Anthropic Ships Claude Motion and Dashboards, Plus the Rest of This Week in AI
A weekly AI news roundup covering two new Claude artifact types, Claude inside Google Workspace, a 90%-cheaper Haiku model, Grok Bot's switch to multi-model routing, and ChatGPT's new interactive UI.
Posted
yesterday
Duration
Format
Talking Head
educational
Views
88K
1K likes
57 · 43
Big Idea
The argument in one line.
This week's AI releases show assistants increasingly picking their own output format and even their own backend model, letting Grok Bot swap in Claude Opus 5.5 for hard tasks without the user ever choosing a model.
Who This Is For
Read if. Skip if.
READ IF YOU ARE…
You use Claude, ChatGPT, or Grok Bot for work and want a fast rundown of which new features are worth testing this week.
You build reporting or dashboards and want a quicker way to turn raw data into something presentable without opening a design tool.
You're weighing AI agent platforms against each other and want to see how model-routing changes what each one can actually do.
SKIP IF…
You want an in-depth tutorial on any single tool — each release gets a few minutes of hands-on testing, not a deep dive.
You don't currently use Claude, ChatGPT, or Grok Bot — most of this won't be actionable yet.
TL;DR
The full version, fast.
Anthropic released two new Claude artifact types this week: Motion, which turns a CSV or report into a short code-built animation (not an AI video model, so every number stays editable), and Dashboards, a live-data artifact template available on every paid plan. Claude also launched inside Google Docs, Sheets, and Slides, and Haiku 5.5 arrived with up to 90% lower pricing plus a monthly API credit on Max and Team plans. Grok Bot's biggest update lets it route tasks to outside models like Claude Opus 5.5 instead of only its own, while also gaining live X search. ChatGPT shipped GPT-6 with Intelligent UI, building interactive charts and tools instead of plain text answers.
Free for members
Chat with this breakdown — free.
Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.
Cold open teasing all five releases covered in the episode: Claude Motion, Claude Dashboards, Claude for Google Workspace, Grok Bot's model-routing update, and GPT-6's Intelligent UI.
00:28 – 03:41
02 · Claude Motion
Anthropic's new artifact type turns a CSV, report, or product walkthrough into a short code-built animation. Demoed live on YouTube audience-retention data and a 'growth city' visualization built from a follower-count tracker.
03:41 – 06:21
03 · Higgsfield Katana
A third-party MCP tool, Higgsfield Katana, connects to Claude and claims to be powered by Claude Motion, producing more polished motion graphics. Demoed by feeding Claude a reference video plus a written shot breakdown; generation took about 20 minutes.
06:21 – 09:22
04 · Claude Dashboards
A new artifact template, available on every paid Claude plan, that connects to live data and builds an auditable dashboard with visible queries and refresh timestamps. Demoed by connecting Blotato social data into a live 'Social Media Scoreboard.'
09:22 – 11:39
05 · Claude for Google Workspace
Claude now opens in a sidebar inside Google Docs, Sheets, and Slides and works directly on the open file, defaulting to an 'ask before edits' review mode. Demoed answering a growth question against a live Google Sheet.
11:39 – 13:36
06 · Claude Haiku 5.5
Anthropic's cheapest, fastest small model drops input/output pricing up to 90% and jumps its computer-use score from 15.7% to 72.4%. Max and Team plans also get a separate monthly API credit that must be enabled in billing settings.
13:36 – 20:40
07 · Grok Bot Can Now Use Other AI Models
Elon Musk announced Grok Bot will use the best backend model for any task, including Claude Opus 5.5, Midjourney, and Suno, instead of only Grok's own models. Breaks down three reasons this matters and four things still unknown about how it will work.
20:40 – 21:52
08 · Grok Bot Can Now Search X
Grok Bot can now search, read, and monitor the X platform automatically, without the user first connecting an account. Demoed pulling and ranking the top reactions to a ChatGPT announcement.
21:52 – 25:25
09 · GPT-6 Intelligent UI
ChatGPT's new GPT-6 release decides on its own when to answer with an interactive interface, like sliders, charts, or buttons, instead of plain text, available to every plan including free. Demoed with a compound-interest calculator and an interactive Texas Hold'em practice hand.
25:25 – 25:36
10 · Final Thoughts
Sign-off and weekly-series subscribe ask.
Atomic Insights
Lines worth screenshotting.
Claude Motion builds animations as editable code instead of AI-generated video, so changing one number in the source data updates the whole animation without a new render.
Claude Dashboards is a new artifact template, not a new product category — Anthropic says so itself in its own help center.
Haiku 5.5 cut Anthropic's small-model pricing 90%, from $1/$5 per million input/output tokens to $0.10/$0.50, for prompts under 100,000 tokens.
Haiku 5.5's computer-use benchmark jumped from 15.7% to 72.4%, a bigger capability leap than the price cut alone suggests.
Max and Team Claude plans now include monthly API credit that does not roll over: $100 on Max 5x, $200 on Max 20x, up to $500 on Team.
Grok Bot no longer runs only on xAI's own models — it can now route a task to Claude Opus 5.5, Midjourney, or Suno automatically.
Every prior agent platform has been locked to its own company's models; Grok Bot is the first major one to open that up.
Claude for Google Workspace defaults to 'ask before edits,' surfacing every proposed change on a card before it touches the document.
GPT-6's Intelligent UI decides on its own whether to answer in text or build an interactive tool, without being asked to visualize anything.
Four basic questions about Grok Bot's new model routing are still unanswered as of this release: timing, who picks the model, task-to-model logic, and pricing.
Takeaway
Five AI Platforms Shipped Convergent Upgrades This Week
AI RELEASE RECAP
Five separate AI updates this week point the same direction: assistants that pick their own output format and, increasingly, their own underlying model, so the tool you talk to matters less than what's running behind it.
02Claude Motion
Claude Motion generates animations as code rather than AI video, so you can open the result and change any word, number, or timing without regenerating the whole thing.
As of this release, Motion is limited to Claude's Team and Enterprise plans, so check your plan before expecting it in a standard Claude account.
Feeding Motion a CSV of real data, like a retention curve or growth tracker, turns a spreadsheet into a shareable video explainer in a few minutes with no editing software.
03Higgsfield Katana
Higgsfield Katana connects to Claude through an MCP integration and produces noticeably more polished motion graphics than Claude Motion alone, at the cost of a roughly 20-minute generation time.
Feeding a reference video into Claude alongside a written, step-by-step breakdown of the desired result improved the output more than a single simple prompt did.
Third-party MCP connectors are becoming a real differentiator: the base model handles the request, but the connector decides how polished the final output looks.
04Claude Dashboards
Claude Dashboards ships on every paid plan, Pro through Enterprise, unlike Motion, which is currently Team and Enterprise only.
A dashboard is technically still a Claude Artifact built from a dedicated template, so the underlying capability existed before this release, just not in this packaged form.
Each chart in a generated dashboard shows the exact query behind it and when the data last refreshed, making the output auditable instead of a black box.
Connecting a dashboard to a live data source keeps the numbers current without re-asking Claude for an update each time.
05Claude for Google Workspace
Claude for Google Workspace works inside the file you already have open in Docs, Sheets, or Slides, removing the old copy-paste-copy-back loop.
It defaults to 'ask before edits,' showing every proposed change on a card before touching the document, worth leaving on for anything you didn't write yourself.
In Sheets it can read a full dataset and do downstream math on it, like turning a subscriber count into a daily growth rate, without you writing the formula first.
06Claude Haiku 5.5
Haiku 5.5 cut input pricing 90%, from $1 to $0.10 per million tokens, but only below a 100,000-token prompt; past that it jumps to $0.50 in and $2.50 out.
Its computer-use benchmark score rose from 15.7% to 72.4%, which matters more for repetitive, agentic tasks than the price cut alone.
Max and Team plans now include separate monthly API credit, $100 to $500 depending on tier, that has to be manually enabled in billing settings and does not roll over.
07Grok Bot Can Now Use Other AI Models
Grok Bot is no longer limited to xAI's own models — it can now route a given task to Claude Opus 5.5, Midjourney, Suno, or other leading APIs automatically.
This removes the old tradeoff of choosing an agent for its interface versus a model for its reasoning power, since one chat can now use both.
Treating the agent, meaning the interface, memory, and scheduled tasks, as separate from the model underneath means a better model can be swapped in later without switching platforms.
08Grok Bot Can Now Search X
Grok Bot can now search, read, and monitor X without the user first connecting an account, a step that was required when the X connector launched in August.
A demonstrated use case: setting an agent to monitor specific accounts and alert on new posts works as a lightweight news-monitoring tool.
09GPT-6 Intelligent UI
GPT-6's Intelligent UI decides on its own whether a question deserves plain text or an interactive interface, sliders, counters, charts, without the user asking for a visualization.
This feature is available on every ChatGPT tier including free, though free users get a lighter model while paid users get the full one.
As of this release, Intelligent UI only applies in normal chat, not inside ChatGPT's work or coding surfaces.
Glossary
Terms worth knowing.
Claude Motion
An Anthropic artifact type that turns data like a CSV or report into a short animation built entirely from code, so every element stays editable and exportable as an MP4.
Claude Dashboards
A Claude artifact template that connects to live data, takes a plain-English question, and builds a dashboard showing the query behind each chart and when it last refreshed.
Claude Artifacts
Anthropic's feature for generating standalone outputs, such as code, documents, or visuals, inside a Claude conversation, separate from the regular chat text.
Higgsfield Katana
A third-party motion-graphics tool that runs inside Claude through an MCP connection and is marketed as powered by Claude Motion, producing more polished animated output than Motion alone.
MCP (Model Context Protocol)
A standard that lets an AI chat interface connect to an outside platform's tools and data, used here to link Claude to Higgsfield and Blotato.
Grok Bot
xAI's AI agent platform that logs into a user's tools, runs scheduled tasks, and can now route individual tasks to outside models instead of only Grok's own.
Intelligent UI
A GPT-6 feature where ChatGPT automatically decides whether to respond with plain text or an interactive interface, such as a chart, form, or tool.
“the model behind all of that is just a part that can be swapped out”
crisp one-line thesis for the whole Grok Bot segment→ TikTok hook↗ Tweet quote
15:42
“I personally really like how Grok Bot communicates with me... it actually remembers how I like to work.”
personal, specific preference that explains the agent-vs-model distinction→ IG reel cold open↗ Tweet quote
05:37
“This is not After Effects. No timeline, no keyframes.”
tight reaction line reading the Higgsfield output on screen→ newsletter pull-quote↗ Tweet quote
24:39
“Instead of people adapting to software, software will adapt to people.”
OpenAI's own framing of Intelligent UI, quoted verbatim, works as a standalone thesis→ newsletter pull-quote↗ Tweet quote
The Script
Word for word.
Read-along
Don't just watch it. Burn it in.
See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.
17px
metaphoranalogy
We have yet another big week of AI releases, and in this video, I'm going to break them all down for you. Anthropic just dropped two new products called CloudMotion and Dashboards. On top of that, Cloud for Google Workspace was released, and Grokbot is no longer limited to Grok's own models, and there are three reasons why I think this is such a massive deal.
As well as ChatGPT now has a completely new way to interact with it called Intelligence UI. I want to make this video quick and to the point, so let's dive right into it. The first thing I want to cover in this episode is clod motion.
Basically, you can give clod something like a report, a chart, or even a product walkthrough, and it will turn it into a short animation that plays just like a video. The interesting part about this is it does not use a video model. Quad actually writes the entire animation as code, so there's no AI -generated footage.
And because it's all code, you can go in and change any word, any number, or even the timing, and the rest of the animation stays exactly the same. Then, once you're happy with it, you can even export it as an MP4 file. As of right now, Motion is only available on the Team or the Enterprise plans, so I actually went ahead and signed up for a Team plan just so I could test this out for you in this video.
but I would assume that they'll probably be rolling this out to the other plans in the coming days or weeks. All right, so first things first, I'm inside of my Cloud account right now. And what I'm gonna do is I'm gonna come over to the left -hand side where it shows artifacts right here.
And now we're gonna see a couple new things. So inside of artifacts, we are going to see dashboards, which I'm gonna show you here in a little bit. And then we also have motion.
These are two new artifact types that we can create directly inside of Cloud. So if I click on this motion one, I can now go and begin. creating a video animation with my specific data.
Now what I'm going to do is I'm going to give it this prompt and I'm going to explain exactly what these CSV files are right now. Basically what this is is this is my audience retention data for one of my YouTube videos. What I'm going to have this do is turn this into a 30 second explainer where I could basically see where people are clicking off my videos and I can go and share this with my team for example if we want to like optimize retention.
So I'm just going to click send and as you can see right here it's analyzing the drop off via this CSV file so we could upload any data that we want to use in order to create one of these interactive dashboards but also motion designs that could help us explain more complex data. Now a few minutes later we actually got this output right here so let's go ahead and play this.
It says where viewers leave, it shows the actual graph, it shows when people are clicking off. what percentage of people are still watching, and then it shows the entire retention graph right here. Before, this was just a CSV file, so it took that boring data in CSV format and turned it into this interactive video that I can go and share with my team in a meeting, for example.
Now, here's another example. Basically, I fed it this following tracker CSV file where I basically track all of my followers and growth on YouTube and my other channels, and I told it to create a city. And what it did is it - they created this fully animated visual of this city showing what this looks like in terms of my YouTube growth.
So you could see month over month, it shows how much my channel has grown. And it gives me this cool little visual that kind of like explains it in a real life use case. It shows how big this city has grown, which the city is obviously just my YouTube channel.
I think this is a pretty cool visual that we can now create with this. These are just two use cases that I'm personally using, but... I'm sure you have a bunch of different data you could use to visualize with Cloud Motion.
Now, as you can see, Motion on its own is really good for the basic use cases like explaining your data. But if you want to create much better motion graphics, you could actually do it with a brand new tool from Higgsfield called Katana. Higgsfield says it's powered by Cloud Motion, and it runs directly inside of Cloud through the Higgsfield MCP, which is basically just a connection to the Higgsfield platform.
So all you have to do is connect Higgsfield to your Cloud account, slash Katana in the chat. So let me show you exactly how this works.
All right, so I want to show you this quick demo that I actually want to use as a reference video that we're going to use with Claude Motion in the Higgsfield Katana model. So as you can see right here, there's this really cool launch video that Higgsfield had on Twitter. And this apparently was all generated with the Higgsfield Katana MCP.
And I want to see if I could actually recreate something like this. I mean, this looks really cool. This looks much better than the normal just Claude Motion graphics.
So what I'm going to do is I'm going to feed this video as a reference into Claude and try to recreate something similar. Now the first thing we need to do in order to generate this inside of Claude is to connect the Higgs field MCP, which allows us to generate these visuals in motion graphics without leaving the Claude platform.
So what I'm going to do is come over here. There will be a link in the description for you to come to this exact page. We're just going to click on connect Higgs field.
Going to click add custom connector. Click continue here. I already have mine set up, but you're just going to go through this very simple setup process.
Click continue, and then you'll have to go and approve your account inside of Pigsfield, and it will bring you back to Claude. Next, what I'm going to do is paste in this prompt right here inside of my Claude chat. Basically, I just said use the Katana feature inside of the Higgs field MCP, use the attached video as a reference style.
And then I actually had Claude break down exactly step by step how I want this video to look as well in case it's not able to pull the full reference video. So I just linked the reference video right here. And let's go ahead and click send.
All right, so we now have the output from Higgs field in Claude. And I'm not going to lie, this did take quite a while. Took about like 20 minutes to generate this visual.
But let's take a look at it right now. All right, here we go. It says.
watch this. This is not After Effects. I mean, this actually looks pretty good.
No timeline, no keyframes. Connecting Claude to Higgs field to Katana. Yeah, I mean, I think that this looks actually really solid, much better than any of the outputs that I've gotten so far when using just the Claude motion design.
I mean, yeah, this is very good. Under the hood, it's all code. It even shows this code background.
And I guess we could have added some voiceovers to this as well, but right now it's just the visuals. So that is pretty good. I'm very happy with the output here.
And again, this generated in about 20 minutes, but I only gave it one very simple prompt. So if I want to give it... an even more enhanced prompt, or I want to go back and forth with this quite a bit.
I'm sure we could fine tune this even more, but I'm very happy with this initial output. Now let's talk about cloud dashboards. And this is one on every paid plan.
So that's pro users, max team and enterprise plans, unlike cloud motion. So how this works is basically you just connect your data, you ask it a question in plain English, and cloud will build an entire dashboard for you. And Anthropic says that it stays up to date as your data changes.
So this means that all of the backend data that is connected will update on the dashboard in real time. Now, I'm not going to lie, when I first saw this release, I was a little bit confused because it feels pretty similar to Cloud Artifacts, which we've had for quite a long time, where Cloud could already build you a dashboard inside of an artifact.
So I went and looked into this and a dashboard actually is an artifact and Anthropic says this themselves in their help center. It's just built from a new dashboards template. So the real difference is what a template adds on top.
Every chart shows you the query behind it. So you can click on any number and see exactly how Claude got that data. And every chart also tells you when the data was last refreshed.
So you know how fresh those numbers are. So this all looks great, but I wanted to go and test it out myself. So let me show you exactly how it works.
And if we come over to artifacts, we're going to also see. cloud dashboards right here, which I could go ahead and toggle on right here. Or I could just go into a new chat and say, hey, I want you to generate a dashboard for me.
So what I'm going to do is I'm just going to come here. I'm just going to say, hey, I want you to generate a dashboard that visualizes all of my Instagram data using the Blatato MCP. I want to be able to track which posts I have, the analytics of those, how many views they got, how many comments, etc.
So I'm going to send that off now to Claude and it's going to go and generate this artifact dashboard right here. And if you don't know what Blitato is, this is a platform that I use to help me publish and set up DM automations for my Instagram. So I could come in here and take a look at my inside of my account, or I could just have Claude now spin up a dashboard for me where I could see all this information without me necessarily leaving the Claude.
platform. So if you want to try out Blotato and set this up yourself, there will be a link in the description for you to do so. As you can see, it creates these Instagram posts for me automatically where I have Claude generate the graphics right here and it will automatically post it and set up the DM automations for me.
Now, just like that. A few minutes later, we have this social media scoreboard right here, which basically pulls all the information that I just showed you inside of Blotato. So it shows the amount of views that I've had over the last 30 days on the specific posts I've scheduled through Blotato.
It shows top posts by views. And as you can see here, there's a table of all this information. And the main part that I want to highlight here is at the top right of this, it shows updated three minutes ago.
We could click refresh. It's now going to pull live data from this dashboard. And we could go and share this with a team member.
or we could export this as code as well. So I think this is a pretty cool feature, but it does feel like it is a cloud artifact, which it is, but now it's almost just a template that you can use to create dashboards for you automatically. Anthropic just released Cloud for Google Workspace.
So Cloud now lives right inside of Google Docs, Sheets, and Slides. It opens up in a sidebar right next to your file and works on the file that you already have open. So quickly think about how you would normally do this before.
You'd copy your spreadsheet, paste it into Claude, get an answer, and then have to paste it all back in. Now, you could ask it right there and it just makes the changes for you.
In Google Docs, Anthropic says that it can fix a sentence or restyle a heading in place without touching the formatting around it. In Sheets, it could write formulas, build pivot tables and charts, and even add new tabs. And in Google Slides, it could build out new slides that match the rest of your deck.
Then it checks its own work and flags text that's hard to read or anything that's running off the slide. So I want to - and tested this out myself.
In order to set this up, you just go to the Cloud for Google Workspace page, click Try Google Workspace, and install it. Then inside of any Google Sheet, you go to Extensions, then Cloud. then open up Claude.
And from here, it opens right on the side of your sheet. You can even pick which model you want to use. And on top of this, we can even choose the effort level we want to use.
So I opened up a Google spreadsheet where I track my followers and I asked it what my growth of my YouTube channel has been since I started tracking this. It read through all the data and it told me the channel is up about 82 ,792 subscribers since I started tracking, which is about two and a half times. And it even did the math and worked out that that's about 333.
new subscribers every single day. And it told me that it only answered in the chat and didn't change my sheet, but it offered to add a summary with the formulas if I wanted to add it to my Google spreadsheet. Now, that's how you use it in Google Sheets, but it works the exact same way inside of Google Docs.
Now, the part that I like is how it handles the edits. By default, it's in a mode called ask before edits. So every change shows up on a card with a summary and waits for you to approve it because you don't want to make these changes to your spreadsheet or your Google Docs without you approving it.
It could go and make some bad changes. And if you don't want to have to accept everything, you could switch it over to accept all edits, and it just works through the whole task without stopping, just like your normal cloud would.
Now the question is, who gets access to this? As of right now, it's on every paid cloud plan. So that's pro users, max, team, and enterprise.
Next up this week, we have a new release from Anthropic, and that is Cloud Haiku 5 .5. Anthropic is calling it the cheapest, fastest, and most capable small model they've ever released. And if you're not familiar with the lineup, Haiku is the smallest and the cheapest Claude model.
And this is coming right after Anthropic dropped both Opus 5 .5 and Sonnet 5 .5, both of which have been extremely impressive after testing. So I've been pretty excited to try out Haiku 5 .5 myself. Now, the biggest thing here is the price.
For input tokens, the old Haiku was $1 per million and Haiku 5 .5 is only 10 cents. For output tokens, it went from $5 down to 50 cents. So if you do the math, this is nine.
90 % lower. Now there is one catch that I do want to flag. That 10 cent price only applies to prompts up to 100 ,000 tokens.
Once you go over that, it's 50 cents in and $2 .50 out. It's obviously a plus that this is much cheaper, but it also got quite a lot more capable too. On computer use, the old Haiku scored 15 .7 % and Haiku 5 .5 scored 72 .4%.
Now, Sonnet and Opus are still the better choice for the harder stuff and Anthropic says it's themselves, but Haiku 5 .5 is good for those tasks that you run over and over again and don't necessarily need a high level of intelligence from these different models like you'd get from Opus or Fable. Now there is a part to this release that I think most people are going to miss.
If you're on a max or a team plan, you now get a monthly API credit. That means you now get $100 a month on the max 5x plan, $200 a month on max 20x, and up to $500 on team. Now if you want access to this, you do have to it in your settings under billing and it doesn't roll over month to month as well as if you are a pro user you won't have access to this if you want a full deep dive on the cloud haiku 5 .5 model i did make a video about it so there will be a link in the description to go and check that out all right moving on to our next release and this is from grokbot and this is one of the biggest updates they've shipped since it launched and i'm not gonna lie i am extremely excited for this i think it solves the biggest problem with grokbot this move from spacex puts some serious pressure on the other ai companies like OpenAI, Meta, and Anthropic so let's break this down right now.
Elon announced that Grokbot is no longer limited to Grok's own models. That means it's going to start using top AI models like Cloud Opus 5 .5. So basically how this works is it gets the best model for whatever task you give it and it's going to pick that model for you automatically.
Now there are three reasons why I think this is such a big deal and the third one is the biggest. On top of that there's also a few important things that we don't know yet and I'm going to get to those right after. So here is exactly what Elon posted.
And I'm going to read it for you word for word, and then we'll break this down. Going forward, SpaceX will use the best backend model for any given task, including Cloud Opus 5 .5, MidJourney, Suno, and other leading APIs. Whatever is most likely to give the best outcome.
Now, if you've never used Grokbot before, it's basically an AI agent platform that you can text like a friend or an employee. You can create multiple agents that work together and delegate tasks to specific agents. So you can text it like an employee.
It can log into your tools and keeps working even when your computer's closed. And this platform has been amazing. I've absolutely loved this.
I've been using it every single day over the last two months as my main daily driver, but there were some limitations to this. Up until today, it only ran on Grok's own models and you couldn't pick a different model if you wanted to. Well now, Grokbot is going to look at the task you give it, and then send it to whichever AI is best for that specific job.
And I think this fixes the one thing that was really holding Grokbot back. So the first reason this matters is the model itself. Grok's models have always been a step behind, or two or three steps behind, Claude and Chachabit, and even Elon has said this.
A couple of weeks ago, he said in an interview that Grok is not as good as Opus 5 .5, and that they'd probably catch up sometime next year. And so instead of waiting, until next year, he just decided to plug Opus 5 .5 and the top leading models right into Grokbot.
Now, personally, I think Opus 5 .5 is the most powerful model that you can use right now by far. And I've been building some incredible things with it over the last week. But I personally really like how Grokbot communicates with me.
I think it's much more easy to use and beginner friendly. It honestly feels like I'm texting a coworker or a friend and it actually remembers how I like to work. So up until now, I kind of had to choose.
I'd go to Claude for hard stuff like coding or more specific and complex workflows. And then I'd go to Grokbot for... all the different admin tasks and everything else that was easier for this model to handle.
Now here's a quick example of what this could look like going forward. Let's say I text my Grok bot and ask it to build me a dashboard that tracks all of my different invoices. I'm still talking to the same agent in the same chat but behind the scenes Opus 5 .5 could be the one writing the code since it's much better at that.
This way I get the agent that I like talking to in the model I trust with the hard work which is Opus 5 .5. Alright moving on to the second reason on why I think this is such a big deal, and this is that you no longer have to pick a model.
Right now, if you want the best results, you kind of have to know which AI is good at what. So you'd go to Claude for writing and code, then Midjourney for images, and Suno for music, and you get the point. And as you're probably familiar with this, that's a different app and a different subscription for all these different AI tools and platforms.
But the way that Elon worded this in this tweet, Grokbot makes that choice for you. So you could ask OneBot to write a launch post, make the image for it, and even make a little jingle song to go with it.
And then each piece of this specific task goes to each specific model that is best for it. Now, there's also a chance that this might save us some money. If Claude, Chachabiti, and Suno all end up being included in the Grokbot plan, that could be one subscription instead of having to pay for all four.
Now, I do want to preface and say that pricing has not been announced yet, so I wouldn't go and cancel anything today off. obviously. But this is something that I'm genuinely thinking about and excited for going forward in the future.
And that brings me to the third reason, which I think is the biggest one. And it's what this means for everybody else. Because in the last couple of months, pretty much every big AI company has launched their own version of Grokbot.
Meta came out with their platform called Muse, which has been taking the world by storm over the last couple of weeks. And then last week, OpenAI launched Chachapiti Dots. And I was personally extremely excited to test it out.
And so were tons of other... people, but it's honestly been extremely underwhelming so far. But here's the thing that all these have in common.
Chachapiti Dots runs on OpenAI's models, and Muse runs on Meta's models, which are very limited. So you're kind of betting that the company that built your agent is also the one with the best model, and this changes almost every week when these companies are releasing new models every week. And Grokbot was the exact same way until today.
But I want you to think about what Grokbot actually is. It's a computer in the cloud that's logged into all of your different tools. and runs your routines and scheduled tasks while you sleep.
And after this change, the model behind all of that is just a part that can be swapped out. So when a better model comes out, Grokbot could just plug it in and you wouldn't have to change a thing. And if you're anything like me, this will make your life so much easier so you don't constantly have to be switching different platforms.
And in my opinion, I think this puts real pressure on all of the other AI companies. Anthropic already has Cloud Cowork and I'd bet they have a bigger version of this coming too. by launching a competitor to Grokbot, Metamuse, Chachapity Dots, etc.
So the big question is whether any of these different AI platforms are going to follow and open up their agents to other models. My guess is they probably will not anytime soon because their models are the thing that they're selling. And this is exactly why I think this is such a big moat going forward for Grokbot.
Now, before you get too excited, there are four things we still don't know. And a couple of these are really important. The first is when this actually goes live, because all Elon said in this tweet was that this was going to be the case going forward.
So that could mean today, this week, or could mean a month from now. We don't necessarily know. But as of right now, we don't know when this is going to go live.
The second is whether you get to pick the model yourself or see which one did the work. The way that this tweet reads is that Grokbot is going to be picking for you.
And we also don't know which tasks actually go to which specific model. How is it making these decisions? So it's possible that Opus would only get used for the really hard stuff and everything else will run on Grok models.
We just don't know. And if a task comes back wrong, you don't necessarily know which model did that. And the third is the price.
Right now, extra usage on Grokbot is billed by token cost and Opus is a more expensive model to run than Grok. So there's a chance your usage doesn't stretch as far when Opus is doing more of your work.
We also don't know if every plan is going to get this or just the more expensive Grok bot plans. And the fourth big thing here is your data. Your Grok bots are logged into your emails, your files, and all of your other tools that it's working across.
So if a task gets sent to another company's model, some of that information could be going with it. And right now, we don't know if you'll be able to turn that off. I'm going to be testing this the second it shows up in my account and I'll make a full video on it because I am extremely excited for this.
We also got another big update from Grokbot and they announced that it can now search, read, and monitor the X platform. Back in August when they launched Grokbot, they added an X connector, but you had to sign in and connect your account in order to use it. Well, now you don't have to connect it to anything.
It just automatically can pull live data from the X platform. So I went ahead and tested this out myself. I basically had my Grokbot thought go and pull the top reactions on X to the new ChachiBT interface that just got announced and then give me the top 10 reactions.
It went and searched this for me and it came back with this. First, it gave me a quick summary of the announcement itself and then I get a table of the top 10 reactions with who said it, what they said, and how many likes and views each of these posts got. So this is a really quick way that we can now survey what's happening on the X platform without having to scroll through it for hours, which I personally get a lot of my different AI news.
from this platform. So this is going to save me a lot of time. A perfect example of a way I'll be using this is having Grokbot monitor X and text me whenever there's a new tweet from both Anthropic and OpenAI.
So this way I could be one of the first people to see all the new product releases from them so I could cover them on this YouTube channel. All right, moving on to the last release that I want to talk about here. And this one is from OpenAI.
They just rolled out GPT -6 inside of ChatGPT along with a brand new feature called Intelligent UI. Everybody has access to this, even people on the free plan. Here's exactly what intelligent UI means.
Up until now, when you asked ChatGPT a question, you mostly got a huge paragraph that you had to read through, and this could be extremely annoying. Well, now it could answer with an interface that you can actually interact with. It will go on and create forms, interactive charts, and even little tools right inside of the chat.
And the part that I really like is that you don't have to ask for any of this. ChatGPT looks at your question and decides what the answer should look like, whether that's text or an interactive form and it will build it for you.
And I think this is why they're calling it intelligent UI. The intelligent part is that it's making the choice for you. So it decides when a visual would actually help explain something and then it builds the whole interactive experience on its own as opposed to you having to ask to visualize something.
Now here's the example that OpenAI gave us in this release article. Somebody asks for a plan for a Sunday lamb roast and they say they still don't know how many people are coming. Well the old model would have given you a long text answer that you'd have to go and sort through.
Well, GPT -6 now gives you a little counter for the number of people. And when you change it, the whole shopping list changes with it. Then there's even a button to copy the shopping list and cooking checklist that you could tick off as you go.
They also showed a road trip where the stops show up on a map and you could just ask it to build you any tool that you want. Now, these are cool, but I wanted to go ahead and test this out myself. So here's what I had to do.
First, I asked it to help me understand how compound interest would work if I invested $2 ,500 a month into this. S &P 500. And instead of just explaining it to me, it built me this.
I get a chart of what my money could grow to with how much I put in and how much it earned on top of that. At an 8 % return over 30 years, it says that it could grow up to about $1 .75 million. And then as you can see right here, it gives me these sliders.
So I can change things like the average return or how many years I'm investing and the whole thing updates right away. Then I tried this on a second use case and I think this one's actually pretty cool. So I asked it to teach me how to play Texas Hold 'em?
And it started walking me through the rules and actually showed me the different cards in the game. It then deals me a practice hand right inside of the chat. And as you can see right here, it gave me ace king and gives me three buttons, which are fold, call, or raise.
I personally hit raise and it tells me that it was a great decision. And then it deals a flop and asks what I want to do next. The crazy part here is that it's teaching me along the way in this interactive experience that's customized based on the initial chat that I gave it.
And this is why I think this is a much bigger deal. than it looks like on the surface. For the last few years, every AI gave everybody the exact same thing, which was a text output.
You had to take that answer and go do something with it yourself or digest it and read all the way through it. And OpenAI actually said this in the announcement. They said, instead of people adapting to software, software will adapt to people.
And it feels like we're moving into a world where software is built for us all the time and customized to each individual person. And I personally think this is where all of AI is heading, where it feels a lot more personal. to you and how you like to work.
Now, the question is who gets access to this. So as of right now, this is available to everybody, including the free plan, which is pretty cool. So you could go ahead and test this even if you're not on a Chachapiti plan.
Paid users get GPT -6 Sol and free users get a lighter model called GPT -6 Luna. And as of right now, this is only in the normal chat. So nothing is changing inside of either Chachapiti work or Codex, but I'd assume they're going to roll this out over the coming days or weeks, maybe.
Anyways, if you enjoyed this episode, subscribe for more. more as I make these episodes every single week so you could stay up to date with AI without being overwhelmed. Anyways, thanks for staying to the end and I'll see you in the next video.
Cheers.
The Hook
The bait, then the rug-pull.
Another AI news week, another pile of releases to sort through: Claude picked up two new artifact types, Grok Bot stopped being locked to its own models, and ChatGPT started building interactive tools instead of walls of text. Brock Mesarich tests each one live and tells you which releases are actually worth your time this week.
Frameworks
Named ideas worth stealing.
15:36list
Three Reasons Grok Bot's Model Switch Matters
The model itself improves immediately — Grok catches up to Claude/ChatGPT by borrowing their models instead of waiting to build a competitive one.
You no longer have to pick a model yourself — Grok Bot routes the task to whichever model fits, inside one conversation.
It pressures every other AI company, because an agent platform that can swap in any model reduces lock-in to any single company's models.
The host's own breakdown of why letting Grok Bot route to outside models, not just Grok's own, is the biggest story in the episode.
Steal forAI tool comparison content or 'why this release matters' explainer segments
19:39list
Four Unknowns About Grok Bot's Model Routing
When this actually goes live, since Elon's post only said 'going forward.'
Whether the user picks the model or Grok Bot picks it automatically.
Which tasks route to which model and how that decision gets made.
What it costs, since Opus is more expensive to run than Grok's own models and billing is token-based.
A checklist of open questions the host flags before recommending anyone get too excited about the Grok Bot announcement.
Steal forAny 'new feature announced, not yet shipped' coverage that wants to separate the claim from the confirmed details
CTA Breakdown
How they asked for the click.
VERBAL ASK
08:23link
“if you want to try out Blotato and set this up yourself, there will be a link in the description for you to do so”
soft mid-roll plug for an affiliate tool tied directly to the Claude Dashboards demo just shown, not a separate scripted sponsor read
Add Modern Creator as a preferred source and Google shows you more of our breakdowns in Search, Top Stories, and AI Overviews. It only changes what you see, and you can undo it in your Google settings anytime.
Add to Preferred SourcesOpens your Google source preferences with us pre-loaded. Tick the box and you're done.
Anthropic merged Claude's memory between chat and Cowork so both remember the same things, but the sync only works if Cowork runs in the cloud, not on your own machine.
A tour of five real workflows built inside xAI's new multi-agent platform, agents that email, invoice, generate video, scrape the web, and text each other to get work done.
A hands-on first look at OpenAI's overhauled ChatGPT voice mode across five demoed use cases: booking a flight from an old chat, hunting down a file mid-task, generating and assembling a product website, and running it all remotely from a phone paired to the desktop.
00:12
00:19
00:31
00:49
01:14
01:42
01:54
02:05
02:47
03:12
03:32
04:01
04:26
04:48
05:01
05:26
05:42
06:02
06:26
06:42
07:13
07:25
07:51
08:03
08:51
09:11
09:46
10:04
10:30
10:50
11:06
11:31
11:59
12:22
12:47
13:08
13:18
13:41
14:10
14:20
14:50
15:12
15:37
15:50
16:17
16:39
17:06
17:20
17:42
18:03
18:54
19:20
19:37
19:57
20:43
21:01
21:27
21:51
21:54
22:13
22:36
23:13
23:38
23:53
24:20
24:41
24:48
25:28
25:32
25:34
10:15
22:18
15:30
09:30
20:36
09:35