The New King of AI Coding & More AI Use Cases

So, let's get into it. Um, how do I look? It's okay. — There's a plant in my ear. Oh, no. Welcome back to another AI use case roundup. This week, we're going to be focusing on a Chinese model that presumably is on Claude's level when it comes to building things, but it's fully openly available. We'll also be following up on Chat GPT's Agents release from last week and the Grock release from a few weeks ago, giving you kind of a follow-up review to those really significant releases and much more in this week's episode of AI News You Can Us, the show that rounds up all the AI news stories, filters for the ones that you can actually use today. And then me and the team do our very best to show you practical examples or comparisons to other models and then hopefully you add some of these to your tool belt or at the very least feel inspired to do whatever you do best. Let's begin. Okay, so we need to talk

about this release. It's a Chinese model from Alibaba called Quenfree Coder. This is a highly specialized model for development output. And without overexplaining, let me just tell you that this thing basically matches the best closed source models in the game like Claude for Sonnet and Opus at its coding ability. But this is open, meaning you could download and run it locally, but it's so damn big that you would need a rig that costs tens of thousands of dollars to even do that at a very slow speed. So realistically, you can use it through their website where it's freely available without logging in, chats. ai. Again, this is a Chinese model running on servers located in Singapore. And the second ways you can access it for the API from various providers. And one interesting addition to this release besides hey there's a new Chinese model that's open and really good is the fact that they released their own command line agent tool with it. So you might be familiar with clot code that started this whole product category. Then a few weeks ago we covered Gemini CLI agent which they fully open sourced and now Quen took that and adjusted it from a Gemini agent to a Quenfree coder agent. Now, look, I'm not going to be reintroducing you to this category, but if you're following the space, you might be aware that these CLI agents have kind of taken over the development space. At this point, they're essentially synonymous with the word vibe coding. And by the way, fun fact, if you checked out the AI advantage site recently, you might have noticed that it got a complete overhaul. I coded the entire thing from the ground up with claw code only. It's hosted on Versell. I used nothing else and I didn't edit a single line of code. Everything on the site is vibecoded and it works really well. So, let me tell you, these tools are powerful and you can do way more than that, but it's a good example. Anyway, back to this model. So, some of the details of how they train it are actually really interesting, but we're not going to go into that because this is about the use cases and how you can use this thing to your advantage. If you care about those, I do recommend the fire ship video that was just released on this model. He talks about some of the peculiarities of the training. But here's what matters. I ran the same test prompt that I tried with Gemini's agent and clawed code. This test prompt just lets it build a little pomodoro timer which is simple but it has dynamic elements so we can really see how this performs on a simple interface. And I have to say even from this simple little example I was already impressed because all of the design elements except maybe the padding here are fantastic. The timer arguably works better than the claw one because you immediately see some progress. And as opposed to Gemini, this thing actually looks good and it even came up with this extra feature that is not functional yet with work break that it added on top. So I can already tell you from looking at this one prompt that this is something I would call a proactive model and some of the best coding models are that's what's really good about claude models. You tell them you want A and B and it kind of figures, hey, if he wants A and B, he'll probably want C too and it just does that for you. Sometimes it's a bit over motivated, but when I build things with these tools, I personally would rather have it be a bit more proactive than underdelivering because you can always really back. And this model seems to do that. People all over the internet are hyping it up and really happy with this. And this is not just for developers. I really want to stress the point that a openly available model like this that produces this level of quote quality is massive for this entire space because a bulk of the apps we've been getting recently are AI building apps and they were always powered by Claude Sonnet 4 or Claude Opus 4. Now developers have an option to take this fine-tune it to their liking and self-host these models or just get faster results as these models are half the size. I can also see some larger companies running these internally to guarantee privacy. So, usually I kind of brush over these new model releases, but I think this one is really significant because we didn't have anything for code generation at the quality level of the clawed models yet. And I personally can't wait to see what builders create with these open models. It will take a little bit of time and for you as a consumer, probably the most significant thing is this chat where you can just use it for free when building something. But I think for the AI space as a whole, this is going to have real implications very soon. All right, so next up, I want

to show you one of the easiest and most powerful ways to build AI agents without much experience, and that is with a platform we've featured many times before, Make. And to do that, I want to take a minute to show you their AI agents and the new Make Grid. Because I'm genuinely loving the way they're going with the smart automations. And yes, this segment is sponsored by Make, but this is such a perfect partnership because me and the team actually use Make in our company and to teach all the time. And I likely would have shown you these updates even if they didn't reach out. So, first up, let me show you makes AI agents. What they do is make your automations intelligent. Makes AI agents can adjust workflows instead of just relying on fixed rules like a traditional automation. So, if you use an agent in your flow, it turns a static workflow into a dynamic one with the AI adjusting things as it sees fit. and you giving directions whenever you need. But hold up, how do you actually know if something goes wrong and if your input is needed? Well, that's where make's new grid comes into play. Make grid gives you entirely new ways to view and interact with your automations. You get a real time interactive overview of your automations where you can see how everything connects. You can see where the data flows, quickly identify performance bottlenecks, and even do troubleshooting all in one place. And honestly, this is such a great way of managing the complexity of modern automations because usually all of these things are in different places and to maintain the automations, especially with AI in the middle, you need this info for you to get them working smoothly. It's perfect for monitoring a brand new AI agent, which you just activated. We're actually already preparing a video that is going to show you makes AI agents and grid more in depth. So, keep your eyes out for that for a full workflow breakdown. And if you want to start building your very own automations, sign up to make today using the link at the top of the description. New users will receive one month for free on the pro plan which includes 10,000 operations per month amongst other perks. All right, now on to the next piece of AI news that you can use.

So for the next one, we'll be needing headphones because this is a brand new voice model out of Hume called EVIE. And what the specializes in is emotive voices. So let's actually go in here and clone my very own voice. 30 to 90 seconds of audio up to 30 mbits. So, I just have a little audio clip of me presenting a YouTube video that I'm going to upload to this, and we're going to be comparing it to what I would consider the best-in-class app at this, which is 11 Labs, where I have a customtrained voice. Now, mind you, this took over an hour of audio and took like a full day to process. So, it's not exactly apples to apples, but I still uh yeah, hello. What's on your mind today? — Oh, is this me? Am I talking to myself? So anyway, I was just thinking about how crazy it is that we're still in this phase where everyone's talking about AI taking over everything, but like there are — Hold up, brother. You kind of sound like me, but not entirely. What's up with that? — Huh? Me sound like you? That's interesting. Look, I don't really know what to tell you about that. I mean, maybe we just have similar ways of thinking. — Kind of different. Okay, so let's see what happens when I switch to podcast host. — Hey, welcome to Radio Hume. I'm Eevee 3. Just uh just started up. — What about uh live coach? So, these are obviously different voices. — Hey there. Welcome. So glad you could make it today. — Hello. Welcome to you, too. Good to have you here. — Oh, that's so sweet of you to say. I'm really excited, too. — Who are you? Where do you come from? Where's this accent from? — Oh, right. Good question. So, I'm Eevee 3 and I come from a place of well, let's just say a lot of positive energy and big ideas. — My accent h it's a mix of things. — Yeah, I don't know. That sounded northern European. Okay, so that was okay. Not great. I mean, I can immediately tell you that if I do the same thing with something like 11 Labs, but again, it's not apples to apples. And then when I enhance this with these new features that make it more emotional. Hey there, what is on your mind? What? To be honest, that doesn't sound like me either. So, I think I might have to switch this back to the V2 model for it to sound more similar. Hey there. What is on your mind? That's better. But I have to say in many cases, this used to work better. As we are liberated from our own fear, our presence automatically liberates others. Okay, so there you go. Definitely a strong model, but emotive voices are really getting there. Seeing new releases regularly, yeah, Hume always

seems to deliver. So, next up, we have a really fun one, and I want to comment on the fact that we were actually sponsored by this company recently, but this video is absolutely not sponsored by them. We're just featuring this cuz we generally think this is really cool, and I think you'll see why in a second. This thing is called Mirage. And I suppose this could be referred to as a new product category in AI. LSD standing for live stream diffusion. Nuts. And the team has told me that I just absolutely have to live test this cuz this is just too fun. There is a little queue here. So, I'll just wait through that. And then what should happen is it should take my camera input and we can transform it live. And I think you get like 10 minutes for free. So, let's just start the experience and see what it's got. Yep, it's got the cam. model is warming up. Oh. Oh my god. Me putting on the Okay. Wait. What else we got? Relaxing anime. Sun-kissed. I'm not sure if this is sun-kissed or sunburned. What about zombies? Okay. Let me take off the headphones. Man, what if I only recorded videos like this from here on out? That's kind of amazing. What about this one? Colored manga cabin core. Jesus. Okay. Wait, wait. Oh, okay. Okay. I need to find a good one here. None of these look like me, but that's not the point. Yeah. Cold AF. What happens if I take my little water bottle here and take a sip? That's a weird water bottle. Holy moly. Okay. Dubai skyline. Look at me. I look like a crypto influencer or something. Okay, I have a little book here. Let me open that up. That is pretty insane. This is actually kind of sick. What about the moon landing? Okay, headphones again. Whoa. Gold digger. What is that about? Oh, hello. Okay. And why does my seat have a face? Not anymore. I mean, this is just fun. A cigar lounge. Actually, this is what I kind of feel like when I'm presenting these videos. I mean, maybe not looks wise, but this is the vibe I like. Welcome to another episode of AI news you can use. In today's video, we'll be covering live stream diffusion and more. So, yeah. I mean, what a fun product. No, and I love that sometimes these companies just like reach out to us and want to showcase their applications that, hey, we would be showcasing anyway. I mean, this is incredible. So, yeah, if you want, you can go try this for yourself. And honestly, this is a really fun one to try. If you're enjoying this video, make sure to leave a like. It really helps the channel out. And now, let's get back to the next piece of AI news that you

can use. Next up, I want to talk about chat agent and also Gro 4. Two biggest releases of the last month and how I'm kind of using them, my team is using them, what we're seeing in the community, and what kind of use cases really matter. So, let me start by Grock because that one is rather simple. The model is really impressive. If you do sideby-side comparisons with some of the other leading models, it objectively does a little better sometimes. And I've been doing that for the first week or two, but now I'm at a point where I fully switched back to a combination of ChachiBT and Claude because that slightly better answer is not worth all the compromises. So, the model is really good, but I personally have switched back and I've spoken to many other people, many other power users of all these tools and they've found themselves in the same spot. Now again the model is incredible but there's a little bit of a platform lockin with the most popular LLM platform that most people are experiencing. So now let's talk about

chat GPT agent which has been integrated into chat GBT at this point has been rolled out to all European users and by the end of this week so that's something like the 27th of July even all plus users will have access to this. Now I've been playing with this a whole lot more over the last week and I stand by what I said last week. It performs better than any product in this category that we've seen before. In some cases, marginally, in some significantly better. And this product category is really, really promising. I mean, an agent using your computer and doing online tasks. That just sounds like the topia that a lot of us want to live in, but it's slow and it can still not do many things that you would want it to do. Now, this video is here to show you what you can do, but sometimes also includes pointing out things that you cannot do. And what this can do really well is one research topics and put them into Excel sheets and two PowerPoint presentations. These are the two use case categories that I personally and my team have found to be the most useful and consistently deliver usable results. For example, I took this prompt that friend of the channel Matt Wolf shared recently and I ran this. We've been testing variations of this on various social media channels, but here for my YouTube channel, this worked exceptionally well. I told it to analyze videos over the past three months, put all the data into a spreadsheet, and then to run some analysis on top of it recommending further videos. Things like this work exceptionally well. Look, it mapped out all the videos. It came up with a category system for the videos. It added notes and it created a pie chart for all of it with useful commentary that I didn't even ask for. It gave me strategic recommendations. It gave me actually more than 10 datadriven video ideas, many of which I actually really like. And this sheet is fantastic and accurate. Putting this together myself would have probably took me something like 16 minutes. But in this case, all I needed to do is customize this prompt. So if you have similar tasks where it includes researching something, then analyzing it, putting it into a sheet, I strongly recommend you test out agent. Same thing goes for slideshow use cases. There's this video from OpenAI showing how it pulls together pictures and creates a first draft of a slideshow. Now, here's the thing. If I'm creating slideshows, I'm already using AI. Something like Gamma to create presentations is just powerful. Been using that for over a year. But this can automate some of that process. And I continue to experiment. Sometimes I have these to-do list items that could use a little bit of prep and research and creating Excel sheet like this. I'm trying to think of agent and trying to integrate it into my workflows. But it's not at this point where it's just an endto-end solution for some of your tasks. Again, some of the best use cases I found are research and analysis, not execution. And I just thought that was important to point out. If you find more use cases for the agent, please share them in a comment below. Not just me, but the entire community watching this video would surely love to see those. And at the end of the day, as I always say, the best use cases are always personal to you. They're always individual, just like you are. Okay, next up, let's cover all the

quick hits of this week. Starting out with claw artifacts on mobile. if you still haven't given this feature a shot. This is one of the funnest things I've done in AI over the past few months and it's available for mobile now. You can just build these preset applications. It works on the free plan. You can do it on the go, show other people. Really great interface with fantastic models

obviously. Next up, it's not really a use case, but something you should know about other people using AI. So, OpenAI model won the gold medal at an international mathematical olympiad. This is actually a really interesting story because Google also achieved gold medal level performance. And if you want to learn more about these, I would highly recommend AI explains video on this where he deep dives into these results. We also recommended this as the video of the week on our weekly newsletter if you haven't checked that out. Link is in the description. And

then lastly, there has been a silent but significant update to anybody who's using GPTs in CHP like myself and many community members of ours are still using these regularly. They're powerful tools that were sort of abandoned over time, but now OpenAI is catching up and actually equipping them with features like the brand new model picker. Yes, you can pick the model that the GPT user will be using. So, if you have some basic prompt and context engineering skills, you can set up a GPT for a colleague, prep the model for him, set up the prompt, set up the starting messages, attach relevant documents, and they get a smooth yet powerful experience where you set it up exactly the way it's supposed to be set up. They didn't tweet about this. They didn't mention this, but now you know. So, go ahead and update your GPTs if you have any. And that's pretty much everything I got for this week. I hope you found something that was useful or interesting to you. It's good to be back home in the studio. I have to say the last months were super busy. I was juggling a lot of things at the same time. And the one thing I wish I did a bit more of is tutorials on this YouTube. So, moving forward August, September, you can expect a few more tutorials, practical workflows for free on here. So, I'm looking forward to sharing some of those with you. And with that being said, my name is Igor and I will see you very soon.