AI News: $500 ChatGPT Plan? OpenAI Agent Leak & Opus 5.5

0:00 It was a crazy week of AI releases and updates, and I know you are probably just as overwhelmed as I am. Openai just dropped two cheaper GPT6 models. Plaude Opus 5.5 is here, and it's incredibly

0:11 good and cheap. Meta announced a brand new AI device that you can carry on a keychain. And there are rumors of OpenAI dropping a brand new AI agent platform next week. So, in this video, I'm going

0:21 to break this all down. Every single Friday, I do these weekly AI updates, breaking down everything you need to know so you can stay up to date without being overwhelmed. So, without further

0:30 ado, let's dive right in. And just like that, Anthropic is back. They just released their new model called Claude Opus 5.5. And I'm not going to lie, usually when a model drops, I'm pretty

0:40 skeptical because everybody says it's a gamecher. And I usually think it's just hype. But there are a few very practical updates to this new model that I'm genuinely excited about. First things

0:49 first, we have 40% lower costs with this model compared to the previous Opus 5 model. As you can see right here, Enthropic said, "We're introducing Claude Opus 5.5, the first model in our

1:00 new Claude 5.5 family. It performs at the level of Claude Fable 5.1 on most work and costs 40% less to run than Opus 5, which Opus 5 obviously is much cheaper than the Claude Fable models. So

1:13 supposedly this is going to be a really cheap model to run at give or take the same intelligence as Fable 5.1. Now let's quickly break down the API pricing. So, first of all, the input

1:22 tokens went from $5 for Opus 5 and it is now down to $4. Output tokens for Opus 5 were $25 and now this is $20. As well as we do get a decrease in the cacheed input as well. On top of this being

1:35 cheaper to run, Opus 5.5 requires less compute to serve than Opus 5. And then here is the chart we are so familiar with seeing whenever a new model drops. So, this is actually pretty impressive

1:45 here. Opus 5.5 on Aenta coding is 66.4 4 and this is compared to Fable 5.1 which is 55.8. So it scores over Fable 5.1 on most of these benchmarks here or I guess every single one of them which is

1:59 actually pretty crazy if you think about it with how cheap it is. And this benchmark right here, computer use for me is the main one that I'm looking at with these new model releases because

2:07 obviously coding is already very good when it comes to models. Even though there is, you know, decent increase here, but computer use is the key thing that I think that AI is still trying to

2:16 crack and when it does, it's going to be pretty gamechanging. So, I'm just waiting for this number to get higher and higher. So, this is just like a a slight 1% increase over Fable 5.1, which

2:24 I wish were higher, but of course, it is nice that it is better for cheaper. And then of course we obviously have the new big dog on the block and that is chatubt6 Astra and Opus 5.5 does compare

2:34 very well to GPT6 Astra as it is better at majority of these tasks as well. Looks like the only things Astra is better at is business workflows and agentic scientific research. Now this is

2:44 very interesting. Enthropic tested this out compared to Fable 5.1 and it was able to perform a code migration in 9 and 1/2 hours whereas Fable 5.1 previously took 12 hours. So, it is

2:55 significantly faster than Fable 5.1 as well. And probably the craziest part about this is it was 51% cheaper than Fable 5.1. And one big thing about these new model releases is to be completely

3:06 honest, it's not necessarily, in my opinion, all about the new capabilities that it unlocks. Rather, how cheap can you run specific tasks compared to other models? Because yes, something like

3:16 Fable 5.1 or GPT6 Astra can do some pretty amazing and mind-blowing things, but if it's going to cost you an arm and a leg, most people can't necessarily afford to run that all the time for all

3:26 the tasks that they run. So, if they could get a model that is close to as capable for much cheaper and in this case 51% cheaper. I think this is a real improvement. Again, here is another use

3:36 case tested Opus 5 and Opus 5.1. So, Opus 5 took 93 minutes, 5.5 took 63 minutes, and then the reported cost reduction here was 50% for a spreadsheet and a presentation test. Now, this is

3:48 the thing that I am personally most excited about, and I think we'll probably make the biggest change for us as average consumers of these AI models, and that is communication. Enthropic

3:57 said, "We've made major improvements to the way that Opus 5.5 writes and communicates." One of the most common areas of feedback we heard about Opus 5. Its messages are much easier to

4:08 understand at a glance. If you are using AI models every single day like me, you probably get very tired of trying to decipher what your AI is actually trying to tell you. I've had this problem with

4:19 Grockbot really bad over the last couple of days and it's really gotten on my nerves. If Opus 5.5 does communicate better and give you better outputs that you could read at a glance and fully

4:29 understand what it's trying to say, this is something that I'm very excited for. Now, it says that they put the most important information up front. It's less likely to use jargon or random

4:38 phrases and follows the writing rules that you give it. So again, this is just all a plus in my eyes. I was just scrolling on X when doing research on this new model drop and I stumbled on

4:46 this and I just found this to be very hilarious and I wanted to share it with you guys. So right here we have Sam Alman scared because somebody just whispered in his ear saying, "Sir,

4:55 they've removed M dashes from Opus 5.5. It's so over for us." Because if you really think about it, it shouldn't be this difficult to make AI sound human. And so I just find this meme to be very

5:06 funny and this is kind of the overarching theme of trying to get AI to communicate clearly. And if I scroll down in this article, they even give me examples of specific outputs comparing

5:15 Claude Opus 5 with 5.5. And you can really see a difference in how much easier it is to read. On the left, Opus 5.5, like this looks like you're trying to read code, trying to understand what

5:25 it's trying to tell you. Whereas on the right, we get this text in bold right here to instantly identify what it is that you're even reading. I think that this looks so much easier at first

5:34 glance to understand what this is trying to say. So this is explaining a bug. I could then click on summarizing a thread, which is probably going to be more practical to everyday users. And

5:43 again, this is summarizing a Slack thread. So this is the prompt that they gave it in this initial testing. And then it's using m dashes, a huge blob of text. I don't necessarily know exactly

5:52 what it's trying to say at first glance. Whereas the output from Opus 5.5 is much simpler. It breaks down the problem. It shows the immediate fixes. It shows the long-term fixes. And I think that just

6:02 looks so much more clean. And then same thing with explaining a design change. Obviously, you know, this like has some code in here, just looks tough to read. Whereas this is much simpler and easier

6:11 to read. Now, I went and tested this myself. So, I basically told Claude Fable 5.1 and Opus 5.5 to go and pull from my YouTube videos and give me a breakdown based on all the data that I

6:21 have. And this is the output here. I'll give you a second to guess which is which. So, on the lefth hand side, we have Fable 5.1. And on the right hand side, the output is Opus 5.5. And I

6:31 genuinely do think that this is much simpler and easier to read. It breaks down the overview. It has this little table view to show the videos and the amount of views per day it's getting. It

6:40 shows the top performers right here. And I like how it makes the important things in bold, which just makes it much easier to read. Shows the engagement, traffic sources, and then probably the most

6:49 important thing, the three things that I need to do today. Now, here are some cool examples that Enthropic gave us of use cases of Opus 5.5. So, I was able to generate this video right here showing a

6:58 short story about our watermelon. I think that's a cool use case. Next, we have a recreation of Apollo 8's Earth Rising. So, it looks like it's generating all the visuals here along

7:07 with this little diagram. Then, this one's pretty cool. Here is a pencil sketch that you could play with. So, it's almost like an interactive uh infographic diagram that it's showing in

7:16 order to explain something. I I think that this could be pretty cool if you're trying to learn a specific concept or teach it if you're a teacher, for example. Now, here are some of the final

7:23 numbers. Astra still leads on specific tasks, like I said earlier. So, business workflows and scientific research, it does have Opus 5.5 beat, but honestly, after taking a look at the benchmarks,

7:33 I'm pretty impressed at how capable 5.5 is for how cheap it is and how fast it is compared to Fable 5.1 and GPT6 Astra. Now, the last thing on this announcement is we have more room before we hit our

7:45 usage limits inside of Claude. So, I think this is obviously going to be good news for people. So, we now have a higher 5h hour limit as well as we have a reset that we could save and use it

7:54 whenever we choose to. And this applies to anybody who is a pro-Max team in seatbased enterprise user. And then lastly, here are all the places that we can use Opus 5.5 right now. Right. So if

8:05 you haven't been living under a rock this week, you've probably seen these viral videos all over ex Instagram and Tik Tok of Sam Alman, Elon Musk, and other people recreating this Migos clip.

8:15 And what I'm going to do is I'm going to show you exactly how you could recreate this with yourself or different characters in just a couple of simple steps. First of all, if you haven't seen

8:23 this, let me just play part of this very quickly. >> Drank. >> I mean, look at the movement right here. Sam Alman literally, I mean, he just

8:39 looks like a G. This is absolutely hilarious. Who would have thought these two characters would be like, you know, making a rap song like this? Anyways, like all jokes aside, lots of people are

8:48 making these remakes with a bunch of different characters. And this is very simple to do using a platform called Higsfield and specifically their video model called Genjitsu. I mean, if you've

8:57 ever seen Breaking Bad, you know these two characters are absolutely legendary. And then take a look at this. I actually recreated this with myself and my business partner, and this is pretty

9:06 solid. I think that this is absolutely hilarious. I need to send this to my mom, honestly, because I think she would love this. And then if you're interested, this is

9:16 actually the original right here, the original music video from Migos compared to the output, the AI recreation that I created using Higsfield Jingjitsu. So what I'm going to do now is show you

9:26 very simply how you can go and recreate this so you could share this with your friends. Now, first of all, I'm going to come inside of Higsfield. If you guys want to try this out, there will be a

9:34 link in the description to sign up for an account. So, I'm going to click on videos and then I'm going to search specifically for this Higfield Jenjitsu model right here that has this little

9:42 new tag. So, I'm going to click on this. So, basically what we could do is upload a reference video right here and then we could add different characters, products or clothes and it basically just

9:50 manipulates the reality of this video. So, this example right here that you're looking at, it's this character basically staying very consistent, but it's just changing the environment and

9:58 the clothes that he is wearing. So, in order to generate this, of course, we need this specific reference video. So, if you guys want to use this specific video, there will be a link in the

10:06 description for free to just download this on Google Drive. It's very high quality, which is really important. Of course, we want to upload a reference video with high fidelity, so that way

10:14 the output looks as good as possible. Next, all we need to do is come into Higsfield Jenjitsu and then upload this reference video. We could just select, you know, basically the frames that we

10:22 want to add this for. So, I'm just going to do this for, let's say, 22 seconds here. I'm going to click confirm. Next up, make sure to upload your reference images in this upload box right below.

10:31 And then, here are the ones that I use. So, here's this image of me wearing sunglasses. And then there's this image right here of my business partner also wearing sunglasses. So, now what we need

10:39 to do is we need to just turn on this prompt here because we need to make a few specifications. So, let me show you the prompt to use and you can just screenshot this and paste this directly

10:47 in to your Higs field when you generate this. Now, here is the prompt that I'm using. Replace the man on the left of the source video with the person in at image one and replace a man on the right

10:57 with the person in image 2. The original sound and movement should remain the same, just swapping characters. So, all you need is a very basic prompt like this. Obviously, just specify which

11:06 character you want to swap and then you should be good to go. In order to get the best output here, I'm going to do 1080p quality. And then I'm just going to click generate. This should take

11:14 probably a couple of minutes. So, give it like 5 to 10 minutes and it should be good to go. All right. So, this actually took about like 12 minutes to generate, but this one actually looks very good.

11:21 So, let's take a look at this right now. >> Hop off the ditch. I know this not a challenger. I get some members with me in the fridge. Get cozy. They some cannibals. They like to get geek. Drink

11:34 a whole bottle. Wake up and repeat. Damn. She took a little missing with the chill out. Now she say she I go in the jungle and I got a coat. I bet I come out with a meme. Do this for the

11:44 bigger than me. >> I mean that's hilarious. Hopefully you guys can use this and maybe make some memes with your friends. Send this to your mom. If your mom is anything like

11:52 mine, she'll probably actually think this is real. Anyways, it's super fun to play around with Higsfield Genjitsu. This is just one use case of it that's been going viral. So, I figured I would

12:00 show you guys exactly how to use it. Now, I want to quickly talk about using Grockbot inside of Tesla because this video because this video on X leaked and it's basically somebody using their

12:09 Grock bots inside of their Tesla Cyber Truck and they're using this in order to manage their inbox, clean up their calendar, and talk through their different files on their computer, for

12:17 example. I think we're at an interesting point in time where all these different LLM providers are starting to build out personal AI assistants, whether that's Grockbot, Metamuse, and now Metam Muse

12:27 Charm. All these different standalone products where we could interact with these agents to go and work across our different applications. I personally found this very interesting cuz I think

12:35 we're moving towards a point in time where you can just wear certain glasses or a physical product and speak with your AI agent in order to get majority of your tasks done. And on the topic of

12:44 Grockbot, I did see this on Twitter as well. So, Grock came out and said that they rebuilt customer support around Grockbot in order to scale their operations without adding their

12:53 headcount. And right here inside of the article, it says, "A new combined team has seen 175% increase in support tickets, but we did not have to hire any new people thanks to Grockbot." And the

13:03 craziest part about this is it said we might have hired 200 additional people otherwise if it weren't for the tasks that Grockbot could automate. This is just a glimpse into the future of having

13:12 these personal assistants handling basically a majority of the stuff we do on our computer. Obviously, we are not there yet, but the capabilities of these different AI tools are growing

13:20 drastically. Next up, we have an update from Enthropic inside of the Claude app on our phone. So, we now have the ability to add multiple different accounts inside of our Claude app on our

13:31 phone that we could toggle between. This is obviously going to be helpful for people that want to separate maybe like their work life and their personal life inside of Claude. So, you can now set up

13:39 multiple different accounts that you could connect to. I'm going to quickly show you exactly how to do this. It's very simple. First of all, this is only inside of the Claude mobile app. Then,

13:46 you're going to slide over to the lefth hand side where you find your chats and then go down to the bottom left where you have your profile icon. Then, at the very top, you can now see I have my

13:55 personal account added and I could click on add accounts, but I already went and configured one. So, I have one with my sponsorships email. So, I could go and click on this and then just like that,

14:03 you could see this is basically just a new claude account. I don't have any previous conversations or anything. This is a small update, but I think this could be pretty helpful if you want to

14:11 separate, like I said, your personal life and your work life inside of Claude. Now, on the topic of using different AI agents like Claude Code, Codeex, Grockbot, all of these different

14:20 agentic AI platforms, it's more and more important that over time we stay platform agnostic. And what do I mean by this? This means let's say a new model drops from Claude and you are inside of

14:32 Codeex. Well, we need to make sure it's very easy for you to go back to Claude if you need to start using that again. Or for example, right now I'm using Codeex for the majority of my coding

14:42 tasks and my really complex workflows, but I use Grockbot on a day-to-day basis to handle all of my admin related tasks. So then the question becomes, how do we share context between these different

14:53 platforms so that way we could switch between them with no problems? And I'm going to show you right now exactly how we could solve this problem. So that way they have all of the necessary context

15:03 inside of this second brain. So you could work across any platform you want. Now I personally use a platform called Neon which is basically just a database that our AI agents can use. We could

15:12 build applications from this. And the best part is is we don't need to be a developer in order to use this because this plugs directly into our AI platforms like Grockbot, Claude Code,

15:22 Codeex, and basically all these different AI tools that have specific connectors, allowing us to take the information and the conversations we have inside of those Gentic AI platforms

15:31 and storing that inside of this database. Now, before I show you how to set this up, let me explain this even more so you can understand it even if you're not a developer. So, I want you

15:40 to think of our Neon database like a filing cabinet. So, we could use our AI assistant like Grockbot or whatever the case may be. And we could basically just tell it to keep an idea and store it for

15:50 later. And then it's going to put that away and store it into this file cabinet that we can access whether that's inside of Grockbot, whether that's inside of Claude, or let's say there's a new

15:59 groundbreaking AI that drops next week, that could access it, too. This way, we now have one unified memory across all three or four or five of these different AI agents, however many you use. And

16:10 that is why it's really important to build out a database with something like Neon. Now, I'm going to show you exactly how to set this up. Obviously, you will need to have an account with Neon. So,

16:19 go ahead. There will be a link in the description to sign up if you want to test this out. And then I'm going to show you how to connect this inside of Codeex, but it's going to be very

16:26 similar in other platforms like Claude or Grobot, for example. So, I'm just simply going to come down to right here where it shows my name, and I just need to go and find the plugins. So, right

16:37 here under integrations, you're going to see plugins. So, I'm just going to click on this, click on browse directory, and then just search up neon. Since I already have a built-in connector, it's

16:47 very, very easy to add this. So, I'm just going to click on this plus button like we do any of our other plugins or connectors to these different AI applications. I'm going to click

16:55 continue to Neon. And what it's going to do is it's now going to ask me to approve. So, that way chatbt can access our Neon databases and create databases and all those different necessary

17:05 actions it needs. So, I'm going to click approve and continue to Neon. I'm going to click on authorize and it's as simple as that. It's going to bring us back to chatbt and then we should see this now

17:15 connected. So if I search inside of my plugins and I search up neon you can now see that we have this database configured and again the process is very similar for rockbot. So we come to

17:25 marketplace go to plugins search up neon postgress and click add and we'd be able to set it up in a couple of very simple steps. Now for reference this is what our database inside of neon looks like.

17:36 This is a short form video tracker that I have created. I use this across my Claude Codeex and my Grockbot. And what this does is it pulls all of my top short form videos that I've ever posted.

17:48 And to show you this in action, I'm going to pull up Grockbot here. And I simply just said pull from the Neon database and tell me about the top performing short form videos that I have

17:56 posted. It pulled automatically from Neon via this connector. Breaking down all this information. Obviously, this is metadata that is stored inside of our Neon database. And the best part is is I

18:07 could come over to Codeex and since this is also connected to Neon, I could give it the exact same prompt and it's going to pull the same information. So, let me show you exactly what I mean. So, I'm

18:16 going to send this off. And now Codex is going to go and pull this information and and it should have an answer very shortly. And here we go. but pulled from that exact same database inside of Neon

18:24 that Grockbot pulled from and broke down in a list view ranking all of the top performing short form videos for me along with a link to them as well because obviously all this metadata is

18:33 stored inside of Neon and it's very easy for us to access it inside of whatever AI agent platform we're using. So this is a very specific use case for me and how I'm using. Now, this is a very

18:43 specific use case for me, but this can be very powerful when you connect this to all of your different AI agents and basically creating a second brain that you could work across all these tools

18:51 on. In typical fashion, like an hour after Enthropic dropped Claude Opus 5.5, OpenAI decided to drop GPT6 Soul and Luna, their newest models to the GPT6 family. So, now on top of having Astra,

19:05 we now have these two cheaper and faster models, Soul and Luna. So, let me break these down right now. At the bottom of the screen, you can now see the difference in the input and output token

19:14 cost here. So, Astra is our most intelligent model for the best results. $10 input tokens, $50 output tokens, which is obviously very expensive. Then we have Soul, which runs at $2 for the

19:25 input tokens, and then $10 per million output tokens. This is best for complex coding and professional work at a lower cost. And then we have Luna, which is fast and efficient for everyday work at

19:34 scale. The input tokens there are 10 cents and the output tokens are 50 cents. So much cheaper than both the Astra and the Soul models. And then here is a breakdown compared to the previous

19:45 GPT 5.6 models. So it is a step below. Looks like GPT6 is cut in half from the API prices compared to GBT 5.6. And then here is the cost breakdown for Luna again compared to the previous GPT 5.6

19:58 prices. Now, if you're interested, here is a chart comparison from GPT6 Astra, the new GPT6 models, as well as Claude Opus 5, Fable 5.1 for business workflows. I'm not going to go through

20:09 all of this, but if you want to take a look at this, pause the video and you can get a sense of how capable these are for the cost. And then here's another breakdown. Same thing, comparing the

20:18 different models based on the factual mistakes right here. Now, here are some examples of demos that I saw on X when I was just scrolling doing some research on this. So, this is pretty cool. Luna

20:26 generated this. I mean, that looks really solid. I feel like we're at a point in time where all these models can basically do whatever it is that we want that's not like super complex tasks. It

20:35 really just boils down to what's the cost comparison to previous models. Right here's a comparison between GBT6 Soul and Opus 5.5. I personally think Opus 5.5 looks much better than the one

20:45 on the left, but this is just my personal opinion. So there's three different places we can now use these new Chacha BT6 models. So we could use it inside of Chacha BT work on a paid

20:54 plan. We could use it inside of Codeex on a paid plan and again via the API. Next up, we have a potential leak of a brand new chatbt plan called ChateBT Pro Max, and it's $500 per month. As of

21:06 right now, this is just a rumor, but this tweet went out today by testing catalog breaking down this chatbt pro max plan for $500. As you can see right here inside of this code, there's this

21:16 new model that was leaked. It says Pro Max amount $500 right here. Now, this is obviously just pure speculation, but I do think it's only a matter of time before these model providers like

21:27 Anthropic and OpenAI begin releasing these ultra expensive models like this right here. We've never seen over like $200 per month from these different platforms. So, this could be the first

21:37 time that we see some pretty expensive models. And to be honest, a $500 per month subscription is pretty crazy when you think about it. But I guess if you are a Codeex power user like I have been

21:47 recently, this could be worth it for you because I've personally been hitting my limits on the $200 per month plan. So I could definitely see quite a few people taking advantage of this new plan even

21:57 though this is pretty damn expensive. Let me know what you think in the comments. Is this just ridiculous pricing or does this make sense with the demand for this? Next up, we had another

22:05 big release from OpenAI, and this is a new voice mode that now gets plugins powered by the brand new GPT6, Astra, Soul, and Luna models that we just talked about. Now, here's a few things

22:15 that the new model can do. It could use your plugins like email, calendar, and Slack powered by these different models, and it can be used inside of chatbt, work, web, and mobile. And this is

22:25 rolling out globally. So, you should have access to this now. Now, the main thing that's new here is not only a better voice. Of course, they keep incrementally making the voice sound

22:33 more human and better, but this could now speak with our different applications. So, we could use chatbt voice mode as the bridge between our email, our calendar, our Slack, and

22:42 basically all these different applications that we're using on a day-to-day basis. It does feel like we're moving towards a world where we could simply just speak with our voice

22:50 agent and get these tasks done for us. Just like I was talking about earlier on in the video with the new Grockbot inside of Tesla as well as the brand new Metam Muse Charm product that they just

23:00 released a couple days ago. Of course, the best part about this is you could speak with the different applications and the information across the apps you use every single day. When you're using

23:08 any AI agent, for example, chatbt voice, you can connect it to all the different applications you use. But what happens when chatbt only has a certain list of apps and the one you want to connect to

23:18 isn't included? Well, you can use something called the Zap Your MCP as a bridge in order to connect to those. This is a perfect example right here. So, with chatabiki voice mode and the

23:28 zap your MCP server, I'm personally able to go and use and pull from these different applications. School, which is a platform I use every single day, Beehive and Drip, which are both of my

23:39 email marketing softwares that I can't connect to directly inside of Chatbt. As you can see, inside of this MCP server, I have a ton of different apps configured. So, these are all apps that

23:47 I can now pull from just by configuring this simple MCP server and we're able to do this with Zapier. So, if you have a specific app you want to connect to or multiple different apps and you want to

23:57 try this out, there will be a link in the description to sign up and begin playing around with this. All right, so in order to test this out, I'm going to come into chatbt work inside of the

24:05 chatbt desktop app and then we're going to obviously see the voice button right here, which we already had, but this is a new voice mode and can use the specific plugins that we configure

24:13 inside of our chatbt app. So, I'm just going to click on this and start conversating. Hey, how's it going? >> Good. What do you want to tackle today? >> Yeah, I wanted you to quickly pull from

24:22 my Drip account via the Zap, your MCP. And I want you to pull the last email I sent in that email marketing campaign. >> Okay, let me take a look. Got it. Your latest SAD broadcast is what happened in

24:34 AI sent September 20th. I've shown the summary and a link to open the full email. >> Perfect. Could you actually show me the uh clickthrough rate on that email, too?

24:42 I want to see how many people like what percentage open that. >> Sure. Let me pull those stats. Looking at both the opens and clicks for what happened in AI. For what happened in AI,

24:52 the open rate was 28.6% 4,156 opens. The click-through rate was 0.59% 86 clicks. I show them both in a table. >> Perfect. Thank you so much for this

25:03 information. All right. So, there we go. That's obviously the voice mode in action. Like, I don't notice a difference. It doesn't sound any better to me, but it was already pretty solid.

25:10 And then it's able to now go and pull from these different apps that we connect via the plugins, which is what catcht calls it. I think that over time, this is obviously going to get better

25:19 and better. To be transparent, I did clip this up so that way you didn't have to wait as long as I did, but it did take about like 15 to 20 seconds in order to get a response. When it was

25:28 going and pulling information from these different apps, so when I'm just speaking with voice mode and chatting with it, the latency is very, very quick. But when it is pulling from apps,

25:37 it does take a bit longer. But I'd assume that they're going to work on this. So, it is much faster. All right. Next up, we have a brand new product from Meta, and this is called Muse

25:46 Charm. This is a physical device that runs their new AI agent called MetaMuse, which is a personal assistant very similar to Grobbot. They just announced this yesterday at this Meta event. I

25:56 think this is very interesting. we could just speak with this, have it send emails, plan a trip, organize our day. And again, this is going back to the overarching theme that we're starting to

26:04 see in this AI space, and that is having personal AI assistance at our disposal, whether on our phone, inside of our glasses, or now a standalone device. Like, so they only announced this

26:15 yesterday, and they have a shipping target of December, even though they don't have a ton of these produced. Mark Zuckerberg actually said there's only a couple of them that they have these

26:23 different prototypes and we do not have a price yet for this device. But this is something that I'm personally excited about. I think we are getting to a point in time very soon where we will be able

26:33 to carry something like this in our pocket or have these running on our glasses and we can just speak with our AI agent and it can go and do things for us which I love. Now this might be the

26:42 biggest news of the week and there is a leak from OpenAI about a brand new agent called Aeon. I'm not sure exactly how to say this. It's spelled AE O N. I've personally heard some rumors that

26:53 they're going to name it Chachab O. Chachabt Orbit. And this is apparently a new name that people are saying that it's going to be named after. Of course, this is all speculation, but I do

27:03 believe next week OpenAI is going to be releasing something very, very similar to this breakdown right here. According to this tweet, here is the speculation. AON is going to be more than just a

27:12 chatbot. It could have its own profile, DMs and groups designed to compete directly with Grockbot and Muse. Could launch as early as Thursday. Also, it could launch on OpenAI dev day, which is

27:22 coming up. And he goes on to speculate that OpenAI is building an entire social layer around its AI. I'm pretty sure we've all been anticipating this coming from both OpenAI and Enthropic. I also

27:32 heard that in the next couple of weeks, Enthropic is going to release their own Grockbot competitor as well. So, I think that the world of personal AI assistance and agents is really going to start

27:41 accelerating over these next couple of weeks because we've all seen how powerful things like Claude Code and Codeex are, but there is not an intuitive way to interact those

27:50 different platforms and Grock Bond Muse kind of solved that and paved the way for products like OpenAI. And before those two platforms, there was OpenClaw and there was Hermes agent. And don't

27:59 forget that OpenAI actually hired the person that created OpenClaw and I'd assume he is behind this project as well. I personally really hope this news is true. I've been absolutely loving

28:09 Grockbot, but I do feel like there are some limitations because the Grock models aren't on the frontier of these different LLMs such as ChateBT and Enthropic. So, I'd assume they're

28:18 probably going to be even more powerful than Grockbot is right now. And if you do the math with Anthropic and OpenAI releasing these new models this week, this could be them laying the foundation

28:27 for this new product to be released. Anyways, there hasn't been much leaked about this on the internet. With that being said, there hasn't been much leaked on this topic. So, who knows if

28:36 this is real or not, but if I had to put my money on it, I would expect them to be dropping it next week. And there we have it. This week was loaded with AI updates and product releases, and I

28:44 believe next week is going to be even more hectic. So, if you enjoyed this episode and you want to see these every single week, make sure to subscribe to this channel for more, as well as let me

28:52 know what you think in the comments below. Hope you have a good day and I'll see you in the next