Enjoying this issue?
Get tomorrow's AI & engineering digest in your inbox — hand-picked, summarized, and always spam-free.
No summary yet. Click “Regenerate summary” to queue one.
Stop scrolling. Start reading smarter.
Receive the day's most important AI & engineering updates in one concise email. No spam.
Transcript (captions)
This was the biggest week of AI releases this year and your life is too short to sit around and research all of this yourself. So, in this episode, I'm going to break down everything that you need
to know in order to save you time. Openai dropped chatbt dots space and a new GPT 6.1 soul model that is Astro level at 1/5if the price. XAI released team Grockbots. Enthropic dropped Claude
Sonnet 5.5, which is incredibly powerful for very cheap. Every single Friday, I break down the top AI releases so you could stay up to date without being overwhelmed. So, grab a coffee, sit
back, and let's get into it. OpenAI just dropped their brand new personal AI assistant called DOT that is powered by the ChateBT6 Astra models. They also just dropped Chache 6.1 Soul, which is
their newest model that is 15 the price of GPT6 Astra with near Astra level intelligence. We also get a brand new platform called Chatbt Space, which is very, very interesting that we're going
to talk about later on in this video. And they even released a $500 per month plan that gets 25 times the usage of Plus. All right, so first things first, we obviously have their brand new
product called DOT. This is their personal AI assistant that is a competitor to things like Grockbot and Metamuse that we've seen over the last month. I personally have been using
Grockbot as my main daily driver every single day along with using codecs on the side for my more specific workflows that I want to run. So I'm very excited to test this product out as I would
prefer to stay within the OpenAI ecosystem as opposed to Grock since these are more Frontier level models instead of having to use the Grock models that in my mind are a bit
limited. Right here is the demo video that they showed when they announced this. So let's just skim through. Basically dots are these personal assistants that we create. They have
this cute little design here, which honestly looks like they just stole this design style from Grockbot. But this lady right here is speaking with her dots. And one really cool part about
this is not only can we text with these over the chatbt app, but we could actually speak with these over the phone. So we could call our Dot, we could text our Dot on SMS, we can email
our Dot, we could add it in Slack messages. All these different ways to communicate are now possible with our personal assistant. And the thing I'm most excited for is I have a feeling
over time they're going to release a hardware product where we could use that to communicate with our agent instead of having to carry our phone. Now this right here is an example of the
interface of chatbt dots. I don't have access to it yet otherwise I would test it out. I will be covering it in the next couple of days when I do have access on this channel. So make sure to
subscribe. On the right hand side is our little interface here. And this looks very very similar to Grockbot. We have our little icon up here where we could name our specific agents. This is not
like MetaMuse how we only have one agent. We can configure each individual specialist agent for given tasks. This isn't going to be anything groundbreaking. If you are familiar with
platforms like Grockbot and Metamuse, if you've been testing it over the last couple of weeks, but dots are basically able to be proactive and work even when we're not communicating with it. So
right here in this example on stage, it shows that Alfred Sam Almond's dot is saying, "Good morning, Sam. I've already caught up on your Slack and your email. Nothing urgent or just three things to
look at when you get to your office. So, basically, it's proactive. It's able to follow up with you. We're used to using chat GBT and codeex and having to give it a prompt. So, now that we could
actually text with it and it can be proactive while we're not necessarily engaging with it is the main difference here. Now, let's quickly talk about how do work and the infrastructure behind
it. So, basically, we have our dot which is where we're interacting with it and this is using a few different things. It is using codeex. It's using chatbt work and a cloud browser along with the
applications that we connected to. Basically, it's able to go off and use codecs without us needing to be inside of the codeex app. We could just chat with our dot and it's able to go and
perform actions across the projects that we're working on inside of codeex. And again, just like rock, it does have a cloud computer, which if you don't know what this means, this basically means it
has a designated area where it can go and work that's not on your physical device. This way you could have it go and look through Instagram for example once you give it a sign into Instagram
instead of it running on your own computer. It's running in the cloud so you could still continue using your computer while it's going and doing that task. This is basically what we already
have inside of codeex if you have used that platform as well as chatbt work. So this isn't necessarily anything new but these three things are what make up our chatbt dots. Now of course people are
going to ask about the permissions and the security behind this. Is this safe to run? Is it going to go and text your ex-wife telling her that you miss her? Things like this you don't want your AI
to do. So, we're able to set specific permissions before delegating tasks. So, we could connect it to only the apps that we want to use. We could then have it require approval for immediate
actions if we don't want it to go and run rogue, which most people are going to prefer to do that. Now, let's quickly talk about pricing with chatbt dots. This is very, very surprising and this
is going to be good news to basically everybody. Our first dot is included in our pro or business premium plan at no extra cost. It's available 24/7 to talk, help you think, and stay on top of what
matters. Conversations with your chatbt dots don't count toward your chatbt usage limits. Now, when it does use our credits is when you ask your dot to start or manage tasks inside of codeex
or chatbt work, then those are obviously going to count towards our usage limits, which makes complete sense. But if we're just chatting back and forth with the dots, it won't charge us. Next up,
probably the second biggest part of this release in my opinion is chatbt space. This is basically a new workspace that you can work across not only with your team but also with these chatbt dots. So
it's almost like a unified work environment that they're trying to get us to work inside of. Now I personally don't have access to chatbt space yet along with dots, but this right here is
a live demo of what this looks like. And this looks very similar to notion. From my understanding of watching this initial demo, it seems like they're trying to make a unified space like
Notion for both our team members and ourselves to interact with. So, for example, right here, her chatbt dot created this little feedback diagram dashboard. She's also then able to tag
her dot right here in order to go and make changes to this specific page. Right off the bat, after looking at the demo of this, I'm not sure exactly how I like this. It does seem like it could
get overly complex. I'd rather just communicate with my chatbt dot in a text thread like I could do with Grockbot. So, I'll be interested to see if this is something that I'll actually use on a
day-to-day basis. Now, moving on to the new model that dropped, we have GPT 6.1 Soul, which is now 80% cheaper than GBT6 Astra. And OpenAI is even saying it's near the same level of intelligence as
Astra 6. Right, here's a breakdown of the price. So, we have $2 per million input tokens on GBT6.1 Soul, whereas Astra is $10. For cashed inputs, we have 10 cents as opposed to $1. And then for
output, we have $10 compared to $50 on GPT6 Astra. Now, at this point in time, it's getting incredibly hard to keep up with all these new models because Anthropic literally dropped Sonnet 5.5
yesterday, and a couple days before that, they dropped Opus 5.5. It seems like both OpenAI and Enthropic are dropping a new model literally like every other week at this point. So only
time will tell if this is actually that capable of a model for cheap even though openai is saying that it will be. Now where is this available? This is available in chatbt work codecs and the
API as of today for all plus pro business and enterprise users. Next up we have a new $500 per month plan inside of chatbt which I'm personally excited about. I'm definitely going to be
upgrading today. On this plan we get the new ultraast model which was also released at devday. We have no 5h hour limit. we could use it across our partner apps and it's 25 times usage of
the plus plan. Next up, this one gets a little bit developery, but I want to talk about it very quickly. So, they also released the new decisions API, which is very similar to Jev, which has
been taking the world by storm over the last couple of weeks. This is a new way that our AI agents are able to make decisions very, very quickly. And in return, this means that computer use can
be much better. So, as you can see on screen right here, we have our decisions API going and using computer use in order to go and book a flight for us. And as you can see, this is extremely
fast. If you've ever used previous computer use, it is very slow when it comes to being able to make decisions on what buttons to click, how to use specific interfaces. So, this should
make a big difference when it comes to computer use and running our different AI agents inside of chatbt. All right, so those are the top things I wanted to cover from OpenAI dev day. I'm very
excited to begin testing out both chatbt dots GPT 6.1 Soul. I'm going to upgrade to the $500 per month plan and let you guys know if I'm still running into hitting my usage limits. One of the
biggest releases from OpenAI devday was completely overlooked chatbt space. This is a new way to interact with AI and our teams. And I'm not going to lie, I was very skeptical about this at first, but
I'm growing to really like this. So, what I'm going to do is break down exactly what it is and how to use it. They released about 20 different products. A lot of them are developer
tools, but there are a couple that are important for us average consumers. So, let me break this down very simply in like 30 seconds. First of all, they released DOTS, which are a new personal
AI assistant that we could use that use codecs, a cloud computer, and can basically just perform tasks for us and we could text it like a normal person. Next, we have ChatVt plugins, which
allows developers to build their apps directly into the chatbt app, which I'm actually going to show you later on in this video. and then chatbt space which is the main thing that I'm going to talk
about in this video. A completely new way to interact with both your team members and your AI agents like chatbt dots. They also announced a $500 per month chatbt subscription which I talked
about in that previous video as well. Now let's quickly talk about what chatbt space is before I show you a demo of me using it and what I love and hate about it. We could interact with this between
our team members and our AI agents. It allows us to create and collaborate together. We could visualize anything on dashboards inside of these chatbt sites. And most importantly, we could allow our
chatbt dots to work inside of these unified spaces as well. Now, your next question probably is who has access to this? So, if you're a prouser, a business user, and enterprise user on
web, desktop, and mobile, you will have access to this. Now, enough talking about it. Let me show you exactly what this looks like inside of the chatbt desktop app. So, first things first, I
want to give you a quick tour of the new interface inside of chatbt because it does look quite different. So, first of all, we have chatbt. We could toggle between chat and work mode. We then have
our normal chat bubble here that we can interact with. We have our pinned conversations on the lefth hand side. And then I could toggle between chatubt and codec. So, if I switch over to
codeex, it's going to be the exact same thing. We could work inside of our projects and folders inside of here. And we even have access to our GPT 6.1 soul model which just released yesterday at
devday as well. Now what we're going to focus on here is this little page icon and this is chatbt space. And if I open up my space we're going to see a couple of different things. Now if I click on
this all button this is basically our dashboard across our entire chat space. On the lefth hand side we have a bunch of different assets. These are all the different pages that I've created. And
don't worry, I know that this probably looks and sounds overwhelming like what is the difference between a space and a page and all these different buzzwords. So let's quickly break down all the
concepts you need to understand about chatabt space right now so you can understand the rest of this video. All right. So the first thing that we need to understand is we have different
spaces and a space is basically just a hub that has a few things underneath it. So personally I have two spaces I've created. I have one called AI for non techies which is everything for my
YouTube brand. And then I have one that is called Loom Media, which is everything for my influencer marketing agency. So these are two designated areas that are separate from each other,
which I recommend creating multiple spaces for your personal life versus your business life, for example. And now inside of this space, this is what this looks like. Inside of my AI for non-
techy space, we have this short description on what this space is going to be created for. So, I have yet to fill this out. So, you could see I need to describe this, but this is basically
where we break down. Okay, here's the hub for this specific sector of my life or my business. And we can provide context for this. Next up, inside of our spaces, we have a few things. We have
pages and we have subpages. This right here is my YouTube's video page, and underneath I have a few subpages. So, I obviously have these different videos, which if I click on them, it takes me to
a more detailed view of that video. Now, one of the really cool things about chatbt space is you can see on the bottom right of my screen, I actually have codeex pulled up. So, this is the
little codeex window that I could trigger at any time. This is working inside of one of my folders that I work inside of codeex. This is mainly how I'm using chatbt at this point via codeex.
So, what I could do is I could work with this specific page. So, for example, I'm just going to come here and say, I want you to generate another thumbnail very similar to this one. However, I want you
to make the sun bursting into flames a little bit more. Now, what our codeex is going to do right here is it's simply going to take this thumbnail. It's then going to regenerate it with chatbt image
2.5 and then it's automatically going to add that to my thumbnails right here. All right. So, there we go. Now, we have that image added right below my previous thumbnail, which I think this one looks
much better. We obviously have the sun glaring a little bit more. So, you can begin to see how we can interact with both codeex chatbt inside of our spaces. I'm going to create a new page from
scratch and build this out so you could see exactly how this works in real time. So, I'm going to come over, click on new page, and I'm going to call this product ads. And I'm simply going to upload a
few reference images here. So, we have this little product. We have a Sun Bum can. And let's say that we want to generate product ads from this image right here. This is the reference image
we want to use. And one really cool thing about our pages is we could use the specific plugins we have connected to Chatbt. So, we could use the different applications inside of our
spaces. Now, the specific app I'm going to use is Higsfield. So, I'm going to use this to take that image and generate different videos that we could use for social media posts inside of our space.
What I'm going to do is I'm going to come to Higsfield. If you want to try this out, there will be a link in the description to set this up yourself as well. So, I'm going to click on install
Higsfield plugin. I'm going to click on open chatbt. And now, just like this, I have the Higsfield plugin installed. One really cool thing they added to Chatevt yesterday at Devday is they added
designated plugin. So, what this means is different products can build directly into chatbt. So if you see on the lefth hand side of the screen, we have this Higsfield icon. If I click on this, we
could open up this interface right here, which shows all of my different Higsfield generated assets. So now that we have Higsfield downloaded, what I could do is I could come to the chat and
I could say, "Hey, so I want you to take this image of this Sun Bum bottle and create multiple different product ads that we could use for social media. These are going to be used for Instagram
stories and different Instagram posts. So I want you to get creative to create these visual assets using Higsfield." So what I'm going to do is just type in at Higsfield so it knows to use Higsfield
and then I'm going to send this off. It's going to go and use this application to generate these images and then it will automatically paste them back into this page right here. All
right. So what Codex did is it went and generated these different images here. So we now have I think six different images of this sunbum. These look like honestly amazing. Before we would have
had to generate these in Canva like 6 months ago. So these look really really good. And so what I could do now is I could actually have it go and add this to this page because right now it
basically added these to a zip file that I have to go and download on my desktop. But I'm just going to simply come over here and just say, "Hey, I want you to actually go and add these to this
product ads page. I want you to add the text generated ads." And then below that, you could add the images. All right. So now we have all of these different assets generated and added to
our page inside of our chatbt space. So it breaks down the actual layout for each of these. So we have our Instagram story, we have our Instagram feed, golden hour for Instagram story, and
then Instagram feed. And then obviously, as you can see, it breaks down the aspect ratio, so we know which one we should post to which. I mean, these look really solid. Now, what we could do is I
could come back to our little CEX B bubble right here and say, "All right, I want you to take this product and I want you to now turn it into an Apple product style video. On top of that, I want you
to generate another video with Higsfield." taking the bottom image that you generated showing your son's sidekick and turn this into a video as well in the same style as that asset you
generated. So, I'm just going to give it that prompt. Again, I'm just going to tag at Higsfield so it knows to use the Higsfield MCP here and it's going to go and generate those assets and then add
it again to this page inside of spaces. All right, so a few minutes later, Higfield generated these videos for us. As you can see on the bottom right of our screen in our little codeex bubble,
it added this check mark. So that means that this task is done and has been added to our page. Also, one thing I didn't even mention is I could click on voice mode and we can interact with it
this way in order to make changes. Now let's go ahead and click on these. So if I click on this, it's going to bring us to this page. Looks like we can't pre preview here, but instead I could just
download this to my desktop. All right. So here is the first one. This is that Apple product demo styled video that I wanted to generate. Shows our sun bump can. Uh this is obviously very good.
Looks in the Apple style. So that's pretty cool. And then here is our second demo right here. Again, it's in that exact same style as before. We have the palm tree shadows moving. This is pretty
cool. This is something we could just throw up on our Instagram stories, for example. I do want to point out that we could access all of this information inside of our chatbt space in the
browser as well. So throughout this video, I've been working inside of the desktop app, but this will also populate on the web app as well. So if I come to spaces, you can obviously see that we
have all these different pages that we've been working on throughout this video. So, if I click on product ads, all of those product advertisements that we created live right here as well for
us to reference at any time. I also believe that this is rolling out to mobile users as well. However, I don't have access to it on my phone yet for some reason. So, we should over time be
able to see our spaces from our mobile app as well. Higsfield is just one of hundreds and even thousands of different plugins inside of ChatCt. And when we combine our plugins with chatbt space,
this is where it can become very powerful cuz we could pull from all the different applications that we're using in order to create internal dashboards. Now, let's say you go through and you
can't find the app that you want to connect to, even though there are a bunch of different apps that we could add inside of chatbt natively. Well, there is a workaround. So, let me show
you that right now. You can use something called the Zapier MCP, which is something I personally use every single day. I want you to think of this as a bridge between chatbt space or your
AI agents and all the different applications that you use because Zapier MCP connects to 9,000 plus different apps. So odds are the one you want to connect will be included in that and
it's very simple to use. In order to show you what I mean, I have this MCP server configured inside of Zapier and I have a bunch of different apps connected. One of which is Drip. This is
my email marketing software and there's no way for me to connect this into ChatBT unless I'm using something like the Zapier MCP. And what I did is I had chatbt create this separate newsletter
stats page that you're looking at right here inside of my chatbt space. And what this does is it gives me day a daily breakdown on everything that's going on inside of my newsletter campaigns. So it
shows the latest newsletter that went out. It shows when it was sent. It shows the tlddr breaking down all the stats. So I could see the clickthrough rate and the open rate. I could see the amount of
people that are unsubscribing. I could even see the amount of bot clicks that have been tracked from this as well. It then pulled up this little table here with a breakdown for me. So this is an
amazing visual for me to look at and catch space was able to do this with access to drip from the zap your MCP. If you guys want to try the zap your MCP, there will be a link in the description
to sign up to do this yourself. But I wanted to show this as a perfect example of how you could take the apps you're using and then create separate pages inside of your chatbt space because now
we have all this information that we could either view inside of here at any given time and this is now context that our Asian can use as a second brain so to say. All right, so I want to now talk
about why I've used notion in the past religiously and why I think that chat space might be the alternative that I've been looking for for years. So, let's quickly break down exactly how my notion
is set up so you can get a sense of why I'm probably going to transition to chatbt space. I think there's so many different features inside of notion that are unnecessary. I just like to keep it
simple. So, I have my main hub which is called AF for non techies which in this case would now be called a space. And then inside of this hub, I then have different subpages. So, for my
financials, I have a specific subpage for money tracker, monthly investments, wire info, all this stuff. These are all separate subpages that I've created. And this is what my video system looks like.
I have different subpages for all the videos that I work on with my video ideas. And then inside of those video subpages, this is what it looks like. So I'll have my title ideas, I have my
thumbnail ideas that I add inside of here. But one thing with notion is I have to download an image, upload it into notion myself, whereas inside of chatbt spaces, I had chatbt generate
this image right here and it automatically uploads this into my space. It really does feel like chatubt space is an AI native second brain instead of having to use something like
notion like I've talked about in this video. So we obviously have people which is you and your team and we have our AI agents and how we're going to collaborate with them. So over time
we're going to need an interface or a platform that we can collaborate with all these different people and AI agents and it almost feels like chatbt space is that and I could see the vision from
openi long term. Inside of our spaces we have all the relevant context. We have our different pages. We have project context. That includes the projects we're working on inside of our computers
via codeex. And all of this is context we could provide to both the people working inside of our companies and our AI agents working as well. Anthropic is back with a new model. They just dropped
Claude Sonnet 5.5 just a few days after releasing Opus 5.5. And as always in this video, I'm going to break down everything that is new with this model. I'm going to be quick and to the point
so you could save your time. Without further ado, let's dive right in. What I'm going to do in this video is break down this article right here that Enthropic gave us and just give us the
main talking points that matter to you so you could save your time without having to go through this entire thing and try to decipher this yourself. Enthropic introduced Claude Sonnet 5.5,
the second model in the Claude 5.5 family as they just released Opus 5.5 just a few days ago and it is very very capable and I've been impressed by it. They go on to say it's a clear upgrade
over Sonnet 5 and it runs 30% faster and it costs up to 30% less for most work. Along with this, Anthropic hinted that Haiku 5.5 is dropping probably within the next couple of weeks. So, this will
end up being the third model in the 5.5 family. If you're unfamiliar with the Claude Haiku models, it's mainly built for high volume and costsensitive applications. So, it's their cheapest
model that they have and obviously the least capable. However, we should be seeing a pretty big upgrade in this model over the coming weeks, which I'm excited about. Now, moving on to the
first big upgrade in Sonnet 5.5, and this is collaboration and more importantly, communication with this model. So, Enthropic said, like Opus 5.5, Sonnet 5.5 writes more clearly than
our previous generation of models. Early testers describe it as a much better partner for collaboration than Sonnet 5, its speed also makes it well suited for fast iteration on complex tasks. So,
this is one thing that I've really noticed when using Opus 5.5, and I've been seeing a lot of other people talk about this. It's very good at communicating. It doesn't feel like you
have to decipher exactly what it is that it's trying to tell you. And this solves a very big problem when it comes to working with large language models. So, right here is a perfect sideby-side
example of Claude Opus 5 versus Claude Opus 5.5 breaking down this specific issue right here. And it's very obvious that the one on the right called Opus 5.5 is much more concise and easy to
understand at first glance. Now, this is obviously some great news because this is now coming to Sonnet 5.5, which is going to be cheaper than Opus 5.5, even though Opus is pretty cost effective.
So, here's a breakdown side by side between Sonnet 5.5 and Opus 5.5. For cash reads, they are both at 20 cents per million tokens. Sonnet 5.5 is half the price of Opus 5.5 when it comes to
cash rights. the input tokens it is half the price and for the output tokens it is also half the price. So this is getting pretty costefficient compared to Opus 5.5 and like I said that has
already been pretty efficient. Next up, Enthropic says that Sonnet 5.5 requires fewer tokens per task than Sonnet 5. So it's less expensive to run and it also generates output 30% faster and its
efficiency is immediately noticeable. This is something that we're starting to see across the board with all these different model releases. It does seem like they're becoming much more
efficient when it comes to token spend. And this is something that solves a major problem for us consumers of AI because it's not like these models aren't capable anymore. It's more so
about how efficient they are when it comes to how much money it's going to cost for a given task. Now, of course, cost is important, but it doesn't really mean anything if it's not very capable.
Well, here is it compared to Sonnet 5, Opus 5.5, and Chat GBT6 Soul. The thing that really stands out right here is look at the difference in a gentic coding between Sonnet 5.5 and Sonnet 5.
5.5 is at 70.6% whereas Sonnet 5 was 10.3%. So many people were complaining about the capabilities of Sonnet 5. So this does feel like a meaningful upgrade when it comes to a lot of these
benchmarks here which I'm going to walk through. Next, you could see Opus 5.5 is 66.4%. It's even scoring 4% higher than the brand new Opus 5.5 model. And if you go
all the way down the line here, it obviously blows Sonic 5 out of the water on all these different benchmarks by a meaningful amount. And then if you compare it to Opus 5.5, Opus 5.5 is
still better at everything else aside from that agentic coding. So it's just very interesting to me that it is so good at aentic coding. I wonder why this is. Speaking of coding, let's break this
down even more. So it says right here, at high effort on Frontier Code, it scores 10 points higher than Sonnet 5 at the same setting. And here's the crazy part, at about 115th of the cost per
task. So, the cost of this is going down significantly. And the craziest part about this is Epic Games wrote this quote about Sonnet 5.5's coding. They said in Epic's early testing, Claude
Sonnet 5.5 cleared the same quality bar you'd expect from a higher tier model. And this isn't just coming from somebody vibe coding a game and showing a demo on Twitter. This is coming from the company
Epic Games is pretty crazy if you really think about it. The director of machine learning at SpaceX AAI even said that Cloud Sonnet 5.5 delivers frontier level performance on Cursorbench 4.0 at 55%
second only to Opus 5.5 that just dropped. We think it will be a hit with developers looking to balance performance with cost. And again, this is the main overarching theme here. It's
not necessarily about how capable these models are because they are very capable. They have been for quite some time, but can they do it at a price that's not going to break the bank for
everyday consumers like you and I? Anthropic did say that it's about 30% faster than Cloud Sonnet 5. So here is a live breakdown of how long it goes from giving it a prompt to writing the code
to actually generating a HTML file right here. So um looks like obviously Cloud Sonnet 5.5 is burning through these tokens much faster. This is very fast when it comes to coding. And let's wait
and see the output. So looks like Sonnet 5.5 is done. This little animation is now working with these like fish swimming around in this HTML page. whereas we still have Claude sonnet 5
over here coding this. It does seem to be much faster than Sonnet 5. Now I'm going to make a prediction right now. So OpenAI tomorrow has dev day and I have some pretty high conviction that they
are going to be releasing their own AI agent AI personal agent similar to Muse and Grockbot tomorrow. And if history repeats itself, which it does, these AI companies like to play tugofwar. when
one company drops something, the other will drop a very similar release just a couple days later to take all of those eyeballs and put it onto their own products. So, in my opinion, after
OpenAI drops this tomorrow, Enthropic is probably going to drop theirs later this week or maybe next week. And if you connect the dots, this makes sense because Enthropic is dropping models
like Opus 5.5 and now Sonnet 5.5 that are much better at communicating. And this is perfect timing before dropping a personal assistant agent that people are going to be using to communicate with
over and over again. Again, this is just my two cents. Take it with a grain of salt. Let me know what you think in the comments. Do you think that this is the case? Cuz I have some high conviction
that this is what's happening. Now, the prompt that we are going to be using is this. Basically, we're going to generate a three-dimensional world kind of similar to Minecraft for example. And we
have some hard constraints here. We break down specifically what we want the world to be in a bunch of different steps. We actually broke down how to work. So we wanted to plan the
architecture before writing the code, keep the cord code organized and so on. So this is a prompt that we're giving every single one of these models. So let's go ahead and test those out right
now. All right. So here is the one that Sonnet 5 generated. We're going to go through each of these incrementally. So let's start with this one. All right. So we have this cool little game here.
Isn't anything crazy. It's kind of like a Minecraft styled world. I don't think this is great, but obviously we can move around here. I think our goal is to actually grab these elements. But yeah,
so what we need to do is come here and I think that we need to grab these little things. And we don't I don't know necessarily what our goal is. I think that we have to take it somewhere, but
I'm not exactly sure. You can see that there's snow falling even though there's no clouds. We have a little lake over here. And let's see if I could go in the lake. Oh. Oh, guess I can't go in the
lake. I could go under it. So, it's glitching quite a bit. Yeah, this isn't great. I will say I think the other outputs are going to be much better. Let me come over here and try to grab this
little orb. Yeah, I don't really know what to do with this, but I mean, this is a cool starting point. This isn't bad. Now, in terms of pricing for Sonic 5, let's break this down along with how
long it took to generate. First of all, it was $159 to generate, which is very cheap compared to these other models. You're going to see it gets much more expensive than that. And then for time,
it took 7 minutes and 6 seconds, which is pretty fast in the grand scheme of things. And one thing I do want to mention is that we have all of these on effort level high. So, they're all
running on the same effort level, just for transparency there. All right, moving on to Opus 5. We have this little world here. So, I'm just going to click on begin. Obviously, we can move around
with our mouse and whatnot. Okay, this one looks really good. This is Opus 5.5, which is dropped um last week. This looks much better than Sonic 5. I'm not going to lie. I think that this actually
looks really good. I could even throw the rocks. I could come over here and pick up this little orb, which I think is actually the goal of this game. I just need to find a place to bring this
to. Yeah, this looks really good. Honestly, I think that the the world is very consistent. Basically, like the objects aren't moving. before with the previous one, you could see that like
the lake I could go underneath the lake. Whereas this one, I think I could actually go inside of it. So, let's just go over here and test this out. Yeah, so I could kind of go in the lake even
though like does glitch a little bit when I'm inside of it. This is still a level up from the previous one. Not anything crazy, but I think the graphics looks really good. And then the movement
obviously with this character is very good as well. And I could just throw this little warp. So, yeah, I'm pretty impressed with this one. And you can see now fog is rolling in. I mean, this is
basically like Minecraft. This is pretty cool. Anyways, this is great. Let's go ahead and rate this one cost and the time it took. All right, so Opus 5.5 costed, drum roll please, $11.95.
So this was about like 10 times or whatever the breakdown is. I'm not smart with math, but about 10 times more expensive than Sonnet 5. And this took 41 minutes and 47 seconds. This took
quite a long time in comparison to Sonnet 5, which I guess makes sense cuz it's a much more intelligent model. Let's move on to Fable 5.1 and test this one out. Let's enter the world for Fable
51. Click to begin. And here we go. Okay, same thing. We come over, pick on, pick up the rock. This looks great. The shadow of the rock looks pretty cool. My character moves around pretty easily. It
does look like it's glitching a little bit when I'm trying to walk here. So, for whatever reason, you can see the screen shaking a little bit. I like that it added these little like this little
ruins here. Yeah, for some reason doesn't look great in terms of like the motion. Let's see. I could also think I could throw the rock. Yeah, able to throw the rock. We have our little orb
over here. Let's test out the water real quick after I pick this back up. So, we come over here, it come inside of the water. Uh, looks like it doesn't glitch inside the water like the previous two.
So, this is good. Yeah, this one's pretty solid. And one thing that's cool is if you see on the top left of my screen, it shows when the sun is going down. So, it looks like afternoon is
coming. Shows the time moving incrementally. Yeah, this one's pretty solid. I think this is one of the better ones we've generated so far. All right, so for cost breakdown on Fable 5.1,
let's take a look at this now. So, for cost, this was $1745, which is much more expensive than Sonnet 5. Obviously, it's about $6, $55 more than Opus 5.5 as well, which is to be
expected. It's a more capable, more intelligent model. It's their highest level when it comes to intelligence. And then we now have 40 minutes and 30 seconds, which is interesting. This
actually took less time than Opus 5.5. Still, it took quite some time. 40 minutes and 30 seconds compared to Sonnet 5 which was only 7 minutes and 6 seconds. For our final test, we have the
output from Claude Sonnet 5.5. All right, so here is what our world looks like. I think that this looks incredible. Actually, I personally think this might be the most like the best
looking world. Obviously, we could move around and there's no like weird movements going on. We have these little bunny rabbits hopping around. Uh rain is rolling in right now. I could even go
into the water right here. There's fish swimming around. I mean, I'm genuinely pretty impressed with this. This is the supposedly the cheapest model out here aside from like Claude Opus 5. I believe
it's the same exact price, but like this looks better in my opinion than the Fable 5.1 output. This is pretty insane. I could throw that. Um, you can see that it's getting darker outside. This world
looks pretty good. And again, the the movements here are great. There's nothing wrong here. So, yeah, this is pretty crazy. This is honestly my favorite output that we have. So, let's
go ahead and look at the pricing and the amount of time it took to generate this. All right. So, in terms of cost, this only costed $92 compared to Opus 5.5, which is $11.95.
Fable 5.1 is $1,745. And again, I think this is probably the best output that we got here for very cheap comparatively. Then, if we come over to time, this was generated in 36 minutes and 43 seconds.
This is the second quickest out of all of these. There's a significant jump from Sonnet 5 to Sonnet 5.5 and then these other models here. This is very good for the amount of time it took to
generate this and the cost output. I'm very very impressed. And just to show you again the outputs here in case you don't remember, this is what Sonnet 5 generated. Looked very, very bad in
comparison to the other ones. We have Opus 5.5. It looked pretty solid. Nothing crazy here. And then we have Fable 5.1. Looked very, very good, but the movements were quite off. And then
Sonic 5.5, which is the best overall one. The colors are very bright. There were bunnies hopping around. It looked very dynamic and the movements were perfect with this output here. Anyways
guys, there we have it. This is the cost breakdown on all these different models. I know that this might not be the most practical thing. Not everybody's going to be generating video games with these
models, but this is a good way to test how capable they are with a complex coding task like this, seeing how long it takes and the cost breakdown. So, next up, we have a brand new update from
the Grockbot team. We now have Grockbot teams, which are specific AI agents we can use and work across our team with. I personally think that the SpaceX team is crushing it with these new Grockbot
releases. I've been loving this product, and I personally like this over chatbt dots, which everybody's losing their mind about. So, let's go over very quickly what this new Grockbot teams
looks like so you could fully understand this and begin using it yourself. Now, this is a shared bot or agent that has private individual chats that you could share with your entire team. In their
announcement [clears throat] article here, they said, "Although the bot is shared, each person's conversations with it remain private. The bot keeps separate context and memories for each
user while drawing on the skills shared across the team." So, we could build skills that our entire team can use across these team Grock bots. But, of course, all of the individual context in
previous conversations stays with each individual user. Teams can also collaborate with a team bot inside of Slack. Everyone can ask questions, contribute context, and see its
responses. And I want to quickly show you this visual so it could help you understand this a little bit more if this concept doesn't fully make sense to you. So on the left we have our team.
These are physical people that are working inside of our company for example. Each of these have their own private chat or shared Slack conversation. Then we have in the middle
our shared team bot. And inside of this bot we have a few things. We have our shared files, instructions and skills. So that way this shared team bot can do specific tasks for us. And then this
team bot is connected to our different tools. It's using our different apps and plugins, thirdparty APIs, and any approved data that we give it. So think of this as a hub that your team can work
across for specialized tasks. You could have specialized team bots as well. So a good example of this is a marketing bot. This is one of the examples that XAI gave us for Grockbot. This is a bot that
our entire team can use and conversate with to review drafts against your brand guidelines, prepare something like website changes and a preview and can give us the final sign off so it would
message us before going and posting this website for us. I personally think this is a great addition. I think the Grobbot team is adding some really good features to this product to make it probably the
best go-to personal AI assistant on the market. The main problem with it right now is it is bound to having to use the Grock models which aren't necessarily on the frontier of model intelligence, but
I really do like this feature and I'm excited to test it out and share my experience on this channel. Now, we do have another interesting release from Grogbot and they introduced something
called the primary bot. This is a description as well as the card we're going to see when we log into Grobbot now and it's going to have us create or use an existing bot as our primary bot.
So, right here it says primary bot does your work proactively. that is your go-to for everyday tasks, unblocks your work, and checks in when you need help. I'm yet to configure this, but I'm going
to do this right now so you guys can check this out. But from my assumption initially after seeing this, it feels like this is our chief of staff Grockbot that a lot of us have already
preconfigured. For me, I have a chief of staff or main agent that I speak to that then delegates specialist tasks to my other agents. And so it sounds like that is what this is to make it easier for us
just to interact with one specific agent instead of having to go to our individual ones. And that can get a little bit messy. All right. So inside of my account, all I did was click on
this create primary bot button. And then it brought me right here. There's this new little Grockbot icon. It says, "Hey, I'm Grockbot. I'm new here. I'm here to be your primary agent. I can manage your
other bots. Task me with anything. I'll make sure it gets to the right bot." Now, in order to test this out, let me just say, "Hey, I want you to go and pull some information from the demo
brand partnerships agent." And what it's going to do is now communicate with my other agent. And here we go. It communicated with my brand partnerships demo, which is my fake agent. This is
not real information here for demo purposes, but it went and communicated with it. So, if you already have a chief of staff or main agent, this probably won't be valuable to you, but they did
add this feature to make it simple for people to set up this main agent themselves. What a crazy week of different AI product releases and news from all these different AI companies.
If you guys enjoyed this video, subscribe for more like this. I do this every single Friday. Let me know what you think in the comments and I'll see you in the next