OpenAI's Releases "Intelligent UI" and Anthropic's New Model

Welcome to the podcast today. I apologize. It's been a week since I have recorded an episode and that is because I have probably got the bubonic plague from Russia. Probably not, but some horrible cold. So we're on the upswing right now. We're getting back into it and we have an awesome lineup today. The first thing is that OpenAI has added something called Intelligent UI to GPT -6. This is actually very useful and they have a great demo for this. So we'll be getting into that.

If you're following along on Apple Podcasts, you'll be able to see the video for that or on YouTube or wherever else. We also are going to be talking about Anthropix. Claude 5 .5 model that just dropped. Specifically, that is going to be their Haiku 5 .5. Then we have Microsoft that has just released a new NVIDIA chip AI PC.

This is going to be in the new revamped Windows 11. And I think there's a lot of exciting stuff happening in hardware. Speaking of hardware, there's a company called that the former CEO or founder of Cal AI just raised $10 million for his 19 years old. We'll get into that. We also have Manus, who just raised over $500 million.

This is their first funding round since they got split up with Meta. I have some probably controversial takes on that whole situation. And then Vesta has just raised $30 million to help with use AI for mortgage lenders using agents. There's also controversial takes on that. So we're going to get into all of it today. Before we do, I wanted to mention that AI Box has just revamped our playground feature. If you've used AI Box and plot. in the past. That's my own startup.

We've just added a whole bunch of new features to our playground where you're able to do a handful of different things. If you've listened for a while, you know that we have something called the AI Box Builder. That's kind of where you link together different AI models to build tools. And they have something called the Playground where you can talk to a handful of different AI models, over a hundred of them, and you can compare them side by side. We have rebuilt the entire playground from the ground up. You could do a couple of cool things. One is in the playground.

We basically just have one chat interface and you can talk to that. That chat interface can go and build tools. You can also reference tools you've already built. We call them boxes, but you could say, hey, I built a box that helps me to, you know, generate podcast scripts for XYZ thing. Go use that. It will go find whatever box you built and it will be able to use that inside of AI box. You can also say, hey, I want to, you can add favorite models and. I think you can add groups as well. That's a new feature we just added. So you can, what I've done basically is I've made a group with ChatGPT, Claude, Gemini, and Grok. And anytime I ask a question, instead of just going to one of them, I ask it to the

whole group and it generates a response from all four at once. In the past, we had the ability to compare the results side by side, but you had to basically regenerate one question at a time, which was slow. So now you can make a group and you just ask your question to the group. And you can pick the best response. You can kind of see some of the differences between all of them. So if you want to go check out the brand new launch of AI Box and all of the new features in the Playground and the Builder and a lot more exciting stuff coming soon, go check out AIBox .ai. I'll leave a link in the description. It's only $8 .99 a month to get started. So it's super cheap. You get access to over 100 AI models.

You get image, audio, video. If you want to try making really cool videos, if you want to try making amazing images with a bunch of different image generators, go check it out. It's the best way, I think, to try all of the AI models and get the best responses and find out where there's any gaps or holes in an AI model that you might be using. You know, Claude is awesome, doesn't generate images. So you get access to all of that on here. Okay, let's talk about what OpenAI has just released here. So they had a tweet that came out and they said, GPT -6 and Intelligent UI now rolling out in ChatGPT. For everyone, Intelligent UI and ChatGPT delivers fast, interactive answers that make everyday questions more visual,

complex topics easier to grasp, and interactive tools for your task available on the spot. So basically, they had actually a pretty funny, they had a pretty good like... video that came out with this so i actually love the the demo videos opening i came out with they had a guy and he was like trying to paint his house and all of the little paint swatches that he had were just white and they were just text explaining what the paint color looked like there was someone asking about a city and how to get somewhere and it was like a ginormous map that they had opened up and it was instead of a map they were just like text explaining where everything was or someone trying to put together a bicycle and

they just had like a book with text in it Anyways, the point of their whole little ad is like sometimes text isn't enough. And I love that. I think that's very accurate. And then you go on to show basically demos of what the new quote unquote intelligent UI is, which basically it's it has, you know, you can ask it questions about something and it will show you the color that, you know, if you give like a color code that you want to paint your wall, it will show you what that color swatch would actually look like. If you ask it how to get somewhere, it will have this interactive map. And apparently what it does is it says with intelligent UI, GPT -6 can now compose responses using text, visuals, and interactive elements,

choosing how they fit together based on your questions. Responses can include graphics and charts to help explain an idea along with tappable buttons, forms, and interactive experiences you can use directly in your conversation. This reminds me of something that Google kind of announced maybe like a year ago. I have no idea whatever happened to it, but it was one of those Google demo days where they said like basically inside of Google Chrome, you're going to ask Google a question. And if they thought it was a good opportunity, they would generate like a mini app that would help you understand the question.

If you're like, hey, where is Mars and the solar system would generate like an interactive solar system with all the planets on their axis is like moving and would have like a big bubble like this is Mars and give me some information about it. And I think my. My response to Google thing was like cool, but like maybe that's a little overkill. Maybe that's like a lot of compute and maybe you might generate a whole quote unquote interactive game that might take a lot of time, take a lot of compute that maybe someone doesn't really need or want. So I think threading the needle on like maybe not over engineering a response, but at the same time, these are really incredibly useful in a lot of cases. So they had, you know,

cases where you're asking it, how do I fold an origami crane? And then it's like showing you the step -by -step folds to make. It's super useful. Or they had an example of someone that had a big box full of parts and they're putting a bike together for their kid. And it was showing like an interactive image of like the bike and showing you where to put each piece on. As you put a piece on, it would like, it would, you know, glow to the right color and it would tell you the next piece to put on. So I think there's so many incredible use cases that they've gone through. But basically, the point of this is that ChatGPT is going to be more useful and more visual. when it comes to putting things together.

And I love this because this is not like we had a big model upgrade per se in order to get this. It's not like they've reinvented anything in particular. It's just that all of a sudden, the model that we use, they've added tools to make it more useful. And I love all of this. Anthropic has just launched the new Haiku 5 .5. And there's a bunch of different things that are going on with this. But basically, this is a model that they built. for things like, so it's not their frontier model. It's not the best model. You know, they have kind of like a tale of, I don't know if you want to call them worse models, but maybe more efficient models is a more generous way to say it. But this model in particular is

for things like classification, summarizing stuff, extracting stuff. Apparently you can use it for live support. Live support. And I know people are like, why would you want to use a bad model for support? Number one, people get millions of support requests. But number two, these more efficient models are also very fast. So when you have live support, it can answer people really quickly. And if it's pretty good at the answer, then these things work good. Apparently it's good for voice agents and in -app assistance.

It costs 75 % less than Haiku 4. It's interesting though. And I'm like, I hate talking about the cost decreases here a lot because basically what happens is they're like, They're like, okay, Haiku 4 .5. It's like super, I don't know, it's super awesome. And then you have like Haiku 5 .5. It's 70 % less. And they're gonna have a new model that comes out and it might like double in price or increase in price. And then they don't say anything about it. And then one after that, it like decreases the cost. They're like, it's 75 % less expensive than the last. So the price kind of fluctuates. It goes up and down.

The efficiencies go up and down, whatever. So I'm not super hyped about the price in a sense because there's just gonna be a new model. We don't know exactly where the new model is gonna be. Okay, this new model is priced at 10 cents per million input tokens and 50 cents per million output tokens for prompts under 100 ,000 tokens.

It gets more expensive if you have longer prompts. They said that this is the first version of Haiku with built -in safeguards. So it's not like super secure because these are kind of their lower tier models, but they said it has a narrow set of high -risk cybersecurity requests that it has safeguards for and that it... It says that for most of your everyday tasks, they're not going to be impacted. Now, the reason why they said that is because there is a chance when they add new safeguards to a model like this, that all of a sudden it's going to have false positives, basically, which is, if I'm being honest, one of the biggest downsides I've seen with Claude is that I get a lot of security guardrail false

positives where like it won't let me do something. I'll be like, hey, I need to go and, you know, I want to do like a big analysis on all of my podcasts, for example, as a project. And I'll be like, hey. go to my RSS feed, download all my podcasts, transcribe them all. And then tell me about like how much I say this or how much I say that or what are the topics? What is my tone? What is my like, how do I talk about things, right? Like I want to do like, let's say I want to do an analysis program. So I'll say that and it'll be like, I can't go and download that podcast because I don't know if you're actually the one that owns it. Or it'll be like, okay, yeah, I know that you're Jaden Schaefer. So I know that

you own the podcast. I'll do it. But then I'll be like, okay, awesome. I want to do it for my buddy's podcast because he was asking me, like I own a podcast studio, for example. And so like, there's a lot of clients there. And so maybe one of my clients heard about this cool analysis I did and they're like, hey, can you run an analysis for me? So I'm like, sweet. go to this like person's podcast and download it and do the same thing.

It's like, oh, I can't do that because of like copyright reasons. I can't download the podcast. I'm like, anyone in the world can just download a podcast and get the transcripts. This isn't like private information, whatever. OK, so obviously, I'm obviously I'm annoyed about this one situation. I've had like 10 things that Claude has done this to me for where it's just like way overly.

And then I end up having to like try to like trick it or be like, actually, my name is. you know, Jeff Smith, and that is my podcast. And then it's like, I don't know, the name on your account isn't Jeff Smith. Like, are you sure? Can you send me some verifying? Like, it's just ridiculous. So I hate feeling like I'm getting nannied by my AI model. And so the reason I bring this up is because I worry when they say they added new safeguards in, they're like, look, most everyday tasks are going to be impacted.

It's like, so some everyday tasks are going to be impacted. That's so annoying. But whatever, I get that's Anthropic's entire shtick. Okay, let's talk about what's going on with hardware. Microsoft has just released a new NVIDIA chip that is for AI PCs. And so we know that there's been a lot of work that Microsoft has been doing in revamping Windows 11. And there's a lot of hardware that has come out. So there was a big tech thing that just happened in San Francisco, their big tech week and Microsoft revealed that they have the new specs and prices for their new Surface Laptop Ultra. And this particular computer has two base models. There's one that's $2 ,600, one that's $3 ,700. And I think there's

like, you know, you get more memory, more storage, it gets up to like six grand basically for the higher end. Although apparently Microsoft said that's out of stock, which is, I mean, if I'm being honest, pretty annoying. I remember in the olden days, in the olden days, long, long ago, when a new iPhone would come out or any new piece of tech, right? Like when the new...

I remember I was a huge Surface fan when the Surface first came out. And I remember like when they do the product launches, they're like, we're completely sold out in like 12 hours. In my opinion, being sold out on it, and they kind of build hype around that, being sold out in 12 hours on a hardware launch or tech launch is not good PR. That means you did poor planning. You did poor. planning of how much people would want it. The tool, you didn't do your supply chain measurements correctly. I don't know. But just like, why would you sell out on the first day?

That's not a cool thing. Anyways, it used to be a really cool thing. And like PR, like our iPhone is so popular, it's sold out. It's like at this point, you're a big company, Microsoft. If you make a piece of hardware, you think a lot of people would buy, just make it so it's in stock. So that's me. That's me crap talking Microsoft a little bit. But at the end of the day, I think this is a really cool piece of hardware.

And I love this concept. I mean, if they can keep it on the shelves, people can buy it. And that is that I think there's going to be a lot of people that are going to want to run your own AI models on your own device. You're not going to have to pay for them. This is cool. And especially, this is cool when we start talking about all these agents, Muse, Cloud Cowork, ChatGPT work, all of these tools that run on your computer and can control your computer. Well, guess what?

Next time I'm like, hey, you know, ChatGPT, I recently did a project where I was using Cloud. And I was like, hey, Cloud, I need to do a project. I got to generate like 600 images. for like something and they ought to be like these different specs and blah, blah, blah. And of course, Clyde doesn't even have image generation. So I had to use the AI box MCP attachment, which by the way, it's pretty awesome. If you haven't used it already, basically you can use AI box. So in a hundred different AI models, any AI model can be used inside of all of these AI assistants that we use. And even if like, you know, you're like, oh, I got chat GPT and it does like image, but it doesn't do video. Well, you get Google VO3

generated inside of chat GPT with the AI box MCP. But anyways, let's do this project, generate a bunch of images. And I had to, you know, obviously pay for the usage on all of them. Now, imagine if I had a computer that could generate images, I would just, I would just say, hey, or even if I was using a chat GPT, but I was like running low on tokens, instead of burning all my tokens to generate a bunch of images, I could have an image on open source image generation model on my computer and could just say, hey, like, you know, chat GPT work, go use this program to generate the images and go do the project. And I could save a ton of tokens. by using open source stuff on my computer. I think that's going to be very

popular in the future. We're going to see a lot more of that. You'll be able to outsource stuff. So you're not just burning tokens just to burn tokens. Like if your computer can generate it for free, might as well do that. Speaking of other hardware, the former founder of CalAI, he's 19 years old. He just raised $10 million for a new startup. And if you don't know their backstory, he sold to MyFitnessPal because CalAI blew up. It's like a calorie tracking AI tool.

And they pulled in over $30 million in annual revenue in under two years. Okay, what's his new project? His new project, he said, is a personal AI assistant like Meta's Muse. Okay, shocker. And this is the most like viral, trendy thing. So getting into that isn't, I think, what's interesting about him. What's interesting about him is that he's doing it via a wearable. So basically, you have a bracelet that you can tap and talk to.

And it's like Meta's Muse. Now, I actually am not a hater on this at all. I think this is probably a great use case. for a couple of different reasons. And I don't think this is, I think we'll actually probably see some of this from the big labs too. But basically, if you can imagine, especially if this is something that can attach to your computer back home. So like, let's say you leave and you have to go to, you know, go out to lunch and you're like, shoot, I need to get this report done for like a meeting I have later. If you're able to just tap your watch and say like, hey, like access my computer and go like build this thing. And here's my folders and files are in these places. Go get the stuff done. I think

something that could like remote. get things done for you, but it's, you know, wearable. I think that's kind of cool. Now, a lot of the parallels that are being drawn to this particular piece of hardware is, you know, people are like, this makes, in TechCrunch, there's an article they say, that makes personal somewhat a persona, that's his company persona, somewhat more like B, the wearable AI device that Amazon acquired in 2025. For those of you that don't know, I actually spared all of you and bought one of those B AI wearable devices after Amazon bought them because I was like, what are they buying?

Like, what is this company? And full disclosure, it is a plastic bracelet, kind of like the one that's in this. And it is the absolute worst piece of hardware I've ever used or seen. It's basically a little microphone on there and it's got an app that like it records everything and then it transcribes it for you. So it's like a transcription bracelet app thing. And it's allegedly quote unquote AI because I know everyone just uses AI for transcription like whisper.

So it's really not an I wouldn't like classify this as an AI company. This is just a transcription bracelet. And you have to like manually start the transcription. So it's not it's basically like you could just have a note taker app on your iPhone that does the same thing. It does nothing unique or cool. It misses a lot of words.

It's really low quality. It's the cheapest piece of plastic bracelet you ever saw. So anyways, all that to say, B, I was very unimpressed with their product and their device. We almost actually made a whole review video on it, but I just decided it was like too bad of a product to even make a review video on. So there you go. Hopefully Persona is a better product and I'll be, you know, I think there's a lot of space and I think Amazon probably bought it with. knowing that it could do something like this. But if Amazon bought B and he's getting into this, I mean, assuming Amazon doesn't just let it die because it's a useless piece of garbage and they actually try to revamp it and make it useful, then I think

he'll have a good competitor. And I think this is a good space that there'll be some, you know, there'll be some good competition and some good products people actually use out of it. Okay, let's talk about Manus's China or Manus AI and their new round of funding that's $500 million. This is the first round of funding they've been able to do since they got split off from Meta.

If you don't know the backstory, they had like a viral video that like a viral demo. They're kind of like an agent like Meta's Muse, which actually I think is kind of interesting because I think that Muse basically was like, it's just a clone of Manus. And so after they acquired the company for $2 billion, it went viral. They moved the whole team from China to Singapore and then sold it to Facebook for $2 billion.

And then China said, hey, you're not allowed to do that because, you know, it's a... You're just outsourcing, like you're getting rid of Chinese tech jobs or whatever. So they made Meta unwind it. Meta originally said like, hey, we legally bought this. We're going to like fight this in court, but good luck fighting China in court. Like that's ever going to do anything.

I think they were holding the CTO and the CEO hostage in China. So like, anyways, it all got unwound. And I think at the end of the day, Meta probably realized that's fine to unwind this. They already had the IP. They already had. Well, I mean, DIP, but like, I think they learned everything they needed to learn from Manus. They figured out what they were doing. They unwound it, gave it back. And then, you know, a couple months later released Meta's Muse that went super viral.

And I mean, I use it to triage all of my emails every day. It goes through like 11 different email accounts for me, writes drafts, sorts, deletes, does like a whole bunch of stuff. It's completely free. It's amazing. I really should get sponsored by Meta's Muse because I've hated, I mean, not hated, that's a strong word. I haven't been a huge fan of Meta for like basically the last 10 years. And now I think they did a great product. They did a good job. They made it free for everyone. Super valuable. So I think, you know, that's where I'm at on it. Okay. So where does that leave Manus though? A lot of people were like, oh man, like China just completely destroyed them and they're going to be down in

the dumps. They got unwound. We got sent back, but it looks like they were able to raise $500 million. And I'm also going to say. When you see a big, you know, quote unquote national deal like this, where it's like literally the government of China is unwinding it and forcing them back. I think China realized, look, we had a good piece of technology and AI that a lot of people liked, especially from the viral demo. And, you know, liked it enough that Meta wanted to buy it for $2 billion. Let's bring it back and let's make this like a, you know, our own. competitor to a bunch of these American companies. They're not going to let it die, basically, is what I'm saying here. And so when they unwound it and

they're going to do this new $500 million round, it's coming from Bayou Capital. It's coming from IDG Capital. And then they have a bunch of existing stakeholders, Tencent, HSG, which was formerly Sequoia China. They have the Zen Fund. So basically, I think the government of China is like, hey, bring it back. let's get all the big Chinese investment funds in. We're going to like fund this thing. We're going to push it. The government put their finger on the scale and said, this is the winner. Everyone have at it. So everyone is now backing this company. And it's interesting if you look at like the pivots and the direction that the company has taken, because after the big viral launch, like I didn't really hear that

much about it, except that I think in like December, they said that they had annual recurring revenue of $100 million. So it was, you know, it was doing really good at the time. But honestly, I... I'm really curious where it's at because I have a friend that used it after it got acquired by Meta because he worked on Meta ads. But now that there's Muse out for free, like he's not using Manus and it got pulled out. So I think that it probably lost a lot of that monthly recurring revenue, but we'll see where it's at. I don't, I also am like dubious. I think it's going to probably do really well in China. I'm dubious of like American companies, actually. If there's a company that the Chinese government has enough

control over to unwind an acquisition by an American company. and force it back to China, I'm like very dubious. They said that they're going to expand, you know, they're going to expand their offices and stuff around the world. So they're going to have like American offices for this company. But I'm really dubious that like American companies are going to put a lot of investment and have this running their whole company when they could use something like Claude or ChatGPT or Meta or Google, where there doesn't seem to be such like, I don't know. those types of risks where the Chinese government has such direct control over this company. So it'd be interesting to see what happens there. There is a company

called Vesta. They've just raised $30 million. The reason I wanted to bring them up is because they're working in the mortgage industry, which I think is awesome. But the headline from TechCrunch was Vesta raises $30 million to bring swarms of agents to mortgage lenders. And I just hate the term swarms of agents. If you read what they're doing, I mean, They've helped people to the revenues up 12x this year. So like they're doing really good. I think they formerly had raised $85 million like all time. So it is a company that started in 2020. So I mean, I think this is I wouldn't call this like a quote unquote AI company at its inception. It started, you know, six years ago in 2020. And they're doing good things with

like mortgage lending. But I think they're just trying to. be a buzzword basically and say that they're, they have swarms of agents just to make it seem exciting and interesting and new. But I hate that phrase because people are like, oh no, swarms of agents are taking over the internet. It's not swarms of agents. It's just like parallel processes. Like it's just so dumb to me that we call these things swarms of agents. It's like you have eight, you know, tools, eight pieces of software calculating eight different things at the same time. That's not a swarm of software.

Like we have software swarms. calculating and helping our our company like work like microsoft uses swarms of electricity to get their work done like we have swarms of light bulbs in our office anyway i mean it's just ridiculous as far as how we how we talk about these things but anyways yeah they have about five percent market share i think and they're helping people speed up the loan processes which is something awesome and and i'm not down i'm not talking down on this company at all because as someone that has been pained by the mortgage process in the past. Like, yeah, it's not fun. It takes really long time. It takes 30 days or 40 days to close on a house. A lot of this is unnecessary.

It feels like there's so much bureaucracy and red tape and paperwork that is not necessary or it just takes way too long, right? Like if you qualify, you qualify. You don't, you probably shouldn't have to wait that long for all the stuff to go through. They say a lot of it is just because you're waiting on humans to review stuff and so they help speed that up with AI.

I'm thrilled with that. I'm thrilled with everything they're doing. I love that. My only critique is just on swarms of agents, that phrase, which I'm seeing pop up all over the place in journalism and when companies launch new features. These are not swarms of agents. It's just useful software automating things and helping us save time, which is great. But yeah, I just think the semantics on it drive me a little bit crazy. Okay, guys, thank you so much for tuning into the podcast.

If you haven't already, go check out AIbox .ai if you want to see all of the new features. We have the ability to pick a group of AI tools and get it to generate four consecutive or a handful of consecutive responses. So I could get Claude and ChatGPT and Gemini and Grok to all generate a response to my question all at once. You also have the ability to talk to it and have it build tools for you, access tools that you've already built. It's a pretty amazing upgrade to the AI box ecosystem. If you want to go check out AI box .ai and use that as your own virtual workspace for getting everything done with AI, you get a hundred different AI models for $8 and 99 cents a month. Greatest deal you'll see on the internet for bang for

your buck. And yeah, we'd love to have you as a user. So go check out links in the description and I'll catch you in the next episode.

OpenAI's Releases "Intelligent UI" and Anthropic's New Model

Kapitelklicken springt direkt hin

Shownotes

Transkriptvom Podcast bereitgestellt · Zeitmarken springen hin