Big AI News: Opus 5.5 vs GPT-6 Sol, NotebookLM Updates, and Muse Charm
A week where Anthropic and OpenAI dropped flagship models on the same day, and one weekly AI news host ran them side by side on real writing, design, and video-editing work.
Claude Opus 5.5 beat OpenAI's cheaper GPT-6 Sol across writing, design, and video-editing tests this week, and new features from Gemini Notebook, Grok, and Meta's Muse pushed everyday AI agents further into daily life.
Who This Is For
Read if. Skip if.
READ IF YOU ARE…
You switch between Claude and ChatGPT for writing, coding, or editing work and want real test results instead of launch-day hype.
You use AI to auto-edit talking-head video and want to know which model actually produces usable zooms and highlights.
You're tracking the shift toward voice-first, agentic AI tools like Grokbot calls or Meta's Muse.
SKIP IF…
You want a formal benchmark breakdown; this is one creator's hands-on impressions, not a lab test.
You don't use Claude, ChatGPT, Gemini, Grok, or Meta AI products day to day.
TL;DR
The full version, fast.
Anthropic's Claude Opus 5.5 and OpenAI's GPT-6 Sol launched the same week, so the creator ran them head to head on script writing, motion graphics, and AI video editing. Opus 5.5 won every round: a more human writing voice, cleaner auto-zooms and highlights, and an edit that finished in half the time GPT-6 Sol took, despite costing more per token. Beyond the model fight, Gemini Notebook added live voice chat and interactive reports, Grok shipped a 4.7 model plus Grokbot voice calls, and Meta's Connect keynote turned Muse into an agent with its own email, video calls, Mac computer control, and a wearable called Muse Charm. The takeaway: Opus 5.5 is now the default daily model, and agentic AI keeps pushing into voice, hardware, and shopping.
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.
Quick teaser of the week's stories: dueling Opus 5.5 / GPT-6 Sol launches, Gemini Notebook, Grok, and Meta Muse.
00:39 – 01:09
02 · GPT-6 vs Opus
Paul frames why both labs shipped a model the same day and sets up a live side-by-side test at high effort level.
01:09 – 04:52
03 · Writing test
Both models research and script a YouTube video live; Opus 5.5's writing reads more human and wins the round despite finishing slightly slower.
04:52 – 06:05
04 · Design test
A blind three-way motion graphics comparison against Fable 5.1; Fable wins, Opus 5.5 is a close second, GPT-6 Sol disappoints.
06:05 – 07:05
05 · Omnisend
Sponsor read for Omnisend, an email/SMS platform that now connects to ChatGPT and Claude through MCP.
07:05 – 10:36
06 · Editing test
Both models auto-edit the same raw tutorial footage; GPT-6 Sol misjudges click positions and zooms, Opus 5.5 reproduces the intended edit accurately and faster.
10:36 – 11:37
07 · Results
Paul delivers his verdict: Opus 5.5 is cheaper-feeling in practice, wins on quality, and is now the best model he's used, edging out even Fable 5.1 overall.
11:37 – 14:54
08 · Gemini Notebook
Live voice chat for Ultra subscribers, notebooks as an @-mention source inside Google Docs, and new interactive reports with inline generated assets.
14:54 – 16:46
09 · Google updates
A pre-orderable Google Book laptop built around Gemini Intelligence, 13 new app integrations, and a new Gemini 3.8 text-to-speech model.
16:46 – 20:54
10 · Grok updates
Grok 4.7 launches to mixed reactions; Grokbot adds local traffic routing, proactive voice memos, and per-bot voice calls.
20:54 – 24:43
11 · Meta Muse
Meta Connect's Muse updates: its own email address, live video calls, Mac computer control, expanded shopping partners (with Amazon blocking access), Ray-Ban Meta Audio glasses, and the Muse Charm wearable.
24:43 – 25:39
12 · ChatGPT updates
The ChatGPT desktop app adds Chrome extension support and multiple connected accounts per plugin; ChatGPT Voice can now use plugins and track credit scores.
25:39 – 26:15
13 · Final thoughts
Paul names Opus 5.5 as his new go-to model and closes with the week's subscribe ask.
Atomic Insights
Lines worth screenshotting.
Claude Opus 5.5 beat GPT-6 Sol on script writing, motion graphics, and AI video editing, even though GPT-6 Sol costs half as much per API token.
Independent intelligence benchmarks put Opus 5.5 about 10 points ahead of GPT-6 Sol despite GPT-6 Sol's lower price.
Opus 5.5 edited a talking-head video in 7 minutes 15 seconds; GPT-6 Sol took 15 minutes 30 seconds for the same task.
GPT-6 Sol's AI video edit zoomed into the wrong spot on screen and misjudged where the creator had actually clicked.
Both ChatGPT and Claude tasks in this test used under 1% of the creator's weekly usage, making both effectively cheap to run.
In a blind three-way motion graphics test, Fable 5.1 ranked first, Opus 5.5 second, and GPT-6 Sol last.
Gemini Notebook now supports live voice chat for Ultra subscribers, letting you interrupt the AI mid-answer.
Notebooks can now be referenced directly inside Google Docs with the @ symbol, feeding notebook sources into a live document.
Grokbot can now route its internet traffic through your own desktop computer, fixing cases where cloud-based requests got blocked.
Each Grokbot now gets its own phone-call voice, and voice mode simulates an actual call you can speed up or slow down.
Meta's Muse is getting its own email address, live video calls with an animated avatar, and computer control on Mac.
Amazon blocked Meta's Muse from shopping on its site, even though the integration had worked the week before.
Muse Charm is a Tamagotchi-style wearable with a fingerprint sensor and a camera, designed to carry Muse on a keychain.
ChatGPT's desktop app now supports Chrome extensions, including password managers, and can track a user's credit score.
Takeaway
Opus 5.5 wins, and agents keep expanding
THIS WEEK IN AI
A week of head-to-head testing found Opus 5.5 beating a cheaper rival on writing, design, and editing, while Gemini Notebook, Grok, and Meta's Muse pushed AI agents further into voice, hardware, and shopping.
02GPT-6 vs Opus
When two frontier labs ship flagship models in the same week, price alone doesn't tell you which one to switch to; run your own real task before deciding.
GPT-6 Sol and Opus 5.5 sit at different tiers inside their own companies' lineups, so a head-to-head only makes sense once you know each one's actual position.
03Writing test
A cheaper, faster model isn't automatically the better one: GPT-6 Sol finished its script test first, but its writing read less naturally than Opus 5.5's.
Testing the same prompt on two models at once, with instructions to rely only on live research, is a fast way to see which one reasons well versus which one just responds quickly.
Both models completed a full research-and-write task using under 1% of a paid weekly usage allowance, so cost wasn't the deciding factor for either one.
04Design test
A blind side-by-side comparison, playing outputs with no labels, is a more honest way to judge AI-generated design work than reading vendor claims.
The cheaper model didn't win the design test either: Opus 5.5 placed second behind a dedicated design model, while GPT-6 Sol placed last.
06Editing test
AI video editors can misjudge exactly where a cursor click happened on screen, so any auto-generated zoom or highlight needs a manual review pass before publishing.
The same raw footage produced dramatically different results between two models: one added an unrequested full-screen layout change, the other reproduced the creator's normal framing exactly.
A model that's slower on other tasks isn't guaranteed to stay slower: Opus 5.5 finished the video edit in under half the time GPT-6 Sol needed.
07Results
After a full week of hands-on testing across three task types, one model can pull ahead consistently even when it loses on price and occasionally on speed.
It's worth revisiting a model you dismissed after a bad first release; Opus 5.5 fixed problems that made its predecessor unusable for the same tasks.
08Gemini Notebook
Live voice chat inside a research tool can be faster than typing, and being interruptible mid-answer makes it feel like a real conversation.
Referencing a research notebook directly inside a document editor removes a whole copy-paste step from writing workflows.
Interactive reports that generate an infographic, slide deck, mind map, and quiz inline are a meaningfully different product than a static text report, even from the same source material.
09Google updates
Hardware built around a dedicated assistant trigger signals that voice- and gesture-driven AI interaction is being designed for, not bolted on after the fact.
10Grok updates
Routing an AI agent's web traffic through your own computer instead of a cloud instance can fix cases where sites block requests that look like they're coming from a data center.
Giving each AI agent its own distinct voice for phone-style calls makes it easier to tell multiple automated assistants apart at a glance, or a listen.
A voice memo delivered proactively, without being asked for, is a different and often more useful interaction pattern than a chat you have to open and read.
11Meta Muse
An AI agent that can act as its own email address gives it a channel to receive tasks that doesn't require opening any app.
A platform blocking an AI agent from completing purchases on its site is a real business risk to that platform, not just friction for the shopper, once agent-driven shopping becomes common.
Packaging an AI assistant into a small, single-purpose wearable is a distinct product bet from putting the same assistant into glasses or a phone app.
12ChatGPT updates
Supporting arbitrary browser extensions inside a dedicated AI app, including password managers, blurs the line between an AI assistant and a full browser.
Letting a voice assistant read and act on email, calendar, and chat tools only becomes useful once it can use those tools without a separate confirmation step for each one.
Glossary
Terms worth knowing.
GPT-6 Sol
OpenAI's mid-tier model released the same week as Claude Opus 5.5, positioned below its flagship GPT-6 Astra and above its cheaper, faster Luna model.
Claude Opus 5.5
Anthropic's flagship model update, released the same day as GPT-6 Sol and used throughout this video as the higher-priced, higher-performing model in every comparison.
Fable 5.1
The AI model the creator had been using for motion graphics work before this test, and the model that won a blind three-way design comparison against Opus 5.5 and GPT-6 Sol.
Grokbot
A chat agent product built on Grok models, separate from the consumer Grok.com interface, with its own settings for network routing, voice memos, and voice calls.
Muse Charm
A small, Tamagotchi-style wearable Meta announced for interacting with its Muse AI agent, featuring a fingerprint sensor and a camera.
Interactive report
A new Gemini Notebook output format that lets you generate and insert assets, such as infographics, slide decks, mind maps, and quizzes, directly inside a single report.
“GPT-6 Sol may be cheaper and faster, but Opus 5.5 is better. In fact, I'll even say that GPT-6 Sol is a disappointment.”
the video's blunt central verdict after a week of testing→ TikTok hook↗ Tweet quote
11:04
“It blew me away, it exceeded my expectations... in fact I'll even say that it is the best model that is currently available.”
superlative claim, clip-ready with no setup needed→ IG reel cold open↗ Tweet quote
08:31
“I'm very shocked at how bad these results are.”
reaction line with visual proof of the failed edit on screen→ newsletter pull-quote↗ Tweet quote
22:46
“Even though this was working last week, as of right now it is not working, which I think is a colossal mistake for Amazon.”
sharp, named-brand hot take→ TikTok hook↗ 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
analogy
This was another big week in the world of AI. We have a lot to talk about. Anthropic and OpenAI both released new models on the same day, and one of them is my new favorite model.
We also got some nice new features for Gemini Notebook and a new Grok model, plus voice calls with your Grokbots, and Meta held its Connect event where they announced a lot of new stuff for Metamuse. We'll talk about all of that and a whole lot more in this week's AI News Recap. My name is Paul J.
Lipski, and every week I cut through the noise to bring you the AI news that actually matters. So make sure to subscribe to stay on top of what's new. And now let's get into it.
We'll start by talking about Anthropic and OpenAI. So both Anthropic and OpenAI released new models this week. We have Opus 5 .5 and GPT -6 Sol.
I covered both of them when they first dropped earlier this week, but those videos are really just an introduction. Now that I've had time to actually use them, I want to share my thoughts on each of these and how they compare to each other. Because the big question I had was...
which one of these, if either one, is going to be my new go -to everyday model. So let's run a few tests and compare these models side by side. I'm going to be running all of these tests at a high effort level, and I'll have GPT -6 Sol on the left here and Opus 5 .5 on the right here.
For this one, I'm asking them to script out a YouTube video for me called GPT -6 Sol versus Opus 5 .5. I'm asking them to send the script here in the chat and to use their built -in browser for research so we can see how well it handles that. And specifically telling them not to access any local files and ignore any prior notes I've taken on this subject matter so they don't try to cheat.
So we'll send this one first and then quickly send this one so they're pretty much at the same time. And then we'll look at the results in just a minute. So both of them started working very quickly, opening up their browsers and getting to work.
And they're both moving through very quickly. I'm very surprised at how quickly Opus 5 .5 is navigating the web. This is much faster than we've seen in the past for an Opus model.
But of course, GPT -6 Sol is also doing it quite quickly, which is not a big surprise. But I'm very surprised how quickly Opus is moving. And just about a minute and a half in, and GPT -6 Sol says it's done, or at least done with the research, and now it is writing out the script here.
So before we read that, let me wait for Opus 5 .5 to finish because I'm curious how long that one's going to take. And there we go, slightly over two minutes in, and it is finishing up here. So GPT -6 Sol was definitely faster, but not by a lot.
All right, so let's read these side by side. We'll start with GPT -6 Sol because it finished first. So it says, OpenAI's GPT -6 Sol costs half as much per API token as Cloud Opus 5 .5.
So is Sol the obvious choice? Then you look at independent intelligence benchmark and Opus wins by 10 points. These models launched in the same week, and the answer depends on what you're asking them to do.
If you're choosing one for coding, research, or everyday work, here's where I'd start. Yeah, that's okay. Don't really love it, but it's okay.
So let's read what Opus 5 .5 put together. On Tuesday, Anthropic released Cloud Opus 5 .5. About 90 minutes later, OpenAI released GPT -6 Solve for half the price.
That wasn't an accident. These two models are going after the same people. Anyone who uses AI for coding, agents, or real work every day.
So the question is simple. Which one should you actually use? I went through the launch posts, independent benchmarks, and the first real -world tests people have run.
so just based on the intro i think opus 5 .5 did a much better job it's a little bit clearer the story that they're weaving where the points that opus 5 sorry that gpt6 soul made over here on the left it felt a little bit like it was jumping back and forth So I just finished reading both of them, and overall I think that Opus 5 .5 did a much better job, and that's what I've seen in my other testing as well.
I've been using Opus 5 .5 all week now for helping me with my writing, and I think it's actually the best model I've ever used for writing. It feels a lot more human -like the language, and it's able to replicate my tone of voice a lot better than any other model, even, of course, GPT -6 Sol. So this win I would give to Opus 5 .5.
It's definitely not perfect. It definitely sounds like AI still, but it is one of the best I've seen for writing. Now as for cost with both ChatGPT and Claude, I am on the $100 a month plan.
And the task I just sent only cost me less than 1 % of my weekly ChatGPT usage. And for Claude, it used only 1 % of my current session usage and less than 1 % of my weekly usage. So for both of them, it was a very cheap task.
Next, I wanted to see how these models could handle design work, specifically creating motion graphics, which I include in a lot of my videos. So I gave both of them the same task, asking them to create a motion graphic for a video I'm working on soon, all about the new Gemini notebook. And I actually also gave the task to Fable 5 .1 because that was the model I have been using for all my motion graphics because...
it was the best up until now. But let's see what the results are. I have all three videos right here playing for you, and you can take a look at them, kind of giving you a blind test here, so you can figure out for yourself which one you like the best.
And personally, I think number one did the worst. That one is GPT -6 Sol. But the second one is the best, that is Fable 5 .1, and the third one is Opus 5 .5, which I also think did a very good job.
Not quite as good as Fable, but pretty close. And considering it's cheaper, it probably is going to be the one that I use for most of my design work. So overall, I was very disappointed in what GPT -6 Sol created, and very happy with what Opus and Fable created.
And before we move on, let's take a quick moment to hear from the sponsor of today's video, OmniSend. OmniSend is a powerful, intuitive email and SMS marketing platform for online stores. It now connects to ChatGPT and Claude, through MCP.
That's what lets your AI work with the data in your OmniSend account. You can connect your account and choose what data it can access. So let's ask it which five email campaigns generated the most revenue in the last 30 days.
Include the click rate for each. That gives us a starting point for planning the next campaign. right here in the same conversation.
And with write access enabled, you can also have it create a campaign draft in OmniSend for you to review. If you're switching platforms, OmniSend's support team also handles free migration within five days on a paid plan. To try OmniSend, click the link below or scan the QR code.
With code PAULJL, new customers will receive 30 % off their first three months. And now let's get back to the news. Another thing I've been using GPT -5 .6 Sol and GPT -6 Astra for is editing my YouTube videos.
The software that I use plugs into ChatGPT or Claude, and through that I can have the AI remove all the silences, remove all the bad takes, and just make the video look nicer. Usually like a first pass, and then I have to jump in to do like the last 20 % of polishing for the video. So I wanted to see how each of these new models would handle.
that task. So first, this is the raw footage. I'm in the bottom right -hand corner the entire time.
There are no zoom -ins, there's no highlights, there's no scene changes like format changes, and all of the blank spaces are left in and all the bad takes are left in. So those are all things that I asked the models to change. Let's look at what GBT 6 Astra did.
So notice the first thing it did was it made me full screen here, and this looks very strange to me. I don't know why it chose this layout. This is not the layout I normally use for my videos.
It's just way too much background space. It's a very strange choice. But let's give it a quick watch.
Today I'm gonna show you how to use Gemini Notebook. The first thing you need to do is click right here where it says immediately that's wrong. This is not, I wanted it to zoom in where I click and this is not where I clicked.
It zoomed in on totally the wrong place. Then you can just close this out. That zoom was also awkward because it's zoomed in, out, and then in right away.
I also think these zooms are way too close. So yeah, I'll just be honest. I'm very shocked at how bad these results are.
Even right here where I'm asking, where I'm talking about these three different sections in Gemini Notebook, it's doing the highlight very poorly. You can see that right here. It's not properly outlining it.
And that's the same over here. And just overall, I'm honestly shocked at how bad it is because I'd never get... results that are this bad, even with GPT 5 .6 Sol.
This, on the other hand, is what Opus 5 .5 did. So immediately, this is the outline I expect and use in my video. So it did that accurately.
Today, I'm going to show you how to use Gemini Notebook. The first thing you need to do is click right here. That is where I clicked.
So it accurately moved into the right position for that. Then you can just close this. And instead of doing like another quick zoom, it highlighted that section where you close that out.
I think that was a much better choice. And then that zoom just felt a lot more natural to me. And throughout this entire video, it just does a much better job.
So when it highlights these three sections, you can see it highlighted exactly. It wasn't off like GPT -6 Sol was. And it's just, I'm honestly blown away at how good Opus 5 .5 has done here.
Both of them removed all the silences and the false starts accurately, but Opus 5 .5 produced results for me that are the best results I've ever gotten out of an AI editor before. Now, as for the pricing, both of them use about 1 % of their current usage, which was quite good, but the speed was very surprising. Opus 5 .5 took seven minutes and 15 seconds, but GPT -6 Sol took...
15 minutes and 30 seconds, which shocked me because in most of the testing I've done this past week, GPT -6 Sol has been much faster, at least faster than Opus 5 .5, but here it was so much slower and the results were honestly terrible. So hands down, Opus 5 .5 wins this one. Now, those aren't the only tests I ran, but I don't want this entire news segment to just be a comparison between these two models.
But for me, the conclusion after using both of these models all week is very clear. GPT -6 Sol may be cheaper and faster, but Opus 5 .5 is better. In fact, I'll even say that GPT -6 Sol is a disappointment.
Somehow it feels worse than 5 .6 Sol when it comes to some tasks. Opus 5 .5, on the other hand, blew me away it exceeded my expectations to be fair opus 5 was a pretty terrible model and after testing it when it first came out i never even used it but opus 5 .5 fixes all of those issues with opus 5 and goes beyond that in fact i'll even say that it is the best model that is currently available even better than fable 5 .1 maybe not for every task but when you balance that with speed and price against fable it wins and it certainly wins against gpt 6 soul in my opinion all right i want to talk about gemini notebook next they've rolled out a few features this week that i want to highlight first if you are an ultra subscriber you will now be able to live chat with your notebooks so coming into my ultra account we can come down to this notebook right here it's called family financial records and this is a fake family i put together i put in here a bunch of financial records you can see those over here in the sources like expenses
For a few years, we have budget for one of the years, and then some sources that talk about how to budget. So if we come over to the chat, you'll see the live chat button right here. Let's give this a shot.
Hey, Gemini Notebook, can you give me a breakdown on my family's expenses? Based on your financial logs from 2023 to 2025, your family's total expenses came to roughly $315 ,000. Is that good?
Your sources don't contain specific benchmarks to evaluate if that overall spending or your savings rate is considered good. All right, so that performs surprisingly well. I often find that inside the chat here, it can be quite slow.
So I was happy to see that the live chat was much faster than I anticipated and you were able to interrupt it as well. So yeah, really nice feature. It's only available for ultra subscribers right now, but it should roll out for other paid plans soon, hopefully.
Next, you can now use your notebooks as a source in Google Docs. So if you open up a Google Doc, if you didn't know this, Down here is a Gemini bar where you can ask it to write for you or refine any text you've already written.
But you can also reference other things you have in Drive or now Notebooks. So if you type in the at symbol, you can now start typing in the name of your notebook, what we call the one family financial records, right? What we were just looking at.
So I can select that and we'll now use that as a source to help write this document. They've also added a new type of report Inside of Gemini notebooks over here on the right inside a studio.
These are all the assets you can generate Of course, we've had reports for a while They've been kind of my least favorite feature of Gemini notebook. Nothing wrong with them Just a little bit boring, but now they have something called an interactive report so i just generated one this is it right here and if we open this up you'll see at first glance this doesn't look too extraordinary because we just have a bunch of text here but as you're scrolling down notice that you can add these recommended assets into the report so here it's recommending an infographic i'll click add there and i can select what i want it to look like and click on generate and that is generating now inside my report.
I'll scroll down a little bit more and here it's recommending a slide deck. So I'll select that one and generate that one. And what else do we have?
Flashcards, mind maps. That one usually goes pretty quick. So I'll generate that one.
And then a quiz. So I can generate that. And so let's see if any of those are available yet.
Here we go. Here's our mind map. So directly inside the report, we have our mind map now.
And down here we have our quiz. So the same assets you can generate, but they're just coming into this report, which is pretty cool.
And let's just close out the last of the Google AI news from this past week. First, you can now pre -order the new Google Book. This is a laptop that effortlessly syncs with your Android phone.
But the reason I'm bringing it up is because it is designed for Gemini Intelligence with Gemini Intelligence everywhere. It has some pretty cool features, like you can use your pointer, point at anything. and that will activate Gemini.
You can ask it questions or have it do things for you. At any point, you can use dictation to tell Gemini what you want to write anywhere, and it can do stuff like that. So this is stuff you can obviously do with existing laptops with the Gemini app or some other apps installed, but this will all be pre -installed and just ready to go.
Looks like it would be a really good combination if you have Android. if you're a fan of using Gemini it now works with 13 new apps you can see all of those right here they're adding more and more every day which I love to see And finally, Google released Gemini 3 .8 text to speech.
This is a very expressive audio generation model that takes text and turns it into speech. Now, this is mainly going to be used by developers, but I always like to keep my eye on these models because this is going to trickle down into the apps that we use, that developers create for us. So it's always interesting to kind of see.
how those apps get better. And this is one of the ways. So if we give a quick listen to one of these models, you can hear what it sounds like.
All right, Melbourne, so good to have you with us tonight. Right here, heaps of great music on the way for you. But first off, something that's gonna absolutely pump up your weekend.
I mean, seriously. That's pretty impressive. I mean, even when you're listening to it, you can hear plosives.
That's the noise when someone speaks too close to a microphone. Listen to that right here. We pump up your week.
That's pretty cool. It makes it sound so much more realistic. Next, I wanna talk about Grok because what do you know?
They released a new model this week too, Grok 4 .7. SpaceX AI says it's a notable improvement over 4 .6 at the same price and speed, but early impressions are pretty split on this one. Some people love it, some people hate it.
Now, I only really use Grok in two places. one when i'm doing some research on trending topics that are coming from twitter i'll usually use grok for that and right now inside of grok .com we only have access to grok 4 .6 so i haven't even been using it inside of regular grok i also use grokbot quite a bit and spacexai has not revealed what model is used inside of grokbot all that's to say i don't really have an opinion on grok 4 .7 and if you only use these consumer tools like I do, I don't think it's gonna make a difference for you.
But let's talk about GrokBot because they've also announced some new updates to GrokBot this past week. So first, if you come into your settings and click on the left where it says computer, you will now see the option to route traffic through this computer. and with this toggled on network traffic from grokbot's computer will run through your local desktop computer instead of a cloud and this will make it seem like anytime grokbot is using the internet it is coming from your location your computer and this will make grokbot work better because there were many times where Grokbot would get blocked, probably because the traffic was coming from a virtual cloud computer.
And hopefully this is going to fix that now. The other two big noteworthy updates is that we now have voice memos and voice calls inside of Grokbot. And I love these features.
So right here is one of my bots. This one does research for me every day about trending AI news. And in the past, it would always spin up this site for me where I could look at the trending news.
But now I have it also sending me a voice memo every day. Your afternoon idea deck is up from the For You feed. cloud devs said cloud code cloud sessions are generally available so it's a really helpful feature i am using it for a few of my schedule tasks and you can also set up one bot where you tell it, hey, always respond with a voice memo.
And then you can just use the dictation button, send it messages, and then it'll always send you a voice memo back, which is kind of nice. Or what you could do is use voice mode. So now, well, they don't really call it voice mode.
They call it voice chat. And it simulates actually a phone call with your bot. And what I think is cool about this is that every bot will get its own voice.
So let's call Seeker right now and see what she's up to. Hey, Seeker, can you hear me? Hey, I hear you.
Can you just tell me what's been new in the world of AI? Holding together the latest across the usual spots. All right, I think you get the idea, but here's another one.
Let's talk to commentator. This is one that scans my comments twice a day to make sure there's no spam in there, just to kind of let me know what the overall sentiment is in the comment section of my videos. So let's give a quick call to commenter.
Commentator. Hey, commentator, what's up? Hey Paul, just wrapped the afternoon pass.
Grokvoxis is quiet again, but those two book promo bots are still crawling the GPT -6 soul thread. All right, so notice how it was a man's voice that time. And that's because if you look, commentator is a man.
So it's really cool how each of them can have their own voice. And if you click through right here, you can change the voice. There are many of them that you can choose between.
So you can get it sounding exactly the way that you want, or if it was moving too slow, like I thought that... Seeker was a bit slow, so I would probably change your speed up to one and a half times. I love both these features.
I'm really glad that they added them and it makes interacting with my bots a lot more fun and easier to do. Meta held its Connect event this week, and they spent a lot of time talking about Muse. In fact, it was the first thing they talked about, and they talked about it throughout the entire keynote.
If you're not familiar with Muse, it's Meta's AI agent, and it's one of the easiest agents to use right now. When it first came out, I said that this could be the AI agent that finally gets regular people to start caring about agents. And for once, I might actually be right about something because it's been the number one app in the App Store for a couple of weeks now.
But they announced a bunch of new updates to Muse during that keynote. So first, every Muse is going to get its own email address that you can use to send it emails to do things. You can CC it, and it's just another way that you can communicate with your Muse.
Live calls are also coming to Muse and they're more than just voice calls, they're video calls. The avatar will come to life and talk to you so you can interact with it and it'll move around just like you're seeing inside this demo video right here. also computer uses here for the mac muse app and with your permission muse will be able to control any app on your mac so you'll be able to line up a bunch of tasks walk away from your computer and it'll keep working for you or you can control it remotely similar to how you can already do that with the chat gbt or claw desktop app meta also made it clear that Muse is basically always going to be free there are paid plans available if you want more usage but on the free plan the usage limits are quite high and that is their main business model to offer it for free and they want to make their money on transactions when your Muse buys something for you meta is going to take a small cut and since that's how they plan to make money it makes sense that they're adding a lot more shopping connectors including
walmart best buy sephora wayfair and they're adding shop pay and paypal for checkout there's instacart coming as well and expedia one store that's not on that list though is amazon because amazon actually blocked metamuse from accessing its site and using it so even though this was working last week as of right now it is not working which i think is a colossal mistake for amazon Amazon, all these companies need to enable a way for agents to make purchases on behalf of their humans because this is a way people are going to shop in the future.
It is a way people are already shopping right now. Now, very soon, Metamuse is coming to physical hardware. Over the next few months, Muse is coming to Meta's AI classes.
You just say your Muse's name, talk to it, and it can see what you're looking at and respond and take action without you having to take out your phone. Also, if you've been wanting a pair of Meta Glasses but you're afraid of looking like a creep, well, they have a new pair called Ray -Ban Meta Audio that has no camera in it.
You'll still be able to talk to Muse through them and listen to music and take phone calls, but just no camera. the last bit of hardware they announced is something called muse charm it's this cute little ai tamagotchi style and size device that you can use to interact with your muse there's a fingerprint sensor right here so you can only unlock it and use it with your finger on that or after pressing it with your finger and then you can talk to it and send it tasks the same way you do from your phone but with just this little device that you can keep in your pocket or on your keychain i think there's also a camera right here that you can use to ask questions about something that you're looking at so i don't know i thought that was pretty cute depending on the price i may be getting one of these for myself And before we end this week's AI news recap, let me also show you a couple of other updates from OpenAI to ChatGPT.
So first, if you use the ChatGPT desktop app, you will now have Chrome extensions. Any Chrome extension that you want will be supported, including password managers. So you can just add those the same way you would any Chrome extension on Chrome, for instance.
Also, if you come into your plugins and come to some of your plugins that you use regularly, like Gmail, you'll notice you can now connect multiple accounts. So you're not stuck with just one account like we were in the past. And another pretty significant update is that ChatGPT Voice can now use plugins like your email, calendar, and Slack.
And finally, ChatGPT can now track your credit score if you... want it to and feel comfortable with an AI keeping track of your credit score. So that's everything I got for you this week.
For me, the biggest story was Opus 5 .5. I think it's going to be my go -to model from now on, especially for writing, editing, and design work. And if the price is right, I'm probably going to get one of those Muse charms, to be honest.
But what about you? What do you think about these new models? What are you most excited for from this past week?
Let me know in the comments section down below. And if you enjoyed this video, consider subscribing because I make an AI news recap just like this one every single week for you. So if you don't want to miss out on that, make sure to subscribe, turn on the bell notification, and I'll see you next week.
Bye for now.
The Hook
The bait, then the rug-pull.
Anthropic and OpenAI shipped new flagship models on the same day, so this week's AI news recap turns into a real head-to-head: script writing, motion graphics, and AI video editing, before the roundup moves on to Gemini Notebook, Grok, and Meta's new Muse Charm wearable.
CTA Breakdown
How they asked for the click.
VERBAL ASK
26:02subscribe
“make sure to subscribe, turn on the bell notification, and I'll see you next week”
straightforward end-card subscribe ask tied to the weekly-recap format, reinforcing appointment viewing rather than a hard sell
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.
A YouTuber walks through the eleven-bot staff living inside Grok Bot, each with a name, a job, and a cloud computer that keeps working after he closes his laptop.
OpenAI, Anthropic, and Google all released flagship models in the same week, Gemini Notebook quietly capped how much you can generate, and Grok's new payments plugin lets a bot spend your money with your sign-off.
A screen-recorded build of a mountain-cabin marketing site, from a bare ChatGPT prompt to a scroll-scrubbing hero video, matching icons, and a photo gallery, all generated inside one chat.
One creator ran the exact same week of real work through both frontier models, a website, a game, a script, an inbox, a full business audit, and tracked where they actually diverged.
A YouTube tutorial on wiring ChatGPT and Claude into a $79 desktop editor called Borumi so they cut, zoom, and animate your talking-head videos for you.
00:38
00:39
01:28
02:06
02:27
02:47
03:07
03:26
03:46
04:06
04:25
04:45
05:05
05:44
06:16
06:47
06:54
07:22
07:36
07:55
08:22
08:41
09:15
09:49
10:00
10:20
10:49
11:19
11:34
12:18
12:37
13:02
13:17
13:35
13:56
14:16
14:36
15:15
15:42
15:51
16:14
16:34
17:13
17:33
17:52
18:12
18:32
18:52
19:11
19:31
19:51
20:10
20:30
21:04
21:30
21:42
22:17
22:28
22:43
22:59
23:30
23:46
24:26
24:42
25:20
25:45
15:58
24:02
14:21
25:15
13:02
13:35