for what the outcome is you want in very much the same way, again, you would actually manage an employee, the better your results are going to be. And that's where a lot of the skill set is going to go to is like how to manage these agentic workflows? Yes, and keep in mind, right now, I'm doing all this for free. This is, this isn't a free version of this product, right? And so Kieran, it were. So first
of all, I tried to do it a couple times, and it was the service was too busy. So one of the challenge with startups is that they have kind of more constrained capacity. Then eventually I was able to get it to work. And then here you see it basically said, Hey, I'm going to start going through these steps. And so it
tells you the steps here research HubSpot products, offerings, and it's it's actually browsing the internet. It's creating files, it's executing commands, it's clicking on elements, on pages, right? So it's, like, it's amazing. This, the level of steps involved here, like, this is very complex, very multi step workflow. It can just scrape all of the websites right. Like, that's That in itself is pretty cool. Like, you have to. Have all these different tools that have built to
script websites where Manus can just go and actually background, write that code and start to that uses a tool called browser use. That's how it's navigating all these websites and pulling it into the CSV for you. Yeah. So I think that's really cool. And so this took about 35 minutes, approximately, to do all of these steps, which, if you think about It's wild. What do you think it would take a human? It ended up doing five companies, and we'll talk about that in a second. How long do you think it takes a human to do five to find do all this work for five companies? I can tell you like that, that, like the average BDR, is going to spend the majority of their time doing these tasks and trying to research a company and trying to identify, probably like product problem fit right, like, what is the core problems that this company is trying to solve, and what are the problems that are a good map? And so this is a the most time consuming part of that job. It takes a ton
of time. And so what it what it actually did here, Kieran, is it gave us a CSV, and if I found these five companies, shout out to these companies. But basically it failed five companies. It found the CEO at each of those five
companies because I didn't specify who with those companies. If I were doing this, like at real scale, like it's one of the things I would do is talk about the specific profile of person, right? And it got me their contact information, their LinkedIn profile. And I think what was really interesting to me about this Kieran is the work it did to determine what were its strengths and weaknesses as it relates to the HubSpot product offering, right? So you can see here it's like it basically created a known taxonomy of of things and whether they were good or bad at it, right? So clear value proposition, emotional marketing approach, user generated content, strategy, free trial option, right? These are all things that were marketing strengths for these companies. Hey, everyone. I want to talk about prompting with you today. Being great at prompting
is like gives you superpowers. It changes how you can build and work and do everything with AI. And because of that, we wanted to make a special guide just for you around how to get better at prompting, framework for how you actually build prompts. And so we've got this baseline framework that you can go and build and customize any prompts that you want to. You can check it out in the link of the description below, and right now, we're gonna get back to today's show. The other thing that like for people to think about when
they're following along here is just how disruptive of a force this is going to be. So what the agent is unbelievable, what this agent is doing right to like, make sure people understand what's happened here. What Manus really is, is is like a people have heard a cursor. Cursor is this development tool that is
wrapped around Claude and other models, but for the most part, people are using it with club point three, five to code, full stack apps, and the only two things that I'm really spending my time on outside of like keeping up to date with this and trying to do all of the things that I do day to day in AI outside of that, it's really like cursor and MCPS. And what this really is, is a cursor like equivalent for autonomous agent, which, I mean, it's like a wrapper that's built around Claude 3.5 and we're going to get into some of that. And then it's extended using tools. So very we're going to do an episode on MCPS this. So
these model context protocols, and I'm not going to go really technical here all day, all people need to know is, like, you're able to, like, extend these agents by giving them tools to use, right? You can say, here's a bunch of tools. And so when I give you a task, you can choose whatever tools you want to use. And that's what Manus, really is. It's like a wrapper around cloud so it's built on Cloud Model and then has all of these tools. And then when you
give it a task, it's like, it's like, using all these tools. I'm going to use the browser use to scrape all this website. I'm going to use my coding tool to code something lightweight tool that I need. And it's really incredible to see just how much it how many tasks it can get through in such a short amount of time. And the other thing I was pretty impressed by is, number one, how
fast it is, like much, much faster than, I think open AI's operator. I'm sure they will scale it up pretty rapidly. And then the second thing is just how accurate it is. And so the reason that my original point here is how disruptive this is. So let's say there's all these kind of data source tools, and today you have all these data source providers, and you would have to go and you would have to say to these data source providers, I'm looking for these kind of people in this kind of market, that I can solve this problem for. You can
easily extend Manus by giving it some more tools, and then you actually have the equivalent now in that agent, because it can call it, can scrape all of the stuff on the web. You give it some access to other data sources, and can pull all that in. You have your own kind of like data provider, right? And for B to B, now you can build your own kind of little data provider that actually gives you all of the ideal customer profiles that you need to actually market to. And I think that's what you originally said was like, We believe marketing is going to become, I believe marketing is going to change from, like, a very volume centric play, how we acquire lots of things, to a very like value. How do you target? Yes, get very. Targeted. And how do you like really over deliver for those people to way increase your conversion rates completely. Because the magic here in what this app Manus did for free was it told it found these companies, found the right person at this company, then found out what the strength and weaknesses were for this company as it relates to HubSpot products, right? And so you have some some weaknesses here, limited lead generation tools, limited pricing transparency, the blog organization is a weakness. Limited enterprise positioning, limited nurture, lead
nurturing, visibility. These are all things that, if you are selling a marketing tool, you want to understand, right? Like it's amazing, pretty incredible. The thing that really blew my mind Kieran is it went and it failed out these weaknesses, and then it went looked at hubspots product pages to figure out what HubSpot products map to those weaknesses. And I can tell you that because we have such a broad product set that is also a very it's a hard problem for sales. And so what it picked up, when I looked through your results, it picked up things that were kind of interesting, right? I said, Hey, this this company do not have a blog. This company has not posted many blog posts, so they're not leaning into content. It actually picked
up that this company had a couple of four Oh, fours. So it's scraping their website, right? And actually, that is just bananas. And so it's able to, like, take all the website, use that as data to say, well, here are, here's my assessment, like, here's my grade, and then, based upon their weaknesses, it did a pretty good job of, not only pattern matching to the right HubSpot, the right HubSpot kind of product at its first attempt, right? You didn't, this is no vision. This is, this is one shot, and then, and then did pricing, and then created a pretty good pitch that alone is an entire couple of days of work. Yeah. I mean, if you're really good at this work, what this did would have taken, would have probably taken you two hours. And if you're like, top level, and you
have 35 minutes, you've reduced like that and this thing, and that wasn't active. It's not like I was doing things that 35 minutes, I was doing something else. And that's one instance of Manus. I would likely, I would challenge people to think about when you have I want to show you this. I'm gonna show you a quick
screen, just like, just this one, because it's just like bananas to think about, like this is actually, I think, one of the best visuals and just how wacky things are going to become. So this is my point. You're showing a single show. I love this. You're showing a single agent doing a task. And it looks pretty cool. What I'm showing on my screen is 50 instances of Manus, all running different social all running different, yeah, social media accounts.
And so there's no reason you can't have 100 there's no reason you can't have a company of 10,000 people that are always doing this work. It is, like, incredible to think about, and Manus for me, the operator. I don't know what you felt like, but when I used open AI operator, I thought it was, like, cool, but you couldn't really grok how powerful it had to be. Didn't feel like it had any scale or power to
it. But this is the first, this is the first time where you're like, holy smokes. If I have a bunch of these things running the amount of work done, I can do 10x 100x in a day. The the other one I want to show you before you get into your example, because it's the the thing I'm obsessed about is the the reverse of what you're showing. So let me show what I mean by that. Oh, please. So what we what we think about a lot, obviously, because we're in marketing, is the our ability to better market to people who want to buy B to B software. What I am really
obsessed about is the B to B's buyer, the B to B's buyer, ability to use agents, to not have to do any of the buying themselves. Because that, to me, is more transformational than the reverse of that, right? I couldn't agree more. And so this is a really good example of what I mean. So this is Manus, and what I believe happens in the future is you don't touch anything that exists on the web today to actually choose software. You get it all built for you in a personalized way. So the example I'm going to show here is Manus
is going to build a aggregator, because with the thing for for rubber mats, and the thing I really care about is the price, right? So I actually want to build my own recommendation engine. That is, allows me to, like, like, my own, my own recommendation engine, like my own kind of g2 crowd for rubber mats, and I can actually sort it based upon price. And so what it's going to do here, and I'm going to go to the end, because, like, people saw your example, and could see it, it's running through all these sites, but I want to get right to the end to show you what it's built. So this here, you can see on the right hand side, it's basically built this whole app, and this app now allows me to sort through the suppliers that I'm trying to find in a very customized way based upon what my core criteria bananas. So think about it, if you were buying any
software, I can say I care about ABCD and E now build me a an aggregator app. That aggregates them all together and allow me to, like, I can, I can look through it based on user reviews. I can look through it based upon whatever I want to do. And not only that, I actually think you're gonna be able to get Manus to go in and actually do the trial, yeah, like, use the software. Tell me, does it do A, B and C, and you can actually ingest that back into the app. That, to me, is pretty wild, like I can't wait to actually start to use this to try to build some product sites that I can actually choose what products I want to buy based upon a custom site built for me. I never have to touch the
vendor site only to actually complete the transaction. I mean, I don't. Are we ready for this? I think this year is, I don't know what you think we're, you know, I think a decade happens in this year. We're only in March. I think things are moving. It's the first time I think things are moving too fast. I actually need someone to
just like, say, can I just learn? Can I just finish my cursor course and build my MCP servers before you release any more stuff? Because I just can't keep up. You and I are, I would call us hyper learners. Yeah, like we're obsessive learners. And this is one of the first, the first time in the last decade where I have felt like I cannot keep up. But the interesting thing is, well, you're kind of leverage, because your leverage goes right. I last week I was, I was mean, you and I were WhatsApp, and I was showing you something I built using Claude and MCPS, which, again, is very is what Manus kind of is. It's just a way better, more sophisticated version.
MCPS are, again, just tools you can give Claude, and Claude can go and use and so the example I had was I told Claude to go research marketing agency services and then build a business that could actually automate those services through AI, and create the website and choose the brand then and then and add it all to GitHub. And it, I showed you. It did all of that we can when we go through the episode, I'll show you what it built so that that took some amount of skill, right? Because I actually had to, like, now, Manus, you don't need to do any of that. They do it. They do it for you. So again, it continues to, it continues to, like, remove where your leverage is. Like, what is my say you're like, What is my leverage? I can do this task. Oh, now you Manus, can just do it. I can't
for everyone, right? So, like, it's con a constant, like, Where does the leverage accrue to is, is, I think, a really important thing to try to try to figure out, yeah, so, so, to close out on the like, the last example about like, this BDR and prospect identification, it did write custom scripts, personalized to each of those companies and their strengths and weaknesses, and talked about how HubSpot could accelerate their growth. I could have given some more specific instructions on the call scripts. I think it would have been much better. Again, this is like a pretty basic first shot, but it is wild, wild
that this was like, this was not possible a year ago, just literally not possible, not even, and now it is not only possible, it's free. The only thing I will say, Kieran, before we go to another example here is, there is limits, right? I had, I asked it to find 100 companies. It found it. It could find five, right? And I tried to have it find more, and it was like, Hey, I'm at the end of my capacity for this task. You need to start a new task. So the context window for these tasks are still smaller. One of the things you and I are
just when we talk to each other almost every day, we're just like, Can the context windows get bigger? Yeah, can the models get faster. I don't need them to get smarter. I need them to absorb more, remember more, and connect more with my other technology. They need to have bigger context windows. Again, yeah, they need to connect with your own, your own data, and to be able to pull in and just your own data really easily. I think the context window is a big one, right? Like huge. Let's say you
want to really collapse a bunch of tools into Manus and for your prospect and an outreach and competitor research, it's like, how much that actually eventually cost you because of the amount of additional bandwidth you need. But I still think cost is going to be a problem for AI, and so, like, long term, it will, like, just be way cheaper to do this than it is today by humans. Yeah. So I completely agree. So I showed you Kieran and everybody watching, like
a sales example, right? How you find and reach out to target prospects in a really interesting way. So then I was like, Well, what's in a market example that literally any company would want to do? And so I had Manus say, like, hey, review all of hubspot.com product pages, rewrite the copy to make an emphasis more on our three key benefits, easy, fast and unified. And the target audience for these pages is growth. Mind is business leaders. Can you create a CSV with each page URL, an updated copy? And that's very basic instructions. I
could have given a much better prompt, but it couldn't find and again, just like it couldn't find 100 companies, it couldn't find every single product page. But if you, if you look at what it did, it came up with, it did five product pages and was able to return the exact request I had. It gave me the current copy. I didn't tell it to give me the current copy, right? Yeah. Inferred its ability to do that, then gave me an updated copy, which I thought, like, definitely fit the bill for what it was trying to do, right? And so that's a very simple example that if you're in marketing, you're going to be interested in doing because you're like, hey, I can now edit and create content at scale with way less manual work at context, if you're just like, I don't even have to give you the URLs. It's like, for most companies, they don't have more than
like, five product pages. So that would be a simple way to do a quick update to all their product pages and like, and it did that in like, 15 minutes. Like, the kind of mundane work goes goes away pretty rapidly with this, like, I can imagine, once it's actually integrated into your tech stack, once it's integrated into your tools, its ability to be able to do things for you and take away a lot of the mundane work will be huge. Now it is a Chinese company. So
again, there is going to be a lot of limitations in the way people want to use it. It's a Chinese company based out of where is it based? Out of Wuhan the other one I want to show you actually. It's not too dissimilar from what I was using. The example I give that I was playing around with with a Claude. So this is it, basically scraping the entirety of Apple's website and then cloning it. So let me give you a pretty great use case. I think this is one of the ones I mix. Ones I'm excited myself to go play around with. So I'm a company of any type, I can now replicate any other business and just ask it to make some slight modifications and tweaks. So any website you are a fan of, you can
actually just go and have it create that for you, but then say, like, change the brand in, change this, change that, and it will adapt it to your style. Now lovable, lovable do a really good job of this as well. I've been using lovable to do mock ups versus actually, any kind of design tools. But again, like its ability to just do this in one shot, that's the thing I was saying to you on the WhatsApp. The important thing to emphasize here, when I say one shot is
it can just, like, replicate the website. It makes no errors, and you don't have to give it any revisions. Your stuff was very similar, like, it just did the thing, and you didn't have to, like, correct it or tell it to do something differently. It just got it straight away. And that, that is the claude's latest model. Its ability to one shot task is pretty incredible. Like the task I give it to do research, provide services, get brand and do website, all within one prompt, no revisions, no error correction, and it did it all out of the gate. And so I think this, the latest Claude model with these wrapper tools are really impressive. The other one, I just want
to quickly show you one thing. The other one that might be just what's possible now is just insane, that it's too much. It's actually the possibility is really too much. But to give our listeners, you know, a kind of tangible one that they
probably do care about, is like recruitment. So it's actually very similar to what you did, but it's actually for recruitment. So this is an example of it, basically going through and actually taking a big loss a long list of PDFs resumes and being able to assess them and stock rank them based upon criteria you have. That's a good one. Now, what I would say is a better
addition to this is take those existing PDFs, ask it to scrape LinkedIn for those profiles and then ask it to ingest any writings or interviews that person has done externally, and combine it all and provide a grade based upon like that, overarch all of that data. Yes, because you actually combine it really easily, and actually it's actually so Manus to be a really great recruitment tool. The one thing that does make me think about is if more and more companies are going to use agents to be able to do prospecting of candidates as well, you really need to have better ability to showcase your work online, like, I don't know LinkedIn. Need to have some sort of ways so you can showcase work in an interesting way, so AI agents can pick it up and use it as part of this criteria. Well, I think this is a very, very important point, as we're kind of wrapping up the show today, is that with these big fundamental innovations like agents, like Manus, like we're talking about today, it creates a second order effect where, like, you need all new infrastructure on the internet, websites need to work differently, social networks need to work differently. All
of these things are likely going to be reinvented over the next decade, right? It's kind of insane, yeah, and it might be the current companies reinventing themselves. It might be new companies coming along. We don't know the answer there, but it's clear there's going to be a bunch of reinvention you have because you're having to build you have a new user on the internet that's a that's basically becomes a predominant user of most of these internet or most of websites and apps, which is an agent, correct? And we haven't, and you haven't, we haven't thought through how to build for agents, right? Because the user is now telling the agent to do all the things. It used to do, but the agent is going to work very, very differently in how it wants to consume that data. We want to end on something like, I thought was also, like, pretty eye opening, right? So, yeah, actually, let me flick to two before I introduce anus, I want to quickly use guys. Guys got a sense of humor. I actually want to introduce why anus is super interesting. So we're gonna, we're
gonna, we're gonna give you a new agent, a bonus here called anus. And so why is anus Interesting? Okay, well, what this is kind of what I started with. So what is Manus, right? So the founder here confirmed this is true. Which is, it's
built on entropic cloud sonnets, right? It's not their own foundational model. It has access to 29 tools. So again, think of that as what you're hearing as MCPS model context protocol, and allows you to extend apps and give them tools to use. This one here browser use is how it's doing a bunch of the browser control, all open source, and then gives, just gives it a little bit around how it communicates with different agents. And then, like, outperforms open AI's deep research on a benchmark. So, like, it's actually pretty incredible that the agent actually can actually do deep research as well, as well as automate all these tasks. That those two things combined are incredibly powerful
and actually open AI hadn't even done that, which really does show the power of combining these tools in really intuitive ways. And so the thing I want to end with was, okay, well, if that stuff, none of it's their their tools or apps, it's all kind of open source or available for other folks to use. What this person did was like, super smart. They basically said, Well, I'll just have Manus create itself that is super smart. And so basically they had Manus create a new agent called anus. And anus is this pretty like Dharmesh would like
this, right? It's pretty good Venn diagram. So Manus AI is a powerful AI agent. The I don't know if you know this, but the invite codes were going for $20,000 online, and so unbelievable. There's a waiting list you're in. Then
there's the open source, which a lot of the tools it's using. Is the open source. Now you have anus. So what is anus? Anus is basically an open source version of itself. So they've created this agent that you can go use in GitHub now that I'm going to actually go install tonight. It's created by cloud
point 3.7, right? So it's created by the model that it runs itself, and it basically just one shot it and rebuilt itself using open source tools. So just insane. We try to just try to figure out, I want everyone to, like, realize how mind blowing this is. There is a business that is went viral over the weekend that has a pricing model that everyone wants to use. They want to use
it so much that there are trade in $20,000 invite codes online. But that model is that tool is so good and cloud sonnet 3.7 is so good at coddin that it was able to replicate itself and put itself onto GitHub, and now you could just use it for free. And I can't get over how mind blown that is like, what happens if, okay, we don't need you in this path, but, like, what happens if I can just one shot any software and open source it? Because it's, I mean, over time, they'll get more and more possible. It's just we
are moving to a world in which brand connection, point of view, the human aspect, is going to be more and more important, because the technical aspect will become more commoditized over time. Darius, the CEO of entropic, said, in 12 months time, AI writes 100% of old code that's try to try to wrap your head around people should try to wrap their head around that. It's freaking mind blowing. I'll be honest with you, I think this was one of the most mind blowing episodes we've ever done on the show. And I think we will look back and this
will be one of the kind of tent pole moments in the history of the show and the history of AI for when, like, major breakthroughs have happened. And this is, like, the first real consumer agent that can do legit complicated work, and we showed you some great examples. Highly recommend you go and get on the wait list and everything for for Manus, if you want to check it out yourself and get on Manus, don't forget about anus. You
just want to keep saying, It's hilarious, but it's you know. But regardless, this is going to be the start of a wave of these type of consumer agents. Kieran, I suspect we're going to be talking more and more about this. I suspect Claude and Claude and OpenAI will both have some version of
this coming in the not that distant future, and we'll be back with those updates when they happen. But if my advice to everybody, as we close out, is you have to obsess about how you're going to get leverage for your company or your in your role, and these types of agents are going to be a critical part of how you're going to get way more leverage in the future than you had today. Anything else? No, all right, this was awesome. My head spinning. Thanks, everyone. Everybody for checking out today's show. We'll see you real soon on Marketing Against the Grain
2025-03-24 05:05
