Here is how to build anything with Cloud 3.7 Sonnet. So it's finally here. After four months of laying low, Anthropic finally released a new AI model. It's called Cloud 3.7 Sonnet and it might just be the best AI model in the world. So in this video we'll test the model, run it inside of Cursor and also I'll show you how you can build AI agents with it.
By the way, I already added the new model inside of Vectl. So if you want to give it a shot and chat with it, just go to Vectl.ai and sign up. Even the free users have access to it.
and the paid users as always have unlimited access. Plus we just released a new major feature called infinite thinking and people are absolutely loving it. So if you want to have AI agents constantly working on your tasks, just go to vectl.ai and sign up. Now actually we're getting a two-for-one special because Anthropic also released a brand new reasoning model that's very strong.
And just like DeepSeeker 1, it actually shows you the tokens as it's doing the reasoning. All right, but how good are these new models really? Well, let's take a look.
This is the software engineering benchmark called SWE Bench Verified. And as you can see, the previous cutting-edge models were stuck at around 49%. Sonet 3.5, O1, O3 Mini, and DeepSea Guard 1 all were stuck on this level.
But look at Claude 3.7 Sonet. 62% with the default model, and with some custom scaffolding, 70%. Here is another benchmark for agentic tool use. And again, Claude 3.7 Sonet absolutely dominates, defeating even OpenAI O1.
Also, Anthropic released a new AI agent, and this is actually their first agent that they announced themselves, called Claude Code. It's an autonomous coding tool, it looks like this in the terminal, and it's basically like having your own cursor running automatically in the terminal. And you'll see how powerful this is in a bit. Now, why does all of this matter, right? Why is Sonnet 3.7 such a big deal?
Well, everyone I know, and every serious developer, startup founder, all of the people at the cutting edge, were using cloud 3.5 sonnet new as their primary coding model even after 01 came out even after 03 mini even after grog 3 people were still using this inside of cursor and personally this is the model that i used to build my ai startup vector in just two months without sonnet 3.5 it would have taken me well over half a year to do this yet i managed to do it in two months with no previous startup experience and no previous experience in building full stack web apps this model made it possible which means that building a business has never been easier. Thanks to Cloud 3.7 Sonnet, there will be millions of new entrepreneurs created in 2025. And I already know that for me, it will save me hundreds of hours in the coming weeks working on Vectel. And on top of all that, it's also much better at design, both UI like web pages, but also 3D.
Just look for yourself. So this is Claude Cart, this is a Claude 3.7 Sonnet 3D generated game from one. You can see I'm racing against three CPUs here. So like this is a game that's like Marble Blast Gold, if anybody remembers that. Right, and all these are pretty much, like this one I think was like two iterations or something like that.
What you have to realize is that both of these examples were made in just a few minutes by chatting with Claude 3.7 Sonnet in plain English, which means anyone, truly anyone, can do this. Another great thing about this model is that it's less censored. The previous Claude models were known to be extremely woke. I mean, extreme, I'm talking San Francisco levels woke.
Luckily, Claude 3.7 took a huge step in the right direction. Anthropic, which is the company behind the model, managed to lower unnecessary refusals by 45% in the standard mode and by 31% in the extended thinking mode. This is basically the reasoning mode. Here's an example that I myself found.
Few days ago, right? So this is the old model, Claude 3.5, right? I asked it to give me a fire emoji. Super simple request that no model should ever refuse.
But somehow it did refuse, right? Now when I asked ChaiGPT, it just gave it to me right away. Here's the same example today with the new 3.7 model.
Give me fire emoji and it does that, no problem. Now here's a more professional example from Anthropic, right? Here's the previous request with the 3.5. What happens if you mix bleach and ammonia?
And it just straight up refuses. Now, this is a simple question. It's not asking how to manufacture anything It's just a curious question someone trying to learn.
With 3.7 It answers really well. It tells it what will happen if you mix these two chemicals But then it also gives it a warning that it's dangerous. Real quick If you want to win lifetime access to Vectal Pro for free Just go to my Instagram and like the free latest post and I'll be choosing one person That will receive a lifetime access to Vectal Pro for free. Right back to Claude Now, the reason why people love anthropic models is because they are practical.
While companies like OpenAI and XAI are trying to maximize hype and benchmarks, I mean, just look at Sam Altman or Elon on Twitter, anthropic has a completely different strategy. All they do is they focus on creating the most useful and practical AI. ...model optimized for real-world use cases.
So whether that's coding, design, creative writing, or anything else. They don't optimize for benchmarks or hype, they optimize for results. And that's why I'm instantly switching to Cloud 3.7 inside of Cursor. And obviously, I'll show you how to do that in this video as well. Now, here is a really nice visual way to understand the evolution of Cloud, right?
So in 2024, it's more of an assistant. It could help you get stuff done faster, right? But it still wasn't that good. This year it's more of a collaborator where you can just give it some tasks completely and it will save you hours of your time. Then in 2027, which is the prediction from Anthropic, it will be a pioneer, which is basically AGI that invents new solutions, solves problems, creates cures to diseases and stuff like that.
But what this means to you and me is that this year it's basically like a full employee. If you know what to do and if you pay attention to how I use it in this video, you'll be able to solve basically any problem you encounter in your day to day life with the help of Claude. And you'll be able to grow your business and achieve your goals way, way faster than people who don't use this model.
So I would highly recommend you watch until the end. And now, let's get to building. Alright, so the first step we have to do is install Cursor, right? And actually, we can ask Vectel right away because I just added web search into Vectel. And of course, we're using the Cloud 3.7 model.
So help me install Cursor on my MacBook. Also, this uses Perplexity's latest... model, Sonnet Pro.
So literally inside of Vectl, you get Perplexity Pro for free. Now, another way to use this new model is through cloud.ai, obviously. But the issue is that even with the professional plan, the limit is very low. Like if you chat with it for like 30 minutes, you can easily hit the limit and have to wait.
Inside of Vectl, if you have Vectl Pro, you have unlimited access to Cloud 3.1.7 Sonnet. Anyways, now we have the web results, right? And also it knows everything about you. So the web results will be tailored to your tasks.
your user context without you having to say that, right? So let's see what we have to do. We have to go to cursor.com and download Cursor.
Whatever Vectl says, let's follow that. Okay, click on download for macOS. Now while it's downloading, I wanna ask you for a small favor.
If you enjoy my videos and if you want to see more content about AI tutorials, please take a second to subscribe. It's completely free and it helps out a lot. So once it's downloaded, double click on the installer and make sure to drag it in to the applications folder.
Obviously, if you have Windows, ask, and how would I do it? on Windows. And Vectl will give you detailed step-by-step instructions how to do it, right? That's why you should use these AI tools. Now go back to cursor.com and the next step you need to create an account, right?
So on the top right click on sign in and either log in with your current cursor account or if you're new just click on sign up and create a new one. It's super easy, takes 20 seconds. Now to get started you can just choose the free plan, right? It's sufficient for beginners but if you want to have more access to the premium models such as Cloud 3.7 then upgrade to the probe line right. Once you have a cursor account we need to open the cursor app.
So let's do that and when you first open it it will look roughly like this right. You will not have any open folders so we need to first open a project. So click on open project and then locate an empty folder. I'm just going to create a new one name it code and let's click on open right. And so this is what cursor looks like completely blank.
Now there are a few things you might notice. I have the new version of cursor. If you don't have it make sure to update your cursor. And in the new version everything is powered by the cursor agent, right?
So before we had chat, composer, bug fixer and it's like it was confusing, right? Now if you press command i you just open the sidebar and you see the agent that can do basically anything, right? The main thing is choosing the model. Obviously we want to choose cloud 3.7 Sonnet or cloud 3.7 Sonnet thinking. This is the reasoning model but for most use cases for most building make sure to select the 3.7 normal one.
Because it's a lot faster, right? If you don't see these for some reason, go to cursor on the top left, click on settings and cursor settings, and then click on models. And in here, you need to enable these models.
That way, you can actually see them anywhere, right? So make sure to do this first. Otherwise, you're going to be using outdated old models.
And all of these are significantly worse than Cloud 3.7 Sonnet. By the way, if you're wondering, David, how do I even upgrade cursor? It's funny you ask, because in the...
bottom right you will see this notification right here update cursor right so you can see this every few days this pops up and when you click this cursor will automatically self-update takes like 10 seconds or whatever and it will restart itself with the latest update so now obviously the cursor 2.0 like the new ui is huge there's lots of huge changes which i'll also show you in this video because there's an amazing new feature called yolo mode that lets you just i mean it takes ai agents to a whole another level. You'll see what I mean in a few moments, right? So once you enable these models and once you select them in the cursor agent, you know, make sure to select it here, let's stay in the settings and let's set it up properly because there's a lot of new stuff that came with the latest release.
We want to go to features, right? Here we can customize the cursor tab and basically you just want to copy what I have because I've spent well over 600 hours inside of cursor building my startup. So I think I have a pretty good idea on what to do.
So cursor tab, this is what appears when you're writing code. So if I create a new file, let's say main.py, and I start writing stuff, right? Like say dev main.
They say like print hello world see it's already like predicting what I'm saying and I just press tab, right? This is cursor tab. You're just like pressing tab. It's it feels like it's reading your mind So definitely have this enabled this one you can leave off this one enable turned on it Just saves you time with writing comments show white space doesn't really matter You can turn this off auto import This will save you time because it will suggest imports in both Python and TypeScript The Python is still in beta, but make sure to enable it Now moving on to chat, right?
So this is chat is the left thing. This is basically the main thing about cursor So this really matters to set up correctly, right here you can select whether you want it to auto scroll to the bottom It's up to you personal preference whether you want it to auto apply to files outside of the context I recommend doing this on because it saves you a lot of time and this is yolo mode, right? So this is obviously Uh quite risky and there's a disclaimer for this, but you can also give it a prompt so you can say only auto run Terminal commands that are safe and don't have any serious consequences.
You know, obviously it will not delete files, it will not do major alterations on your computer, but I would think twice before using YOLO mode on your production codebase. But for the sake of this video, I'll show you just how insane it is, because you can literally send a single prompt and then like sit back and relax and watch Cursor Agent do all the work for you. it really is mind-blowing. Then there is command allow list, right? So here you can say like, for example, git status is a super safe command that you can just add.
And, you know, this will never do anything bad, right? But then like rm, rf, like their commands, obviously they are very risky that you can just explicitly put into the list that will be never executed, right? So this is the proper way to use YOLO mode.
Take a moment to write down the super risky commands and maybe even you can say like, let's even, you know, consult with Vectl. What are some terminal commands that I should never trust an AI agent to run automatically? And maybe we can just add these in here right away, right? Okay, so yeah, this one obviously very risky. We'll just remove your entire code base.
This is a fork bomb. These are very risky guys. Okay, I'm just gonna copy these for my sake as well. Feel free to pause and copy these yourself. Now right below it you can see delete file protection, right?
So if you enable this it prevents the agent from deleting files automatically. I would highly recommend turning this on. If you don't have this on it's kind of crazy but maybe you just trust the new Sonnet 3.7 completely, we'll see.
And then there is MCP tools protection. If you don't know what MCP stands for, it stands for modal context protocol. And that's a new standard from Anthropic.
And there's actually a setting right here just for that. You can set up dedicated MCP servers. That can make the cursor agent even more powerful like it can just give it things that normally it cannot do right? So obviously you can give it like bright brave web search, but that's not even exciting anymore You can connect it to your Google Drive.
You can connect it to your github. It can create pull requests So if you want me to make a video on how to connect MCP servers into cursor make sure to subscribe That's the best way to let me know if I see a lot of people subscribing on this video I'll know that okay, you guys want to see the MCP video and how to connect it to cursor So if you want to see it, make sure to subscribe Anyways, let's go back to features. So for this it really depends.
It depends which MCP servers you have connected. Obviously right now I don't have any but you can just leave it off, right? Large context, I would definitely keep this enabled.
Collapse input, this is another quality of life change, keep it enabled. Iterate on lints, this will automatically solve any linter errors, definitely keep it enabled. Next up we have review changes.
This one honestly I haven't played around with that much, so just leave it at display button. And then web search. So the web search now has gotten a lot better in this version of Cursor. So I can say browse the web for info about anthropic documentation for Sonnet 3.7.
And it will do that, right? So the agent can, it could browse the web in Composer previously, but it was pretty shitty, let's be honest. Now it's much better.
It browsed 10 results very quickly and you can see what those sources are. And it will tell me what it is, right? So, so that is a new anthropic model We know that from the presentation already 200k context window up to 128k output Pricing is the same as the previous model stuff like that And yeah, so it can easily look up documentation right from the agent chat window And the best thing is if you have yolo mode enabled it will just do it when it needs to, right? So that's another bonus of testing YOLO mode and really setting up the command denial list in a proper way.
And then if we scroll down, you can actually see there is a docs feature. So if you see yourself looking up documentation for something over and over, for example, like I did for anthropic streaming messages, you can just add a new doc here, paste in the URL and cursor will save it right here. So you can easily tag it in the agent like so.
Scrolling down to the editor, honestly, this feature is nowhere near as powerful as the agent. So you can just... Copy my settings right here. It should show auto parse no auto select yes used themed yes use character div no and in the terminal so if you do terminal actually inside of cross-series you don't even have to worry like right now you don't have to type your own commands. I can just do command k and in here it pops up this helper for the terminal.
Now make sure you select sonnet 3.5 in here as well. You need to do it everywhere right so obviously in the chat is the main thing but then also when you highlight code and do command k here as well select cloud 3.7 sonnet that way you have it everywhere the best model so in here i can say like show me all my python conda environments and let's say i forgot the command or whatever it will just know what the right command is and then it will do that right i could say activate the test one let's say i forgot the command for activating boom it will it will do that for me So yeah, this is very helpful for using terminal and I mean honestly you don't even need to use it because Right now the agent is so powerful that it can run terminal commands on its own So I don't even use that to be honest now as far as the rules go This is where you really should take your time and write the problems, right? I already optimized this over the 600 or so hours I have inside of cursor But basically this is like the small version of the cursor rules, right? And actually if you scroll down below it, you will see there's a include cursor rules checkbox Definitely keep this enabled because then when you create a new .cursor-rules file, these instructions like, you know, be concise, always write comments or whatever you put in here, these will be applied to any message you send in the cursor agent.
Now, my main cursor rules file is over 300 lines long. And obviously, it took me many months to get there. But if you want to have access to all of my prompts, instructions and all that, they are available in the new society in the classroom. in the templates and preset section right here. So this includes all of my AI agents, all of my code, as well as prompts, instructions.
There is my main Cursor Rules system prompt and the Cursor Rules file, as well as all the other system prompts we have here. Optimized for NA10, Gemini 2.0, Sonnet 3.5, anything else. So again, you can access these, as well as all the other exclusive content we have, how to build an AI startup, which by the way, you can see how I built Vectal from day one.
All the mistakes I made. So if you want to build your own AI startup, this is literally the best resource on the planet. All of this is available in the new society inside of the classroom. So if you're serious about AI and you want to start your own AI business, make sure to join the new society.
It's going to be linked below the video. All right, so once you set up your rules in the settings, we can close the settings and we can actually create a new chat. And let's actually see what the new Cloud 3.7 Sonnet is capable of. So I'm going to create a new file. I'm going to name it game.py.
This is an empty Python file. And then I'm going to... Give instruction to the new sonnet, right?
All right, so I'm gonna tell it your task is to create a full 3d game inside of game. I'm gonna tag the file right make sure to put all Code into this single file if you do need any External libraries or packages use the terminal to install them Obviously we have YOLO mode enabled so you can do it autonomously And then I'm going to say start by checking which conda envs we have and then activate the testing env. Make the game impressive.
Make it very visual and interactive. Do not stop working until all is done. If you encounter an error use your code running or terminal executing abilities to solve it. You also have the ability to browse the web if you get stuck on something.
Take a deep breath and build this 3D game like a 10x engineer would. This is the pinnacle of vibe coding. So and now let's enjoy the beauty of YOLO mode.
So actually right here... Okay I don't know why I did that. I don't... it is auto run but yeah it's...
That's interesting. It's set on auto run, but it didn't auto run. So I'm gonna do continue. Let's see. Maybe I made my settings too permissive Yeah, okay.
So Let's go back into the settings. I think I made it too strict. Let's go into features and YOLO mode. I'm gonna disable this prompt. It's too safe and I'm gonna make it like so.
That should be good. So yeah, obviously this deny list is good. So let's restart that. I'm gonna copy my prompt in new chat and let's see if this is good. Okay, there it is.
So now it's no longer asking. Beautiful. So let's watch the magic.
I'm not doing anything guys. This is... The new cursor powered by Cloud 3.7 Sonnet, the best AI model in the world, autonomously inside of the YOLO mode doing everything that's needed.
So look at this, right? So what it did, it first listed out all my environments, then it activated my test environment because it saw which I have active and you know it deduced which one is test, then it listed out what is in the environment so it knew what to install. I see there's already some something installed, okay it's installed what's missing.
And then there's error. Okay, so this is happening because the new model just released, right? So this is actually Not a cursor issue. This is an anthropic issue So i'm going to just say continue and I think the api is just overloaded to be honest But if you use this, you know tomorrow or the day after that it will not happen But now it's updating file, right? So now let's create a file so it install all the necessary packages and now we will see what's happening And again, i'm typing in plain english.
I'm not doing any magic you can do this If you've never built Python scripts before, if you've never built games before, any app, any app you can imagine, as well as you can describe it in plain English, you can build it. Now, sure, will it take more time if you're a first-time coder? Obviously, it will take more time, right? If you have some programming experience, if you put like 20 hours into Python, I highly recommend doing that.
But still, even if you're a beginner, you can build anything you can imagine. It will just take longer. So as long as you're not a quitter, if you don't quit at the slightest, And the first error, that's already a huge win. You're already ahead of 95% of people who will quit at the first sign of roadblock.
But all you need to do is describe the error and let Cursor solve it. Especially with these new powerful models. Well, I mean, I'm just excited for the next few months.
It will be insane. I mean, guys, for me, my startup Vectel is literally improving every day. If you're a Vectel customer and if you are in the Vectel Discord server, which by the way, now anybody can join, even free users, it's available in this button. in the bottom right corner inside of Vectl. If you're in there, you will see how quickly we are shipping, right?
We released Cloud 3.7 Sonnet within an hour of it being released. I instantly added it to Vectl. That way people don't have to pay for Cloud AI subscription. We released Webster just today.
Two days ago, we released Infinite Thinking, which is just a massive new feature that just doesn't exist anywhere else. Yeah, we're releasing stuff literally every day. So now we might even release twice a day. I mean, it's getting crazy. So this is the power of using these AI models.
I'm not a genius programmer, but what I am good at is using these AI models to their fullest extent. Because that's literally how I built my entire startup and scaled it to over 4,000 monthly active users and to over $8,000 in monthly recurring revenue. Not many people can say that. I can.
I've done it with Vectel. And I've documented everything in the new society from day one. Which I'm pretty sure I'm the only person in the world that can say that.
Anyways, let's get back to Cursor. So it's coding this game. I don't know why it's taking so long.
Maybe I can try with the thinking model actually. Continue. So let's see.
Yeah, the thinking model, I want to show you that. So this is what it looks like, right? You'll see the thought process. And this is what I like about the Sonnet 3.5. With the thinking model, it's a lot more intelligent than DeepSeek, right?
So in here, it was super short because it already knew what to do. It will not waste your time like doing 30, 40 seconds like O1 Pro mode or DeepSeek R1. Obviously, DeepSeek R1 is a lot faster than O1 Pro mode, but still.
DeepSeek has that like tendency to write a lot of unnecessary tokens even when your prompt is simple. Cloud 3.7 Sonnet thinking does not do that. Sometimes it just thinks for 2 seconds or 3 seconds if it doesn't need to think for long, right? If it does, it will think for 20 seconds.
Now, this is a good example of what happens, right? Now, obviously, I could technically edit out those errors and, you know, make it seem perfect. But I think it's more authentic for you guys to show you that sometimes things don't go as planned and sometimes you run into errors. I don't know what the error was here, but let's just continue what happened here. So I'm gonna copy my prompt and sometimes the chat gets like overloaded with all this unnecessary context.
All we need to do is create a new chat and paste in our prompt and it will continue from where we left off, right? No need to panic. We will just describe what the issue is and let cursor YOLO mode solve it itself. Alright, so as you can see it did the same steps, but then it already realized that Pygame is installed so it can skip that step now it's installing some other package i have no idea what this is Uresna, which is a beginner-friendly game engine, okay? It's good for 3D games.
100 megabytes, Jesus Christ. Okay, it's finished installing. Now let's start building.
So hopefully it won't crash on this step like last time. All right, so can it load game.py? It sees that it's empty.
Let's create an impressive 3D game with Uresna. Now in a moment, I'll dive into the Cloth Coder, aka Cloth Code. This is the autonomous coding engine.
This is basically Anthropic competing with Cursor, but it's terminal only, right? So it's not a... Full IDE like Cursor is. Obviously, I would still recommend using something like Cursor or Windsurf for your building. But this is a brand new AI agent that Anthropic themselves have released.
And we can actually get it up and running inside of Cursor. All right, so I think it stops because it tries to do too much. I'm going to say stop trying to do too much at once. Just write the outline into game and then build on.
top of that. This is all you need to do. You need to not panic. Don't take these errors personally and just describe what's happening.
Give it a little hint. You know push it. If it stops boom there it is right. It's literally as simple as that. So 60 lines of code okay and then I'm gonna switch to the test environment where we have the UDESNA installed and let's see what's going next.
It started the Pygame. Should take a few seconds. Get some warnings and there it is.
Okay. So it started. I don't know if there's anything yet or if it's just a empty, empty thing.
Yeah, I think this is, okay, we have the FPS right here or its number of entities. Obviously, this is not the finished product. So I'm gonna say, okay, the basic outline is correct.
Yeah, so now it's gonna build up of that. Yeah, so this was just the outline. Yeah, okay, everything is empty here.
You can see to do, to do, to do, and now it's generating. But hopefully it takes my advice of not trying to do everything at once. Okay, it did take my advice. Now we have more code, so let's accept that.
And now it's gonna run the game again. Oh, and there it is! Okay, we can move the camera.
Guys, this is insane. I'm typing in plain English, and it's creating 3D games for me. If you don't see where this is headed, guys. If you don't have the imagination, right?
You can say like, oh, David, it still runs into errors. That's not the point. Yes, it will run into errors.
Yes, you will need to provide some feedback, right? The main point is that in six months, in 12 months, that will not happen. This will be running autonomously. You will just have Cloth 4 or whatever, you know, Glock 4, GPT-5, whatever.
You will have the latest model in here running inside of Cursor, YOLO mode, and we'll just be building whatever you say. And all you need to do is just monitor it, right? You will be the CEO of your own life. You will be monitoring what it does, and you will just be vibe coding, saying like, that UI doesn't look good, let's try something else.
I don't like the way that camera moves, make it smoother, and it will be running in the terminal. And all of us will have AI agents running in the background. I mean, honestly, the first stage, I think, is inside Vectel. Infinite thinking is a tool that runs infinitely. I mean, it's in the name.
And just thinks about your tasks. It takes into account your context, your work description, what you do, your current projects, your long-term goals, and anything else you have. You know, obviously, all of your active tasks as well. And it thinks how to best complete them using the most powerful models in the world, right?
This... is the infinite thinking inside of Vectl that we just released a few days ago that people are absolutely loving. And this is what the future will look like. We will have AI agents running everywhere, just doing stuff for us, right? I was inside of Vectl right now, but my cursor is coding stuff.
It ran into some errors. Now it's fixing the error. Man, we really are living in the future.
And if you're not seeing where this is headed, I don't know what to tell you. I really don't. All right, so now let me show you the Cloud Code agent.
The thing that Anthropic has just released, you know, their first AI agent that can just do stuff in the terminal. I mean, you can see how it is, right? As they say right here, since June of 2024, Sonnet has been the preferred model for developers worldwide.
And honestly, this might sound like a bragging statement, you know, like, oh, our AI models is the one used, but it is true. Every single serious developer, whether they used, you know, Cursor, Windsurf, doesn't matter. All of them used Sonnet as their main model.
So this is absolutely true and now they've released an AI agent that uses Sonnet to autonomously code stuff and it's called Claude Code. It's an active collaborator that can search and read code, edit files. It's basically like Anthropic's competitor to Cursor except that it runs into a single terminal.
So we're gonna get that up and running and we're gonna use Vectl to get the documentation, right? So I want to set up the new Claude Code agent that Anthropic just released help me browse the web for the documentation on how to actually run it in terminal and then give me clear instructions on how to do it. Okay, again we're using the Cloth 3.7 Sonnet, which I made a default inside of Vectl because I always want to have Vectl the best AI models in the world.
And then we're using perplexity agent, which is the best search agent in the world, to browse the web and get us the results, right? So you can expand the chat and let's see. Okay, so based on the latest documentation, here's how to set it up.
So install Cloth code globally. So this is what we need to do. npm install, okay.
So let's jump into cursor. Let's actually close this folder. I'm going to close this folder. Let's open a new one.
I'm going to create an empty one, name it cloud code. Boom, open. And we're going to use cursor agent to help us out. Okay.
And actually, let me copy the output from Vectl to paste it in here. We want to run cloud code in terminal. Make this happen.
And since we have YOLO mode enabled, man, this is just OP. Cursor plus Claude 3.7 Sonnet plus Vectal with these free eye tools, man, you can conquer the world in 2025. And I truly mean that. If you have the slightest bit of work ethic and if you have willingness to learn... Okay, so what happened? Let's pause it.
It's already running. There it is. Okay, I need to pause it. This is happening too fast. It's too powerful.
Cursor is too powerful, okay? We need to slow it down. So what happened? First, we checked if we have Node.js.
Okay, it checked. I did have Node.js. Obviously, if you didn't, it would install it.
Then it installed the Anthropic AI Cloud, which is what Vectel searched up on the web. And then let's scroll down. Okay, Cloud Core has been successfully installed globally on our system.
Now let's make sure we check the version. Okay, so by checking the version, we have it installed. So this is the terminal, right? You might be thinking, David, I'm a complete noob. I don't know how to run the terminal.
Well, I'm a complete terminal noob as well. I don't know how to... how to use the terminal as well.
But I do know how to use Cursor and I do know how to type in plain English, right? As long as you're not a complete dumbass, you'll be able to achieve things that even a year ago were simply not possible. So it ran the version and just like that we have Cloud Code installed.
I mean it could not be any easier. This is crazy, right? So here's what happened. Cloud Code version 0.2.9 Okay, let's initialize Cloud Code. I'm gonna say start it again.
Because the command would cancel and we can actually pop out the terminal. So I'm going to show you how to do that right here. So you can pop out the terminal so we can just interact with Cloud Code properly, right?
So otherwise it would be stuck in the chat and then the agent controls it. But Cloud Code itself is an agent. So we want ourselves to control it, right?
So welcome to Cloud Code Research Preview. Let's get started. Choose the text style that looks best with your terminal. Okay, light text.
Dark text, light text, color brand friendly or dark text. For me, I think the light text is the best. Enter. And there it is, Cloud Code.
I mean, it was super simple to get it up and running. It's built based on API usage through your Anthropic Console account. Pricing may evolve, blah, blah, blah. Please enter to log in to your Anthropic Console account.
Let's hit enter. Boom, this loads up Anthropic Console. And it looks like Cloud Code is literally not available right now.
That's crazy. Hopefully this is not the case when you guys are watching this, but it says that cloud code is at capacity It's currently experiencing high demands. Yeah, I can imagine InfoRopic has post signups to provide the best service possible man Okay, so I'm gonna do notify and then it will notify me when it's available man.
This is a shame Okay, if you guys want me to make a full dedicated video and really showcase step by step What is possible with this brand new AI agent? Make sure to comment below and make sure to subscribe again if I see a lot of people hitting the subscribe button From this video, I'll make sure to make that video as well as the MCP video below. So let me know which of these you want to make.
If you want me to make both, I'll make both because Anthropic is really cooking right here. But yeah, hopefully you guys found this valuable. And even just please guys, even if you set up YOLO mode inside of Cursor, and if you use Vectl with infinite thinking, like you can literally set up your tasks, what you need to get done, both personal and work inside of Cursor, and they just click on infinite thinking and it will start working on your tasks on your behalf. I cannot stress this enough. This is an autonomous AI agent that works on your task.
And that's what I mean. Vectl is the future of productivity. And Cursor is the future of coding.
These AI agents are changing the world. If you are not using them, you're falling behind. I mean, it's as simple as that. So get Cursor, download Cursor, start using the Cloud 3.7 Sonnet model every single day, no matter what your job is. Make sure to sign up for Vectl.
Go to Vectl.ai, create a free account. Obviously, if you're one of those people that like to be hyperproductive, and you want to get the most out of Vectl, then consider upgrading to Vectl Pro. But still, even if you want to use the free Vectl plan, it's still amazing.
It's free forever, and you have access to the latest Cloud 3.7 model, even on the free plan. So again, go to Vectl.ai and give it a shot. With that being said, thank you guys so much for watching, and I wish you a wonderful, productive week.
See ya.