Speaker 1: This past weekend for AI has been pretty hectic. DeepSeek R1, as we all know, is a revolutionary open source AI model that launched last week, but its impact on US markets became clear over the weekend and today, we all saw that Nvidia and many other AI companies as well as crypto saw a big dip as there were fears that were growing over China's lead in AI innovation over US. This competition is fantastic for advancing AI, but there will be concerns coming from US as well as their government that may impose restrictions on Chinese AI products like DeepSeek. To prepare potential bans, I recommend downloading Olama's distilled R1 models now. I also covered this in my newsletter, so I definitely recommend you take a look at it because there's a lot of details that I highlight about this new craze for the DeepSeek R1 model. DeepSeek R1 is just truly remarkable. It surpasses GPT-4 Omni as well as Cloud 3.5 SONNET on almost every benchmark. I showcased its coding capabilities with Klein in a video where I built a full stack app that wasn't possible to be built with the combination of having Klein with Cloud 3.5 SONNET, but whereas when I used DeepSeek R1, it was capable of performing greater performance and generation. This truly inspired me to explore its potential with RooCode, which was formerly known as RooKlein. It's a fork of Klein offering enhanced features. For those who are unfamiliar on RooCode, it's an IDE extension and a coding agent that is capable of creating, editing files, executing commands, and even browsing autonomously. It has a key difference between Klein and this is something that we're going to be exploring throughout today's video. Now, you may wonder why use RooCode over something like Klein. Now, I'm not saying it's better than Klein, but some people say there's a lot more advantages associated with RooCode and this is because they have a thing called custom modes. It allows the creation of unlimited modes, which is where each of these modes could be tailored to different prompts, model selections, tool sets, and essentially enables you to define specific roles such as a Q&A engineer or product manager. You have constant updates. This is just something that came in recently, which is where you have code actions. It integrates directly within VS Code's native code action system, so you can provide quicker fixes. You have the ability to have this directly within VS Code, and in general, you essentially have more updates given out to you on a regular basis in comparison to Klein. To get started, it's fairly easy. Go ahead and install your IDE. In this case, you can use VS Code, Cursor, Windsurf, and many others. I'm going to be using VS Code. You're also going to need to have a DeepSeek API. Once you have gotten all those things fulfilled, you can then open up VS Code, go over to the extension store, and go ahead and search up RooCode. Once you find this, you're going to need to go ahead and install this. In this case, I already had it installed, but you're going to need to restart the extension to have the new updates installed. Once that is done, you can then open it up in your left-hand panel, and once you have opened it up, you can go ahead and press on this settings tab. You can go over to the API provider, and then you can go to DeepSeek and paste in your DeepSeek API key, and then obviously, you would want to change it to the DeepSeek Reasoner model so that you can use the R1 model with RooCline. Now, there's a couple of things that you'll notice that is super similar to Klein. Now, if I switch over, you'll see that Klein also has the auto-approve feature that you would see with RooCline. This is exactly the same where you have the ability to read files and directories. You can have it so that it autonomously does your execution for you. It can also retry failed requests, and you also have the ability to manage MCP servers where you can create these different additional tools to have it perform different actions with those tools. Now, if you click on this button over here, this is where you can manage custom modes prompts as well as mode-specific prompts. So, you have three different things. You have code, architect, and ask. So, essentially, these prompts will have Roo perform in a different type of manner. For code, you can see that you are a Roo, a highly skilled software engineer with extensive knowledge in many programming languages, frameworks, design patterns, and best practices. Now, if you are to use this, it's going to be able to use its defined prompt to help you execute code generation. In essence, these modes are going to help you generate your code better for desired prompts. So, I'm going to have it use the code mode, which you can select on the bottom panel, and then you can obviously change your default prompts. I'm going to just leave it as default. But now, let's go ahead and have it generate something. Let's start off by having it create a functional CRM dashboard which shadcn packages. The dashboard should include customer management where you can view, add, edit, and also have it so that you can even delete customers. There's going to be tasks tracking, you have data visualization with charts, and we're going to add a couple of new features as we continuously work with it. So, unfortunately, there seems to be a lot of people using this model at the moment. So, this is why I can't be using the DeepSeq API key at the moment. I'm going to be using the API key that's provided from OpenRouter that uses the same model, but it's going to be a bit more expensive, which is something that you guys might need to do if you really want to use this model right now with RootClient. But you can see right now, this reasoning model is working on creating this plan, and it continuously talks to the different AI agents within the model to help you generate this. So, right now, it is working on generating the CRM dashboard for me, and we're going to watch through as it develops this. So, I spent around $3 with the OpenRouter API key, but what I was capable of generating is pretty cool. Now, the first generation I got from the DeepSeq R1 model is definitely not impressive. It's super basic, and it doesn't look so good. But then, after prompting a bit more, I was able to build out the CRM dashboard where you have your different tasks that you can add. You can see over here, I've added a follow-up call, a send proposal to these two different clients. You have all your customers, you have a sales overview, and an activity metric. But you can see that you also have the ability to add new customers and add tasks. Now, obviously, it would be better if you are to strategically place these different graphs as well as these customers and tasks within this dashboard. But from a couple of generations to get something like this, it's definitely quite impressive with this DeepSeq R1 model. Next up, we're focusing on debugging code snippets. Now, I had provided this Python code that has approximately 11 errors at the moment. RootClient is working with the DeepSeq model, and it is analyzing through each of these different lines of code. Now, if you were to see the previous code snippet that I was working on, it had analyzed all of the different issues that I had found with each line of the code that I had provided. Now, it is working on figuring out a plan to find a solution for each of these different errors. Let's see what it ends up doing. Now, it looks like it is focusing on finding a solution. Right now, it is writing out all the different solutions for the different errors that I had found for each of the lines of code. There will be approximately 11 errors. There we go. We have the new Python file that it was able to generate. It fixed all 11 errors from library imports, which you can see from the start. It had requested on importing the wrong library, whereas with this, it focused on fixing that. There have been fixes to the factorial function, the function call, the syntax error, the division by zero, and so many other errors. It was able to do it quite quickly. Obviously, it's going to be a bit more expensive over here because I'm using the OpenRouter API key. You can see that if you combine it with RootClient, it's going to be able to debug and provide you a good new file that is going to provide you all the solutions in terms of debugging quickly. If you like this video and would love to support the channel, you can consider donating to my channel through the Super Thanks option below, or you can consider joining our private Discord where you can access multiple subscriptions to different AI tools for free on a monthly basis, plus daily AI news and exclusive content, plus a lot more. Now, this is just a short little example of what you could do with RootCode. I keep saying RootClient because I'm used to it, but RootCode is definitely a great extension that has a lot of new features that you wouldn't see with Client. The new feature of code actions plus having different modes is going to essentially help you work and tackle with different sorts of coding-based tasks. I definitely recommend that you try it out, and especially with this new DeepSeek R1 model, you're going to get better generations than you would with Cloud 3.5 Sonic. So I definitely recommend that you take a look at this with all the links that I use in today's video in the description below. Make sure you take a look at the newsletter because this is going to be explaining what had happened over the weekend. Follow me on the Patreon, on Twitter, as well as subscribing to the YouTube channel so that you can stay up to date with the latest AI news. But with that thought guys, have an amazing day, spread positivity, and I'll see you guys fairly shortly. Peace out, fellas.