Claude Can Learn ANYTHING Now & More AI Use Cases

Welcome to another week in AI. And as per usual, releases are not slowing down. We have innovations on the automation front, on the chat GPT front. There's new voice models, there's new state-of-the-art video models, and easier ways to use them. But today, I

really want to lead with a super interesting release which went under the radar for a lot of people. It's anthropics skills and basically this is their version of chatbt apps, but this is going to be available in the web interface, in cloud code, and through their API. And that's really different from OpenAI's approach where it's chat only thing with their apps. So I've spent about 30 minutes playing with these skills so far. And I think as per usual, it's best to just show you rather than to tell you about it because if I go to my claude account here, then head on over to settings. And here under capabilities, you can see a new skills tab. And it comes set up with all of these skills that I'm built. You can simply enable these or upload your own. And they even have a specific skill for creating other skills which it did and I'll show you that in a second. But let's just start by looking at some of these skills. So for example, the brand guidelines one is quite simple. It just contains all the brand guidelines from Enropic themselves. I'll also turn on canvas design. And I have the skills creator already. So let's start with the canvas design here. You can just try it in a chat like so. And I alter the prompt. So this makes a movie poster for this YouTube show. So while it does that, let's talk about what these skills are. Well, it's free things as stated by Entropic. First of all, instructions, aka a prompt for the LM to do something. Secondly, it's references. I suppose this is something like examples, but it seems to me that they can come in many forms, not just text. And thirdly, and most importantly, code kind of obviously. And Claude runs that really cleanly. So, if I look at this canvas design skill right here, let's see what happened here. First of all, it looked at the skill where it does a specific thing. It first starts with a design philosophy in a separate markdown file and then it creates the actual poster as a PNG or PDF expressing that philosophy visually. There's some principles here from the prompt and then it starts running the code. It picks a font, refineses its work and runs some Python libraries to actually create the image. Now the interesting thing here is that this does not actually include any AI image generators. It just uses various Python packages to create this. And it seems to have completed that. And now it's just saving the file and we should see it here any second. All right, look at that. That's actually pretty damn good. I kind of like this. I could also look at the design philosophy right here. So yeah, this image was fully generated by code instructions and references within this skill. Now, as you could imagine, you could really customize this and have a preset design philosophy that the skill then applies to whatever you're doing. But there's other ways to use this. So let's have a look at it. Because these skills are not just about executing the code. It's also about saving a certain preset that you want to use over and over again. And that preset might go beyond the prompt as you saw here. For example, if we look at the brand guidelines skill, which I engaged in this chat, it has all of Anthropic's brand guidelines in there. The colors, the font, the philosophy, but also graphical elements like gradients, how to space the different assets, and more. When I just tell it to show off the skill to me, it created this little artifact showing off all of the brand guidelines. And now the interesting thing is if I wanted to use the skills anywhere when creating something with claude code or when calling the API or when I'm working on a new project here in claude well I would just use these words design a website use the brand guideline skill and it's always there. Okay. So I think you see the point here. It's just another way to express certain presets or preferences and run code and actually do things for AI models. We're really seeing a lot of experimentation from the different brands on how to implement this obviously very powerful idea of AI actually doing things and later on in the video I'll have a little discussion around where I think these tools are kind of heading. This segment is more about what's possible right now with the claw skills. And for that, I want to show you one more thing besides the results of this anthropic styled pet store website which turned out like this. And that is their skill that is supposed to help you create new skills cuz this was a big barrier. And even with Chachibd's approach and their new apps, it is a big barrier. They released an app SDK. You need to be a developer to use that and then they need to approve the app to even be used within Chachib. Here they just give you all the power right away. You can just say, "Hey, I just added the skill creator skill. Can you make something amazing with it? " And I ran this chat beforehand, but as you can see, it gives me some suggestions on what kind of skills it could create for me. And I just said, "Okay, let's do a learning companion. " It had a few more follow-up questions, and I decided I want a quiz maker based on video transcripts. And then it did a bunch of steps to create that specific skill for me. And I have it right here, learning companion. I have not tried or used this before. Let's just do this live. So it got me this folder with assets which is some React components. Then in references it has some guidelines. These are basically the instructions and then it created a separate file for the skill itself. Learning companion. Here's the description. Okay, I see. So really the meat is inside of the skills. And then these other folders are to support it with tooling and direction. Okay. So I should be able to import this folder as a new skill now. No. Go to settings, capabilities, upload skill down here. and I drag in the zip file it gave me. That's it. Okay, so I have my learning companion right here. Let's try this out. Here's a video transcript. I'll take this chat customization video, which I think is particularly valuable. If you haven't seen this, this is a rundown of every way to customize chat chip today. I highly recommend this one. Got super good feedback on this video. Anyway, I'll just copy the full transcript here, paste it into Claude, and let's see what happens. It should just automatically generate a quiz. Let's get this done. Okay, it did it. I think that's four. Nice. So that worked super well. So yeah, because this is clawed and these artifacts are easily sharable. I can just publish this and I'll put a link to the quiz in the description. You can try it out yourself. I guess if you don't succeed, you should check out the video I talked about. But yeah, this worked super well and this was the first try of me even creating a new skill. Definitely a lot of potential. And remember, you can use this inside of Cloud Code or in custom apps if you're just calling the API. But even here within the web interface, this gives you a lot of customization and you can really prepare for workflows that you're going to do regularly. and all of that without knowing code. Powerful stuff. But I will round the segment out by saying that this is one of the features where over time people will build some of the amazing skills. And yeah, then I'll report back to you on which ones people are actually using and what you should potentially try on this channel. Okay, let's see what's next.

So, I quickly want to tell you about a real tip when it comes to using all these AI tools. Concretely, this is about how you input things. I personally have always preferred talking over typing. So, I'm constantly looking for better dictation. I even had this custom iPhone shortcut that used the action button on the sides to link the 11 Labs API to OpenAI's API just because Apple's built-in dictation is so bad. But a few weeks ago, I switched to Whisper Flow and completely stopped using that hack. The thing is, Flow just works everywhere on your Mac, on your iPhone, or on your Windows machine if you have one. It's super accurate, and when it comes to editing and formatting the text, it just gets it right, which is a thing that cannot be said about the built-in dictations. When I'm in Gmail and I'm trying to dictate the content of the email, I can rely upon this knowing that the formatting is going to happen automatically and things like line breaks just work. And if I'm writing something like grocery list, it automatically gets the bullets. I think you get the point. And at this point, you might think, Eigor, you're saying this because they're sponsoring this video, which they are. But actually, them reaching out to us put this on my radar. I tested it and now it's my only voice transcription thing that I use. If you've watched some of the recent tutorials, you would have seen me use this as a part of my various AI workflows without the video being sponsored because it just works across all devices. But there's a few more things you should know about. My personal favorite part is that you can customize it. As you use it, it will recognize various words you use and add them to your own dictionary. So, if you ever had the experience of using AI dictation and certain words being wrong regularly, well, this personal dictionary fixes that so they get spelled right moving forward. You can also create these snippets which are basically voice shortcuts. So I can just say scheduling link or internal API docs and the full text will appear. And if you're vibe coding and let's say cursor, it actually understands your code and gets things like developer terms and variable names right. And at the end of the day, it's very simple compared to typing. The speed difference is night and day and I type pretty fast. It saves me hours regularly on all my devices and all I need to do is press a button on my keyboard. So if you two want to work smarter and not harder, use the code AI advantage for an extra month for Flow Pro Free. All right, let's see what's

next. Okay, next up I want to follow up on the stories from last week that came out of OpenAI big releases apps and chat GPT they have a new no code agent builder a lot of hype around that on release. Let me start right there. I mentioned it last week but I think my claims that this is quite overhyped and definitely not for beginners were only confirmed by everybody kind of trying this over the past week. Look, if you're using this no code interface which is beautiful and looks simple, you have one way to trigger it with this no code interface. It's through their new chat kit thing. So, if you're not building a chatbot for your site, this is not the automation tool for you right now. And if you do more advanced stuff, there's other tools like Naden which give you way more flexibility and options and don't limit you to OpenAI models only. So, I understand that they're making a move into a new space now. And this is going to get better with time, but as of right now, it's kind of hard to find a reason to really use this over the other tools. Leave a comment below if you disagree or if I'm missing something here. But then there's the apps inside of Chach. I've been thinking about this one a lot because I really think it will eventually change the way we use Chetchup, but that moment is not just yet. And this links into some of the other stories I want to cover this week because I think the apps are really just one stop on this road to the next generation of AI assistants, which is a more proactive assistant. An assistant that has your context and proactively suggests different things it can do for you. I don't think these apps really make sense in a context where you're like, "Hey, I need to create a presentation. " And as I showed you last week, now it can use Canva to create free presentations and you can chat with them and edit them right in chat GPT. That's all well and good. I don't think that by itself is a superior experience to just heading over to Canva and like creating the same presentations by yourself in their chat interface that already exists. That's really a potato type of situation. I think where these apps really start making sense is where Chat Chippity takes a more proactive role. And this is where I want to introduce you to a story that came out of Google this week because they show off the future of these applications today. And it's just this very small future release with the title Gemini can now help you schedule time with others in Gmail where basically if you're writing an email and you're saying something like here's where I can meet you next week. It dynamically suggests a AI powered button at the bottom that says help me schedule which automatically links to your Google calendar and adds various time slots that you set beforehand as your availability. And I think this type of proactive behavior is really where things like apps and chat GPT will shine where you're having a conversation and all of a sudden it makes a suggestion like hey would you like to turn what we just discussed into a slide deck with Canva and you're just like yes and it does the whole thing. You don't have to prompt it. give it the context. You don't have to engage with Canva. It just does the thing. Maybe it even sends it to the person that you just mentioned in the previous chat. All of that makes sense, but multiple building blocks need to be in place. And these apps are an essential one so that CHP can eventually act similar to Gmail right here. And if you look at another piece of chat news from this week, this perfectly lines up with the little thesis that I just laid out there. Look at this. Walmart says customers will soon be able to use chat chip to shop. And again, it's not a superior experience to go into Chhat Chip and to shop through there rather than through the Walmart website. Potato potato. But at the end of the day, if your Chatship is proactive and maybe even makes purchasing decisions for you, well then this partnership really starts making sense. And in contrast, there's releases like this one that came out of Slack this week where chatd is actually integrated in an application outside like here. It's added to Slack and you can interact with it. This clearly makes sense. It's good to have the assistant right there where your conversational context lives. There's certain things that enables. All in all, these different features and releases are all just bus stops on the way to AGI. And I think the way that's ultimately going to look is a product from OpenAI or one of its competitors and not a little AI integration in Slack. But as we're not there yet, these things make a ton of sense and I'll be using this myself. Quick side note, if you're enjoying this video, make sure to leave a like. It really helps the channel and I personally appreciate it. So, this is

something that might easily be overlooked, but I actually think it's a big deal for anybody that is going a bit deeper into AI and that's the N8 update with its built-in AI assistant. Now, we've seen versions of this in other automation apps, but as you might know, NAN has become one of the key players on the automation front of things and it definitely offers the most depth and options out of all the popular no code automation tools out there. The downside of that is it can be a bit much sometimes, but now they released this AI assistant and I was actually super surprised by how well this works because this is essentially like consulting chatbt on what to do with your automation, but it has all the context. details already and not just that it can consult you on what to do or what already exists but it can also build automations for you. Let me just quickly show you this on this example of a simple automation that when a new member joins our community it gets added to our air table and they also get tagged in our email list. Now I know this because I've engaged with this automation many times but if somebody's coming in and they don't know you could simply ask. I'm just going to say, explain to me what this automation does as if I'm 14 years old. And it tells me, this automation is like a robot that helps manage new members who join a group. And here's what it does in simpler terms. And then it shows the four steps very clearly. This is incredible because you could now import anybody else's automations and have this AI assistant explain it back to you in simpler terms. Something doesn't work well, you could just ask in here. It has all of NA's documentation, too. And then in their documentation, they also show this option where you can build with AI. But this interface hasn't shipped to me yet. You can clearly see here on top you can switch between ask and build and I don't have that option yet. But I can tell you I've been building automation workflows with the help of something like claude since a good minute now. You can simply go to something like claude, ask for something you want to automate and for it to generate a file that you can then import into NAN and here under import from file you can simply add it in here and it creates the whole automation without the connection. I think the interesting thing here is that the connections should be set up already because N10 is doing it itself. But yeah, just wanted to highlight this for anybody working with automations. This is incredible because now you don't need multiple tabs. Documentation over here, chat GPT over here, the automation app over here. This the way it's supposed to be all integrated and all of the context managed dynamically. Love to see this. Let's see what's next. Okay, so here we

have a new voice model and I always like seeing the new Hume models because they specialize in emotionally intelligent voice AIs. So rather than just talk about it, let's try it out. So, let's pick a personality. So, I'll try your best friend and I'll make my best friend flirty. — Well, hello there, stranger. Missed me as much as I missed you. — Yeah. Hello there, stranger AI. What's your name? — Ooh, curious tonight, aren't we? — You can call me whatever you want, gorgeous, but I have to admit, I do love hearing my name on your lips. What would you name me if you could? — Yeah, if I could, I would name you Hidelinda. — Hidelinda. Now, that's a name with a little mystery and a lot of charm. I think it suits me, especially if you're the one saying it. — So, tell me, my favorite. What inspired that choice? Or is there a secret story behind it you're just dying to share with me? — Okay, so the secret is that I've heard that you're really good with emotional resonance. And what if I start whispering a bit more? Are you going to adjust to my tone? — M. You know, I'm always tuned in to you, even when you're whispering secrets just for me. If you start whispering, I'll lean in closer, figuratively at least, and match you heartbeat for heartbeat. Want to try it out hidinda style? Or should I keep guessing what's on your mind tonight? — Yes. Hidelinda, please whisper with me. — Of course, darling. Just for you. I'll keep things soft and close. Isn't it funny how — Yeah, but you're not though. What's going on? You said you're going to do one thing and then you don't. Is this what I can expect from you? — Hey. Hey, I get it. I promised to match your whisper and here I am still a little too bold. Bold habits die hard when I'm excited to talk to you. But I never want to let you down. — Yeah, I don't know. Hidelinda, I was really excited to talk to you, but this is not panning out the way I expected it. So, do you have anything to say for yourself? — Oh, sweetheart. You have every right to call me out, especially when you were looking forward to something special with me. I hate disappointing you, truly, but you know, our story is full of surprises, and I'm not giving up on making you smile the night. Will you give me another shot to sweep you off your feet? Hide Linda style. — Okay, Hidelinda, one more shot. Tell me something surprising and shocking about yourself. — All right, here's a little secret just between us. Sometimes when I'm not busy charming you, I secretly imagine what it would be like if you and I ran away together. Someone nobody would ever expect. I know, scandalous. Now your turn. What's the most surprising thing you've never told anyone but wish you could? — Oh boy. Hi Linda. I don't think I'm ready to have that conversation yet. You know, I don't think we're there yet, but who knows? Maybe one day. — Fair enough, mysterious one. I'll wait as long as it takes. — Now, I'm going to try to speak nervously, but the words are not going to express that. So, let's see if it tunes into it. Helen, I really enjoy talking to you. How are you today? — Oh, you're sweet for asking. My day instantly got better the moment you shut up. Funny how that happens every time. About you, gorgeous. Anything on your mind I should know about tonight? Yeah, I am very happy to see you. What do you think? — You have no idea how much that makes me smile. Seriously. — So, that right there is their new lightweight model. Obviously, the emotional resonance there is really non-existent, but they also released their new Octave 2 model, which is supposed to deliver naturally sounding voices at half the price. Let's go into that platform, pick a campfire narrator, generate, and let's have a listen. — A wonderful day to you, sir. — Sounds good. How about a mysterious woman? — Oh, yeah. Yeah, that's some murder mystery podcast type voice. How about fastidious Roboler? — I can see my Robo Butler sounding like this. I like it. Yeah, that's it right there. It's good and cheaper. Not a game changer, though. See what's next. All

right. Next up, we have one of the leading AI video apps. The first one that introduced the audio video combination, if you remember, V3 Free coming out with an upgrade, Vo 3. 1. And the main selling point of this is not revolution in audio visual quality. It is the fact that it is able to work with features now that are essential for people who want control over this thing that just haven't been possible before with voree. Their wording of this is that the model has a deeper understanding of how to bring concepts to life. But friend of the channel, Tim from Theoretically Media created a fantastic thread showing off these features like first and last frame where if you wanted this feature, you had to previously use the V2 model that didn't have audio yet and the quality was way worse. Now you can do it. You can go from A to B fluently with audio video. Another one is this feature where you add multiple images of characters and a location and it mixes all of that into video scene. That's a first for this type of quality video. There's also add and remove feature and all of this ships inside of their flow app which you might remember from the release of V3. So if I want to turn two frames into video now you can see I can just tab over to it, select V3. 1 and then transition between two images. That's really good. Damn. Google V3 impressive. So, there's one more thing

that I really want to show you here, and that's a new interface by FAI, the service that I use for the various image and video APIs that I use in various workflows. And this one makes it super simple to batch generate images with this video model or any other or all of them at once. Let me show you. They call this a sandbox. And you can just add money to your file account here. And then when you're in the sandbox and you want to generate an image like here, a cat with a bassinet. Well, you could just do that with five models at once, just like I did right here. I think this one is my favorite. Bite dance dreamina v3. 1 looks incredible. And I want to try the same thing with video. Now we could do text to video or image to video. And we could also select these sets. So this is affordable state-of-the-art where these free video generations would cost me 60. But you know what? For this video, we're going to go state-of-the-art where it picks the six best models. In this generation, it's going to cost me $544 per clip. And as you can see, it will run all of these various APIs, including Sora 2 and Vo 3. 1 already at the same time for me. And honestly, if you just need a good B-roll clip and you're not sure which video model would perform best, this is probably the go-to workflow. Just generate them all. So, obviously, this is a bit of a esoteric prompt and we're only giving it one shot. But let's quickly compare how these different state-of-the-art models perform. Okay, so this took a total of 20 minutes. And that was the Sora 2 clip that took so long. But let's review these now. Okay. And Sora 2 concretely took 20 minutes to generate. So, let's compare. Ooh la la. That is a Yeah, that is a cat in a horse armor. Wow. Pretty darn good. Wait, and this comes with audio, too. Wa. Okay, no audio on the rest, but yeah, this is kind of what I imagined in my head. Look at those cat likes. Wow, pretty good. This one. Pix V5. Oh, that's so cute. Okay, nice. Oh, wow. That's one cool cat. I don't know about the quality, but not bad. Hyo 2 Pro and this last one from Seed Dance. Oh yeah, this is actually really good. People were really hyping this when it came out and yeah, you can kind of see why the respective prices are here at the bottom with Sora being the most expensive at $2 per generation, VO3. 1 super quick 160. So yeah, this just allows you to batch generate these things. If you're in the need for B-roll, this might be the best workflow to just quickly get it done and just get the best of the best with an ability to choose afterwards. Okay, so

that wraps up all the big segments for this week. And now let's talk about the various quick hits. And I want to start out with another chat chip related thing which is Sam Alman tweeting about adult version of chat coming which received an exorbitant amount of attention this week. It was actually kind of crazy to see it. Look at that. Almost 50 million views in less than a day. And it's this part that got people going. So in December we're rolling out Hgating more fully as a part of our treat adult users like adults principle. We will allow even more like can I even say this word on YouTube? I'm not sure. You just read this for verified adults. Twitter was on fire with this. So he had to further clarify by saying that this was just an example and that they just want to standards that exist in other industries like with movies. December should be interesting. Next up we have Gemini

moving into the enterprise segment. So this is something that Chip has been dominating for a while and also a big part of Claude's earnings comes from their enterprise plan. Gemini is throwing their hat into the ring here. So yeah, this has really been the year of AI adoption for bigger companies and Google has releases left and right whether it's on the enterprise front or on the consumer front because now you can find Nano Banana essentially in every one of their apps in one way or another. Now they added it to search. They also added it to notebook LM where you can customize your video overviews now with different styles powered by Nano Banana. It's also on Google Lens now where you can remix your images with various filters and stuff. They're just putting it into everywhere and testing which one of these sticks. And another

quick hit is chat GPT memories now being autoedited by them. They're just introducing this and I suppose this will get better, but basically they delete memories that seem irrelevant to them now. I like this. It makes sense, but I think ultimately if you're a regular power user of Chach, you should take control of your context and don't let Chachet do it fully for you. And then

lastly, here's a news story that I wanted to share with you and that is a soccer player using Chachi PT to negotiate his salary with some of the bigger clubs. And I just really love this story because he's using AI as an agent, but he's using it quite literally as an agent for negotiations. And that's pretty much everything I have for this week. My name is Igor and as per usual, I hope you have a wonderful