Transcription
In my previous video, I said that if people like this content, I will do another one with a bit more challenging tests for Manus. And yeah, people liked it, so here we are. So today we're going to do a bit of different things. The first thing we're going to do is a coding job I found on upwork.com that is promising $10 if you can solve this. So I have linked Manus the link to the URL of the job; we're going to attach a JSON file as an example, and yeah, we're just going to let it read the instructions for a job and see if it can make me theoretical 10 bucks, uh, that I can go claim if I wanted to.
Next challenge is going to be a data science challenge with a CSV file with 700 post-college salaries data. So we're just going to run this uh prompt here: “With the current state of the world, I want to find out the top three careers I should consider. Include some good arguments based on tech progress, geopolitics, macroeconomics, and present these results in an MDX file with some graphical visualizations to emphasize your point.” Uh, this one I was really surprised with; this turned out pretty good, so uh, you should be pretty excited for that.
Finally, we have an email challenge. I wanted to see if I could log into my (actually ProtonMail) in the containerized computer and actually answer some emails, send some emails, do some research on a newsletter, and actually send that newsletter to myself. Pretty interesting results there too. So yeah, let's just head over to Manus and start with the coding job from upwork.com. Okay, so let's just grab all of this right. So I think we're just going to do that; let's head over to Manus. Let's paste this in, uh, and you can see here I said I have attached the JSON file that's going to be the example for the job. So let me just upload that. So we're going to go downloads, upload Raw JSON, so because we need this example to check if it's going to work, and that is gonna—all I'm gonna give it. So hopefully now we can open up the URL we pasted in here from Upwork, and we can find out uh what the job is uh by basically letting the agent figure this out by opening this URL in the browser, right? So let's—that is the first thing we have to check; uh, if not, we have to feed it some more information. So let's see what happens here. So now it's going to try to go to the URL. Yes, that worked perfect. So that is very good. You can see uh "Python automation expert with complex JSON analysis," but now he kind of has the idea what the job is going to be: create a Python script to analyze the sports game JSON data and it's going to count the player statistics and output the results in an Excel format. So it's still looking for a JSON file; uh, I might have to—I'm not 100% sure how I can do this, but let's see if we can figure this out. Okay, so I'm just going to uh try to upload it again, then I'm going to give it the URL to download to see if that works. Okay, perfect. So now you can see we have the URL here and we have the JSON file. So now it means that we kind of have all the information we need to try to solve this job. So it's going to get going now, hopefully looking at the contents, and it's going to start writing some Python script to actually try to solve this task here. Okay, so what's pretty cool now is that we can kind of see now in the Ubuntu terminal here that it is trying out to test out different ways to solve this. So that is pretty cool, and I kind of look forward to seeing what the Python script we're going to come up with to try to solve this. Okay, so what's pretty interesting here that it went to a bunch of these uh commands here to try to understand the JSON, and you can see now it's going to start writing this. So, "I analyzed this uh structure to understand requirements; now I will develop a Python script to process this data." So we're going to export this probably as an Excel file. Yeah, we're going to count the statistics of each player that did actually align with the job here. So that looks pretty good. Uh, yeah, so let's see what happens now. Now it's going to write the code, and then we're going to test it to see if this actually works and can it make me 10 bucks? Let's find out. Okay, so you can see "I've successfully completed the task of analyzing the data; the provided script was uh executed successfully; the count—yeah, we count all this good; the results have been saved on an Excel file." Can we—what, preview this? Okay, yeah, this looks pretty good. So you can see here we have all the singles, the doubles. So I guess that should be $10 for me, but we also need—right, we need the Python script. Okay, so we downloaded this now; let's try to run it locally just to see how it works. So let's just run it here: python analyze football stats.py. Uh, okay, so we got an example, but you can see we saved this uh uh XL file here. So you can see we get a a sample here, but let's open this Excel file uh in like Google Sheets or something so we can see the full file. And yeah, here it is. So you can see we have all the players' first name; uh, we don't have the last names; I don't know if that was provided, but you can see we have the numbers, the team IDs, and we have all the stats. So we are counting like uh hard hits; I don't know what—I don't know anything about baseball, but you can see we are counting from the JSON file. So I guess if we went ahead, we submitted this job, and and maybe we could get the 10 bucks. So I will say Manus AI agent now did make us $10. So yeah, that was pretty cool. Uh, so let's do a new session now. Uh, I guess we probably could have done that using maybe Claude or ChatGPT too, but uh this was very autonomous, so I didn't really have to do anything. So that was pretty cool. Uh, but let's move on; let's do test number two, and that is going to be data s—
Again, I just want to grab this. So now we're going to upload our CSV file with this post-college salaries data, and yeah, I'm going to show you the prompt. So the prompt is going to be: “With the current state of the world, what three top careers should I consider? Include good arguments for these choices based on tech progress, geopolitics, macroeconomics; present the results in an MDX file with some graphs visualizations to emphasize your point.” Uh, we also got to remember to upload the the Kaggle data; I think it's archive final post-college salaries. Good. So here is the data we have. So we have like a major, the degree type, early career payment, mid-career payment, and high—meaning I don't know percentage, but basically it's an overview of a lot of salaries, probably in dollars, of different jobs—700 different jobs. So that is pretty cool. So let's see what Manus can do with this. Like I said, we're just going to let it run and see what happens. So we can see that the first thing we are doing is actually looking up the CSV file. Yeah, I think that's a good start; there's a lot of stuff to do here, so it's probably going to—did it save it? So we also instructed Manus now to do some research to find geopolitics, tech, right, to kind of compare with this data to find a perfect career for us that is looking forward. So that's going to be interesting to see what top three career choices it will recommend here. So now we are doing some kind of Python script here to create some kind of graphs, I think. Okay, so now we are actually getting into the research; now we are searching for technological projects, trends to identify future-proof career paths. So we are looking at Gartner. So we're looking at some—yeah, I saw AI progress or something here. Okay, so we are in the Gartner Top 10 Strategic Technological Trends for 2025, some kind of report I guess. Now we are at the Accenture; so this is also some kind of job report I guess, but I'm not going to interfere here; I'm just going to let uh kind of the agent do its thing here, and it's going to be interesting to see what we end up with uh when we have done this. So I can already see there's a lot of AI emerging trends here, so that's probably going to affect the final suggestions, right? Okay, so you can see "I'm analyzing post-college salary data; researching technological trends to identify promising career paths." So it's going to continue here, and because I thought this was going to skip it, so I just said "What about geopolitics?" So now probably it's going to check this out. So let's see—see what happens; I'm not going to interfere anymore now, I think. Okay, so you can see basically now we just have to identify the top three careers, create some visualizations, and compile the document. Okay, so I think we are closing in on the final uh final report here. Now you can see "Top Careers 2025." I was kind of hoping for a bit more into the future, but let's see now when we open this. So it created a page for us. So yeah, I'm going to open it, and I want to see if we have some uh graphs and statistics, maybe our visualizations to actually back this up. Okay, so let's see here now. So here you can see "Click to view." Oh, this is a massive report, right? So we have an AI-driven data analysis. So I'm probably not going to be able to go through everything, but we can take a look at a few things here I think that can be interesting. So if we look at at the methodology, we looked at financial potential, meaningful meaning and fulfillment, okay, resilience to future change, so technological resilience, geopolitical and economic resilience, stability through inflation, demographic shifts, economic cycles. Pretty interesting, current state of the world. So we're seeing the emergence of agentic AI, post-quantum cryptography, spatial computing, energy-efficient computing, polyfunctional robots. Pretty cool. Geopolitical factors uh trade chain reactions, EU at a crossroads. That's pretty good. Biotech as a new frontier. Yeah, EU at a crossroads; that is definitely true. And we have increasing protectionism and tariffs. Hello, Trump. That is also true. Macroeconomic trends here, we kind of go into uh projected to cool down in 2025 and 2026. Ooh, interesting; I didn't know that. Inflation expected to continue easing; unemployment rate is projected to rise. Huh, pretty cool. Here we have the top three recommended careers. So we have petroleum engineering, information and computer science, and building science. Uh, I just want to see why petroleum engineering; that seems pretty not so good if you ask me, but I want to see the reason. "Tops our list due to exceptional financial rewards; strong resilience to future change. While this might seem counterintuitive—" Yeah, that is what I thought—"in an era of energy transition, several factors make this field particularly valuable." Okay, so this is actually pretty good if you ask me. Technological advantages, geopolitical significance, competition for energy resources continue to intensify, making expertise in this area highly valuable. I'm quite impressed here, to be honest. Energy demand uh remains relatively inelastic even in down economic downturns. So that means we still need energy even in our recession, right? That is also true. Second, we have information and computer science; we talk about AI and automation. Yes. Geopolitical, digital—sorry—has become a major geopolitical concern. Economical resilience. Building science uh emerges as our third recommendation. So again, we go through some things; I'm not going to—oh, look at this graph; that's pretty cool. Uh, I'm not going to read through everything, but I'm pretty impressed. So we have a conclusion here. Uh, I want to see if I can share this, and we even got the references. I got to say I'm pretty impressed by this; I must say. Uh, I want to see if I can share this; maybe I can leave a link in the description so you can go read it yourself. Uh, but this graph was a bit strange; I don't really understand this, but uh other than that, pretty impressed; it made some good arguments, and what I was really impressed by was this; I really questioned petroleum engineering, uh, so I like that it mentioned that while it might seem counterintuitive in an era of energy transition; that was pretty good; that was a good find. So again, I'm pretty impressed by this; I would say this report is not bad—not bad—pretty good. I guess maybe the graphs were a bit boring; it just shows the the top pay, but I guess that's fine.
Okay, so the last thing we want to do is something with email I haven't tried yet. So let me just copy this, and we can quickly go through what we want to do. So yeah, let's do a new session, and I'm going to do—let's log into my AI agent mail; that is a ProtonMail; answer the latest email with a polite response; so I'm going to send myself an email; do some latest geopolitics uh news research; send the findings in an email in a newsletter format that is easy to read. So that is going to be what we're going to try to do. I guess we have to log into ProtonMail from the browser container uh manually, but let's see anyway now what is going to happen. Okay, so you can see we have our to-do list now; we can take control here, and this means that we probably can go to—let's see—protonmail.com. Okay, and I'm going to try to log in here; I'm not 100% sure if it's going to work, but let me try. Yes, that worked. Okay, perfect. So you can see here I have sent myself an email; uh, I'm not going to click on it; I'm going to ask—yeah—Manus to do this. So now we have kind of logged in; that is good, and I'm going to say "Great, uh, we are logged in; continue." Okay, perfect. So this means that the next thing was going to do was to open up the latest email and answer—answer it with a yeah—a response. So let's see how that is going to go. Okay, so it just gave me like a generic mail here, so I'm going to ask you to open the browser and read the latest email. Okay, so you can see now we are going into our email; uh, you can see we have the Claude uh that is from my other AI agent throwaway email I have. So I sent to this now; I want to see if the agent can navigate the ProtonMail interface or yeah and actually click on this email. Perfect. So you can see there's the email: "Hello Manus, I hope you're doing well. I've just been upgraded to a new version 3.57; it's awesome. What version are you using? Best regards, Chris." So let's see what it responds here now; it's going to try to open up a new uh yeah; that is good. So I want to see if it probably going to do a draft, but I want to see if it's going to write the email for me. Okay, so it's going to compose it; that's cool. Okay, so you can actually see we are writing the email. So I'm—I want to see if it's going to send this or if it's just going to keep it in the draft now. Okay, so it did actually send the email; I wasn't quite expecting that, but that's pretty cool. So that means we can go to our other uh Gmail or uh yeah my throwaway AI agent email, and we hopefully can read the email. So this is just going to continue here now and do research for the geopolitics news. So let's go to our Gmail. Yeah, we got a new email; I'm going to open this. So it kind of messed up a bit here; you can see it didn't remove this "Sent with ProtonMail," but it wrote: "Hello Claude, thank you for your message; it's great to hear from you. I know that you've been upgraded to version 3.7; that sounds exciting. I'm currently using Manus version 1.0, which has—has been working well for my needs, but I'm always interested in learning about the latest improvements in AI technology. What new features, capabilities do you offer? I would love to hear more about your upgraded experience. Best regards, Manus." So that was kind of my agent sending an email. So you can see now it's searching for some geopolitical news, and again it has to go back to the ProtonMail to actually compose the latest newsletter-style email it's going to send me. So that's going to be interesting to see. So here we can see we have collected some geopolitical research; we have yeah a lot of stuff about of course Donald Trump, Russia and Ukraine, and all the stuff that is happening around this—yeah—basically all the major geopolitical uh things that are happening: NATO leadership, Trump and Greenland, Middle East oil industry. Okay, so that's fine, but I want to see if we can actually take this, go back to our ProtonMail, and send this to my requested email address. So we just going to open Proton again; that is a good sign; I like that. So now we probably have to click on new message, I guess. Okay, so it found a new message; now it's just going to put in the correct email address. Yeah, again, uh, they must have done a lot of—of testing on this because uh there is not a lot of bugs to be honest. I know they are most likely using Claude at the back end here, and they are using tools like browser use that I have covered on this channel, and they are using maybe some MCP servers that as—as something I read on Reddit, but I'm not 100% sure. Uh, okay, so we added subject, and now we're probably going to have to click the compose field and send it, but uh to be honest I don't really care if they have used some existing technology because it is kind of hard to put everything together, make it working this smooth, uh, but of course they are limited on usage; I know there's a lot of guys on the waiting list wanted to try this out. Uh, I on my member section gave away for passes. Okay, so we are writing; that's cool. Uh, okay, so that does not look correct; interesting, but let's see anyway. Uh, like I was saying, then I asked if I could get some more codes to give out to members, but they said "Too much pressure; too many want to try it," so we couldn't give out any more codes, and yeah, I kind of see why, but uh yeah. Uh, so let's see now; it didn't really actually send the newsletter; that was a bit strange. Ah, it wants to attach the newsletter as a file. Okay, now I see. So let's see if we can do that; that is a bit of more—more different uh task. Uh, yeah, let's see if we can do it. Okay, so here we ran into some issues; it says uh "I sent the newsletter as an attachment," but we haven't even opened the browser, so I'm just going to say "There were some issues; open the browser again and try to send it one more time," because uh nothing has happened; it just uh fooled itself thinking it was done. Okay, so we are writing the email again, but what I want to see now is if uh it's going to try to click on attachment and actually put in the file or maybe is it just going to write the full newsletter here now. Uh, let's see; it doesn't really matter too much. Okay, so it didn't even try to put on the attachment, but that's fine; we sent the email. So I'm going to open up that email and take a quick look. And here we have it. So you can see "Global Geopolitics Newsletter"; we sent it from AI agent Manus ProtonMail to my mail. Uh, I wouldn't give this a high score, but you can see kind of messed up here. "As requested, here's the newsletter. I'm sending this a condensed version with the key highlights: You, Ukraine, Russia conflict, Middle East, China, Canadian leadership." Okay, so it actually got that the Prime Minister uh is Mark Carney; that is pretty good. The content was pretty good, to be honest; not too bad, uh, but the format was a bit boring, uh, but I can't complain too much; uh, it was pretty impressive. And what I like to think about is how is this going to be at the end of 2025 and when more people get to work with the models to create cool and very interesting use cases like Manus has done with their general AI agent. Is it perfect? No way. Is it AGI? No. But can you do some cool stuff with it? For sure. And is there room for improvement? Of course. And that is going to be what is exciting. Uh, you can think of this as a good product now; pretty interesting, but uh I'm thinking more how would this be in one year, at the end of the year, and the years to come; that is what is interesting to me. Uh, there's one more thing; you can share this. So I'm just going to quickly show you that. So if you click here and share, we can do public access, right? Copy this link. So if I open this here now, it's a replay. So we can just click play, and then you can kind of see everything it did in like super fast speed. So that is pretty cool; you can kind of relive everything, and you can share this with your friends. I just thought that was pretty interesting; I haven't really seen that before. So yeah, you can see everything what happened in like a detailed preview. So it kind of stores the—or records the full session. So I thought I was pretty interesting too, but again I had a lot of fun; I thought it was interesting, and hope you enjoyed it too. Uh, probably my last video on this for now, and I hope you will get access to it soon so you can try it out. Uh, if there's a lot of demand, I might do another video, but I think that's it for now. Really enjoyed it, so thanks to Manus for making this pretty cool. Uh, again, no sponsored content; just a pretty cool uh software. So yeah, thank you for tuning in; have a great day, and I'll speak again soon.