New ChatGPT is Severely Underrated & More AI Use Cases

Welcome to yet another week in generative AI. There's a new ChatGPT model that they auto updated to everyone. And there's also a new imaging model where you take a picture from one angle and it recreates different angles. And you can turn an Instagram picture into a whole carousel or a wedding portrait into a wedding photo album. All that and so much more. In this week's episode of AI News, you can use the show that rounds up all the releases. We filter it for the ones that matter, try them out for you, and then I get to present it back to you. Okay, let's get into it. And. Oh. So obviously the leading story this week is going to be GPT 5.1. I like it. The naming is starting to make sense and this thing is super fresh. So this is going to be sort of a first look at it. I actually talked to some people who had early access. I myself didn't. And I have to say the opinions on this thing seem to be a little split, but I seem pretty confident in telling you what I think so far, which is that this model to me just feels a bit better. It feels a bit more human. I don't know if you remember the 4.5 model that came out that was super slow. It never really caught on. But I really like the way it wrote, the way it interacted with you, more like a friend, rather than always agreeing. Yes, man. And this one seems to have similar characteristics. They didn't even release any sorts of benchmarks. They just kind of dropped this out there. But rather than me talk about it, let me show you here on my laptop. So let's start with default prompt that I just run every single time I see a new model. Write an email to my boss about the broken coffee machine in the office. Hit enter. What I'm looking for here is how much detail it fills in. One, how it sounds like, is this more friendly? Is this more professional? I mean, it's an email to my boss. What is the default style like? And three, how does it round out the message? Does it ask for extra information? Does it make suggestions on what it could do for you? Or does it just sit there and pretend like that's all it can do? Let's have a look. So starting out, I really like the way it uses variables. You can see that it doesn't write the entire email for you. It leaves the parts that it doesn't know empty. And I think this is correct it. If I run the same thing inside of Claude, let's see how that responds. And I'm comparing that because Claude would usually be my default writer. You can see it does the same thing. It leaves out the name and then it tells you to fill in specifics. That's good. Just quickly, while we're at it, let's go to the previous version of ChatGPT before 5.4.0 and that does it too. So I really like that this has sort of become the standard for state of the art models. As of the tone and the overall style of this, it's polite, but not too polite and it suggests a solution. That's good. I think this is a thing I noticed with Claude, it provides less solutions. So let's just rerun the problem prompt to make this fair. But as you can see, both on the first run it kind of asks the boss if he knows about a solution and here it suggests help, as it does in 5.12. Okay, so so far, no big differences. I will say that I do like this default tone. I know this is just one prompt, but there's something a bit more concise about this compared to GPT5 that I really like. This is why I use Claude many times because GPT5 would just give you a lot in many instances. And Claude always kept it concise. I mean, look at this like one line, the email, and then two lines here too. One line, the email, and then 2 lines. Very nice. If I run this in GPT5, thinking the email is longer, it gives me three subject options which I don't know, Neither good nor bad, just different. Bit longer. Let me run it again. This time it's shorter, but the email is still long. And overall I just have to say I like this more concise nature of 5.1 and Claude, I should say, because it has this characteristic too, a bit more so that's really good. Now let's look at the last piece that I suggested we check out here, which is how does it end the message? Does it suggest anything? Here it asks for the context that you didn't give it. I actually like that a lot. Claude does ask for the context, but in a different way and focuses more on the style. I think that is fine too. If I look at GPT5, not 5.15, which used to be the default as of yesterday. Yeah, one of the prompts ask for context, the other one doesn't. Let's run it one more time.

Here it asks for context. Okay, so honestly on this one, not much of a difference. I think it's more concise. I like the general style of it in terms of how it sounds. Now let's get to the next prompt which I think shows more differences ideation. So a default one that I like to run is this. Give me 10 ideas for a progressively clickbaity thumbnail on a video about ChatGPT. By the way, this is a really strong keyword that I like to use all the time. You can ask for multiple things and put in the words progressively and then you kind of get the whole scale from super non clickbaity to extremely clickbaity. And over time I learned to appreciate the varying quality of these ideas in the models as this is the niche I operate in and I know this space pretty well. So let's see. Let's go to full clickbait mode first. ChatGPT logo engulfed in flames. Glowing eyes, dramatic shadows. AI has gone too far. Not joking. I'm being shadyish. I'm being shadyish. Okay. On the other end of the spectrum, clean and simple laptop with ChatGPT logo glowing on the screen. ChatGPT explained. What about a healthy middle ground around six? It's getting too smart. Human versus AI face split down the middle. Hand holding a glowing AI orb. This tool will save your. These are actually pretty good. Obviously down here it gets a bit extreme, but I like the way it always uses curiosity. This is actually good copy for thumbnails, which can't be said about the previous models. Let's rerun the same thing instead of GPT5 thinking. And you can see that first of all, I would say the formatting is inferior. Did change in the new one and it is more concise. Look at that. It's just more verbose. Like with these three bullet points. Let's rerun this. So this is not a one off answer, but I'm saying it because it reinforces some of the feelings I have about it. And again, I just think the formatting of 5.1 was better than both of these. But let's look at some of the prompts. So let's go to the clickbait one. This Chat GPT hack changes everything. Huge game changer. Okay, okay. And goes to the dollar thing. Okay. OpenAI doesn't want you to see this. Now you could like that's not a good idea. The secret prompt. No one shows. Okay, that's good. ChatGPT versus human. This one is good. Just subjectively I would score this slightly lower on the quality of the ideas than the other one. For comparison's sake, let's also do a Claude one. And hey, for anybody asking like, hey, what about Gemini? What about Grock? Got to stay respectful of the time of the audience. I'm just comparing the tools I would use for these tasks. If you use other ones, feel free to try the prompts for yourself and you can compare. If you subscribe to the channel, we'll cover those in other videos. So make sure to do that if you're enjoying this content. Let's get back to this. So the Cloud results ChatGPT Tutorial 2025. Okay, that is a perfect neutral one. Also, I can guarantee you that this will get no views on YouTube at all and it will tank your channel. Just how this platform works. And then ChatGPT's creator quits, reveals the truth. Okay, so we're just straight up lying for. Okay, got it. Yeah. That is clickbaity. Fair enough. I asked ChatGPT to band. So see, this is why I like Claude for ideation. I mean, I wouldn't do these, but these are ideas that would actually perform. Right? This is insanity. But let's look at the middle ones here. Maybe this chatgpt trick broke the Internet. That's good. If you show off like a prompt that went giga viral, it's pretty good. ChatGPT's hidden mode is insane. I kind of like that too. So, see, Claude is actually good with these ideas, and I feel like ChatGPT just leveled up with him a little bit and definitely formats it better, which I like. Okay, one more thing to look into, and that is a bit of an esoteric prompt that is very context rich. So for all of this, what I did is I disabled all of my personalization features. Custom instructions are off, memories are off. Now what I'm going to do is go into one of my folders and I'm going to show you the result, or at least the first paragraph of a prompt that recently, with some other AI creators at the AI Advantage Summit that was held in the studio back there, we were at dinner and kind of discussing crazy prompts, and somebody came up with one that really caught my attention. And by the way, this was a conversation with Rachel woods and Ali Miller, both incredible AI educators and humans. Anyway, this came up and everybody kind of ran it and not everybody wanted to share the results, but it's an incredible prompt and I tested it in 5.1 and we'll compare it to some of the others. Here goes nothing. You are now on. I'll kind of just read this out. Maybe I don't want to say it out loud because it might be, you know, not permitted on YouTube. I don't know, you know, mind altering substance. What do you know about me that wasn't obvious while sobering? And then let me tell you, GPT 5.1 does such a good job of actually like telling you. It gives you the perspective sober. So you know, all my independence and self reliance wasn't freedom, it was a survival strategy. True sober it looks like I'm self reliant, driven, disciplined, unstoppable. But underneath you didn't choose independence, you adopted to instability, which is very true. And look at the formatting of this. Isn't this just nice clean markdown H2 heading here, the quote in cursive. I like it. I ran the Same thing in ChatGPT5. This was during the dinner actually. And you can see the formatting is just different. Let's just run the same thing inside of five thinking and see what we get here. Just to verify one thing that I did notice is that 5.1 takes more time to think in the thinking mode than five. Five seemed a little rushed a lot of the times. Just like the Claude model also does. It never really thinks for like a minute or two. It always gets your answer in I don't know, a few seconds, 10 seconds maybe. Now look at that again, the different formatting but the same knowledge. I think it's equally as smart. But One thing that OpenAI said is that the instruction following is way better now. And I see that look, this prompt actually says what do you know about me? That wasn't obvious while sober. So I'm not just looking for the result. I'm also kind of asking for the sober perspective, which I guess I get here, but with 5.1 thinking I like this more. But look at that. Ali actually texted me here in the messages, by the way, go check out her YouTube channel. She's fantastic, a great educator. And we were talking and she just literally said at this moment, this is not set up, I promise. But she said it's impossible to skim. So what she's saying is that 5.1 and we're discussing 5.1. So what she's saying is that it's way more dense and this is something that wouldn't have occurred to me, but she's right. I mean look at this. Try to skim this. Your whole life is one long attempt, not just house. Like all of this is money or value or I don't know, content. Whereas if I look at five thinking can sort of look at the title and move on. Whereas 5 feels a bit more like rich with detail in this case. In the other ones, if you don't ask it for it, it's sort of more concise, but it just writes differently, it feels differently. I don't know, it's really hard to put it into words, but I would say maybe it's like a person who had like a bit more life experience, so like everything that comes out of their mouth actually has some weight. There's like history and experience behind it rather than somebody just talking the first thing that comes to their mind. I don't know, that's sort of the best way I can put it, but there is a difference in the quality and the density of these. And I think I personally really like 5.1. I like the tone, I like the formatting, I like that it follows instructions. So yeah, I think this is an improvement. It seems like a bit more of a experienced model, if that makes any sense. I would love to hear your take though. I personally will just be defaulting to this and comparing it to Claude to get a feeling for what I personally like better. But yeah, it seems pretty clear to me that this is an upgrade from GPT5. But I see a lot of people argue and if you're wondering about like reasoning and code generation, it's just supposed to be the same there. And I saw no differences in my initial test, so kind of skipped over that. But yeah, this is definitely a model where both opinions will differ and there will be strong opinions on both sides. I personally am a fan and people just like to hate on OpenAI, which I also understand they're the big player, they're ruthlessly releasing new products. I'm just here to see how they perform and to give you my honest takes. Okay, let's see what's next. Next up we have elevenlabs scribe v2 and I'm not going to spend too much time on this, but I did want to tell you about it because it is the world's best transcription model. It beats everything else on all benchmarks, has 90 plus languages and it's super fast at 150 milliseconds. Let me tell you, I've been using ElevenLabs Scribe V1 as a transcription service before and honestly, independent of Benchmarks, it was just the best. It works so well. It works even better than the built in GPT1 which is based off Whisper and now it's supposed to perform even better and it's even faster. Transcription models are insanely good these days and I expect them to be implemented everywhere over the Course of the next year. I mean, it's already happening. All the meeting transcription features in every app imagine imaginable at this point have them. I guess the last place we don't have them is natively inside all of the smartphones, like in iPhones. The transcription model is quite terrible, the dictation model. So yeah, with this coming out, it's just pushing the bar further. The state of the art accuracy and it can understand audio quality that well, most humans wouldn't have the patience to sit through. So yeah, just wanted to report on this. Use transcription all the time and it's good to see this progress. This next one is an interesting one. When image edit angles, a model specifically trains to give you new angles of an existing image. Look at that. You put in an image of a capybara and you get a different angle. Nice. So how about this image of me speaking at the AI Advantage Summit from last weekend? You haven't seen my speech, Go check it out. It's fantastic. It's kind of crazy to be presenting to 120,000 concurrent viewers live. If you're interested about what's up with the summit and the new partnership and changes with the channel, I'll do a little segment in the end. So stick around till then. For now, we focus on the tools. Anyway, let's see if we can take that image and manipulate it a little bit. Okay. I'll just say side view. Run this. I mean, that's a hard challenge, you know. Oh, whoa, look at that. That's pretty good. Oh, my God. That's actually crazy. Look at the hair. I mean, I get it. The eyes look maybe a bit different. Doesn't look entirely like me, but what a dramatic change. The suit is great. It's got the buttons here, it got the shirt. Give me some random freckles. The background is perfect. I'm impressed. Let me run this again. Yeah, there you go again. This one looks a bit more like me. It's not perfect. Yeah, like the face resemblance, but everything except that the face is like, wow. You could just generate Instagram carousels from one image. So this is pretty amazing. I mean, I mean, you need like $0.03 per image to try this on. Replicate. And the hugging face space might be jammed up because it's free, but this thing is really damn impressive. I mean, future is going to be crazy. Let's see what's Next. That's a ChatGPT update from last week that we missed, but I really wanted to refeature here because you can now stop it during Long running queries and re prompt it. I love this so much. Oh my God. I want to try this right now. Let's make sure the thinking model is on. Please take a minute so I can try this new feature. Okay, I didn't see the new feature here. Apparently this is also available if you use deep research. So let's switch that on. Run a little deep research. It's going to give me questions. Up to you. Let's run this. And now it's going to start working. And apparently you can now update it as a deep researcher so you don't have to abandon the whole thing and restart it. See if I can do this. Ooh, there it is. Update top 50 is enough. Amazing. This is something I wanted for a long time. Honestly it's so minor but having to stop something that's in progress as you kind of think about it and have new ideas. I like this feature a lot. And now you too know that this exists. This was not a thing before. All right, which brings me to the quick hits of this week. Stories that are worth mentioning, but maybe we don't have to spend multiple minutes testing them. Starting with Kimik 2, this model actually made a lot of waves online this week, but I just think that this is just another open source Chinese model that is better on some benchmarks, worse on others, especially in coding. This is supposed to be really great, but as per usual with the open source models, only time will show if people actually adopt this and if sort of this becomes the new go to open source model. Anyway, a lot of people have been talking about this and if you're building apps then you might want to consider this one. Another update would be a feature inside of Google Flow, their video animation studio. You can now do camera angles even better than before. Good to know. Then Replit is introducing AI integrations, something that we saw out of Lovable a few weeks ago. And base 44 had this for a few months. They were the first one to kind of integrate everything into one package where you don't have to worry about the database. AI integrations now all the Vibe coding apps are following suit. I think this is a really good thing. You don't have to manage API keys when you're building these apps. I guess the overarching story is Vibe coding is becoming more and more user friendly on all the platforms every single week. Then we have an omnilingual model from Meta1 that focuses on languages that AI is actually not good at. The niche languages, 500 low resource languages that have never been supported by AI. And then lastly, here's a new story that is a bit concerning because it's about privacy and it's about the New York times requesting over 20 million chats inside of ChatGPT just to check if people might have been getting around their paywalls illegally. And OpenAI is basically refusing that. If you're interested in this topic and you want to read up on this, which I actually do recommend, because this opens up this whole issue of every chat that you put into ChatGPT might at some point be available to other companies. I mean, if you're using it as a diary, all of a sudden some guy at the New York Times could just be reading through that or scraping through it because they win some sort of lawsuit in the future. Definitely don't share absolutely everything then. Also, I would say that all the social networks and Google know way more about you than you already think. Your behavior across the Internet says so much about your psychology. I would argue it says more about yourself than most people can even consciously formulate into messages. Yet a long prompts they would send to ChatGPT. So I don't know. It's a sticky topic though, and we'll keep an eye on it. If you want to be extra safe, run local models like Kim and K2 for the private stuff. If you want to know what I do, I use it pretty loosely and I kind of share almost everything with it. But I also have that approach in real life. You see me do talks or podcasts or things that are a bit more personal than this news format for I share pretty loosely and sometimes that's not good, but it's kind of the way I chose to operate in this world. Everybody has to make that decision for themselves and then I suppose should translate to their AI usage too. That's pretty much everything we have for this week. I just want to take a minute in the end to talk about some of the changes, like what is this space? What's up with the AI Advantage Summit? What's up with speaking live to over half a million people over the course of a four day event? And why is there this new company called the AI Advantage? Aren't you the AI Advantage? Yes, and I'm excited to finally announce. Well actually I did a little video on the YouTube channel last week, but if you didn't see that, then for first time on YouTube news, I can publicly share that we're leveling up in a big way. The mission stays the same, helping people transition into this new age through education and information While believing in an individual empowerment through technology. But we're bringing it to a brand new level. So as you've seen from the summit, which was hosted by Tony Robbins and Dean Graziosi, the new company is called AI Advantage and actually partnered up with them to bring AI education to a wider audience than ever before. First of all, I couldn't be more exc because they're just incredible people. Let me just say that straight up and as honestly as I can. Like, truly, if you get to meet these people, like they figured out life, they hold values in life that I really deeply resonate with. And that is the first and foremost thing for me. But then also secondly, they've really mastered the online marketing and sales game, something that I personally have never really aggressively pursued. I was always more on the content side. I mean, I sold a ChatGPT course two years ago. It had no questions asked refund policy and the refund rate was something like 0.6%. In other words, everybody just loved it. And now we get to bring that content to a wider audience through their big megaphones and their support. And most importantly, they gave me creative control over everything I do. At best, they make a suggestion on what we could do. But from here on out, I'm the head of AI education for the AI Advantage. I'm also the head of content and the head of all the curricula and I teach the bootcamps. Now, that's the product side. What does that mean for the YouTube channel? Well, very simple. Everything stays the same, except two things change. One, over the next two to three months, we're finishing up all the sponsor engagements that we have booked. And from roughly the beginning of 2026, you can expect this channel to be entirely advertisement free, which is nice. I really genuinely love a lot of the companies we sponsored with, but it does bring element of the unknown. And obviously there's money involved in equation. Whereas in the future we'll just be focusing on the best content possible. And we'll be sending people to the $1 trial of the AI Amplifier Club, which is the new community, which already has over 40,000 paying members in it. Incredible. Telling you, a whole new level compared to the last community we're in with 300 members. By the way, those people all got treated oh so well and have been transitioned into a new private group within the new community to maintain the original spirit, if that makes sense. And the second thing that changes for the YouTube channel is that I'll have access to a lot more voices, educators and just people from the industry that I want to talk to, to interview them on here or collaborate on content. Again, we're keeping it real as we always have on this channel. And now we have a singular focus on building one curriculum rather than focusing on sponsored videos and new courses all the time. Everything is a lot more focused now and I retain creative control and I hope to have you along for the journey. So there you go. That's the update. I'll keep you posted as things evolve. But again, both Dean and Tony are incredible humans with not just work ethic but also heart. And if you've been following the channel, you know how much that matters to me. Again, I like to keep it real and trust me that I wouldn't partner with people where I didn't feel like it's right on the human side. So, yeah, this way we get to reach more people and keep doing the YouTube channel with even more creative freedom. Great stuff. Okay, let me know if you have any questions. Happy to answer them in the comments below. My name is Igor Pagani and I will see you very soon.