NEW Grok 3.5 Update! 🤯

So today we're going to be talking about the new Grock 3. 5 updates that have just been announced. Now these are both brand new updates that are coming. And essentially I'll be just running you through today exactly what they mean, what's going to happen, what we can expect, etc. what's

happening. So basically Elon Musk has announced just a few hours ago if you weren't one of the 25 million people that saw it yesterday. Elon said next week Grock 3. 5 early beta release to super Grock subscribers only. It is the first AI that can for example accurately answer technical questions about rocket engines or electrochemistry. Grock is reasoning from first principles and coming up with answers that simply don't exist on the internet. It's quite interesting as well because I don't know if anyone's studied much about Elon Musk or how he thinks and how he makes business decisions etc. But he's very much a first principles man right which is basically this anything that's going to stop me it's purely the laws of physics that's it and he's really powerful in terms of his thinking how he handles decisions etc. And so it looks like he's training Grock

in the same way. For example, quite often if you look at like Google IO views when you search something on Google, one of the biggest issues is that quite often it's pulling in like old sort of sarcastic responses from Reddit and other sources like that which gives you a really bad answer, right? Whereas for example with Grock 3. 5, if that's able to think for itself, then we should see much bigger improvement in terms of the actual responses we get. However, what I will say is when you're using these models like Grock Safar from what I've seen, it was amazing when it first came out. Great for deep research, but very poor now in terms of UC. I don't really use it at all compared to when it first came out. Seems like almost got dialed down or it got a bit less intelligent over time. So, it'll be interesting to see, okay, what improves and how we can go on with that. By the way, Krypton is asking how, hey, I want

to make a website for free. How can I do that? Any advice? So if you want to learn how to do that sort of stuff, check out the SEO lead circle. Otherwise, just check out the AICO success lab. That is completely free on school and you can just check that out and then learn exactly how to build out a website step by step completely for free. So we're going to get straight into this now and we'll come on to the Gro 3. 5 updates first of all. So what do we know and what to expect? Okay, beyond hype of analyzing Elon Musk's announcement cuz obviously he's announced it, but there's not many details in terms of what we can expect. So, I want to run you through what I would expect. Bear in mind, this is coming in beta next week. That's 100% been confirmed. So, the announcement is next week Grock 3. 5 early beta relates to Super Grock subscribers only. It's the first AI can, for example, accurately answer 10 questions about rocket engines and electro chemistry without simply copying from the internet. So, what we know so far is there's a beta release scheduled for next week. is exclusive to super grock subscribers only. It claims to generate technical answers without copying from internet sources and it focuses on specialized domains like rocket engineering and chemistry. Right? So it's kind of like I'll come on to some use cases how I think like rock will be used longterm but it's quite interesting to see this. Now if you look at the actual timeline it's accelerating right here. For example, the AI development from one first came out November 2023. September 2024, we saw Grock 2. February 2025, we saw Grock 3. And then May 2025, Grock 3. 5 will be released next week. This is going to be very interesting. Grock 3. 5 is still unreleased. Just making that very clear. As with previous models, actual capabilities may differ from pre-release claims. And obviously this presentation just analyzes what's known so far. But independent testing is going to be needed to verify the performance. So it's going to be quite interesting to see how this goes. Let's get straight into this. So

in terms of key features, now capabilities. Number one, first principles reasoning. So it can generate technical answers without copying from the internet. So what does that mean? For example, if you ask chat GPT or any of these other models, basically what they're doing is just mish mashing different answers from different places, right? So they've all been trained on the data from the internet and then that internet trained data is basically giving you chat GPT's interpretation of that data right it's not thinking for itself essentially also what we can expect is enhanced reasoning right so building on Grock's think and big brain modes so according to evolution AI hub let me pull this up here so according to evolution a hub gro 3. 5 may offer more transparent step-by-step explanations Also, what's expected is a broader context. So, improved context window for handling longer inputs. For example, if you've seen Gemini 2. 5 or these other new models that coming out, like the context windows are just getting bigger and bigger. So, it's going to be quite interesting to see what we get next right there. Also, potentially more safety filters as these AIs get smarter, as they can think for themselves, etc. You can expect more independent thinking from that. Now, I don't think this is necessarily like the specialized domain expertise, but if you look at the way that Elon Musk is building out all of his companies essentially, he's got like SpaceX that he's trying to get to Mars right now. I can imagine that Grock will be the AI that goes into space and sort of helps with that. It's going to be crazy stuff for sure. It's going to get wild. So, these are the sort of prompts someone might give 3. 5 about like rocket engines and stuff like that. That's a sort of expected response. Rather than copying text from space agency websites, Grock 3. 5 is expected to generate a unique technical explanation based on fundamental principles of electrochemistry and aerospace engineering, right? So, it's just going to come up with more unique answers, which I think is great as well because if you look at the way the internet is going just in general, when people are writing content that appears on Google or anywhere that you read a blog article, typically it's not very unique news, right? is quite often based on old stuff or a mish mash of other stuff that's out there already. So, it's going to be interesting to see how that performs and whether that actually helps us create more unique content in the future. Also, for technical problem solving, this should be a lot more powerful. For example, if you look at the way that Elon has framed this, it's like this will be like a AI that you could use for example for engineering problems and technical problems and that sort of thing. This quite different for example from existing leading models, right? So if you look at like the big models right now, if you asked a question like design a cooling system for rocket engine nozzle that minimizes weight whilst maintaining thermal performance. Previously with current AIS, you might see that AI explain existing cooling techniques, reference published designs and principles and compare like the known approaches. It's not really going to think for itself, right? Whereas for example with Grock 3. 5, it's expected to generate novel designs not copied from existing sources. might provide like detailed engineering calculations, reasoning through trade-offs from first principles and showing its work in solving multi-step technical problems. Now, one thing I'm going to say right here, right, and this is just a reality check cuz I think in AI, we often need to, no puns intended, but we need to be become a bit more grounded, right? It's nice talking about designing rocket engines and all that sort of stuff, but let's be honest, AI is still hallucinating quite often. It still provides a lot of wrong answers. It still does silly stuff that we're like, why are you doing that? And so, what I would say is these examples illustrate Musk's claims about Grock 3. 5. But so many people on Reddit, so many AI experts, etc. are expressing skepticism. and current models like GPT4 claw 3. 7 chat GP03 even they provide detailed technical answers in these domains already right so the key question is whether Grock 3. 5 will genuinely produce novel insights through reasoning rather than recombining training data which can only be verified through like independent testing after release right so you're only going to find out once it gets released but this is quite interesting as well here's another one so for example for electrochemistry if we put in a prompt like for example design a novel lithium ion battery electrode material that reduces degradation during fast charging. Right? I know you're all typing that into Grock obviously like we're all electrochemistry engineers here. Previously current models might like summarize publicized research and known approaches etc. Whereas this with Grock would suggest like something unique through first principles reasoning and that's going to be very different if it actually comes off in that way. So this is going to be super interesting. for example, like you might see more unique answers that have never been shown to the world before. So, let's talk about

like internet reactions, right? How people responded to this. For example, some people say the unbiased opinion Grock is actually a really good model. Can't wait to see how this compares to Quinn. This could be quite good considering that Grock itself is more than decent right now. Other people saying it's amazing how fast this company is moving and it's been really impressive so far. and a lot of people saying they actually really like Grock, especially for coding and writing difficult stuff. However, there's some negative responses as well, right? And I want to show you both sides of this thing. I don't want to tell you like this is going to be the greatest thing if we don't know. A lot of people saying, for example, that Elon's living under a rock or that actually these models can actually answer these questions already. Also, the image generation capabilities are way behind OpenAI. Gro and flux in its model right there was really powerful previously, but if you look right now and you look at its abilities right now, it's nowhere near the same level as something like for example chat GPT4. Like a lot of my thumbnails now, I've compared literally my designer versus chat GP for generating images and chat GP wins like every time. It's just absolutely amazing for generating responses and generating images. So that's going to be very interesting too. Let's keep going through now. So if we have a look at a current performance of GRO 3 and we compare this on benchmarks versus some of the leading models, right? It's really it is decent coding. It's got a decent conversation memory. One of the main things here is like it's less censored and I think actually chat GBT have improved their models and made those less censored simply because of Gro's ability to give like more direct answers without censoring stuff. I don't know if you remember about 6 months ago, but when I was using Claude, it just wouldn't answer anything that I gave it and everything was a little bit like overly PC, overly politically correct. So, it's going to be interesting to see what we get back from Gro. Some people asking me, "What are your expectations of Deep Seek R2? " You should check out my video from earlier this week about that cuz I actually created a video about Deep Seek R2. if you type in that on YouTube and you'll see what I have to say about that cuz Deep Seek R2 right now there's not a lot of information that's actually coming directly from the source and any claims that have come out from R2 have been like from stock market manipulation images or like random Chinese investing and trading forums right so I don't think that you can have any real expectations from D2 if nothing has been confirmed or nothing has even been seeded from the official source. And the weakness is of course right now is like inferior image generation, limited context window, and inconsistent performance. This is definitely something that I found is like when it first came out, Brock was absolutely amazing. And it just seems like it's dialed down or it's evolved in the wrong way. And now the performance is nowhere near as good when I do the tests. And you probably see a lot of videos on YouTube with my tests comparing each model side by side, but typically 3. 7 Sonet, Gemini 2. 5 Pro, these are the leading models right now. So, what can we realistically expect? I don't want to hype this up and tell you anything crazy right now. I would say it's going to be modest improvements. Bear in mind, this is 3. 5. It's not four. So, it's just going to be like a little moderate improvement. It might be better at technical reasoning particularly in specialized domains but I think for the general population it's not going to make a big difference right so for example if you compare this to something like chat which has a great memory understands everything about you that feels a lot more like it understands me and gives me better reasoning but technical problems personalized to me in my life versus something like grock additionally continued code strength and also I would expect that the context window is going to improve genuinely So, we'll see. And then the hype versus reality, the first AI for technical questions. Well, there's so many models that already handle technical questions without simply copying from the internet right now. And this is a bold claim from Elon because right now all AI models actually generate responses based on training data. So, it's going to be interesting. And then for first principles reasoning, it might be improved, but I think it's still going to use similar techniques as competitors. But we'll see next week and we'll see what happens. This is an interesting perspective. So someone said if 3. 5 can genuinely reason from first principles and generate accurate answers to advanced technical questions, it's a big leap beyond current LLM. The fact that it's pitched as producing non-inderived insights suggests it's leaning heavily into symbolic reasoning or hybrid models. So it's going to be interesting to see what we get back right there. And then of

course we can expect Grock 4. I think this is going to be the big evolution, right? Bear in mind as well like I would expect that Gro's models get more and more powerful because they've got this like super computer trajectory. So for example like there's a rumored 1 million GPU cluster. So there's reports to suggest plans for unprecedented AI compute facility to power Grock 4. Additionally, you got the training data expansion. So we can expect a significant increase in training volume and enhanced architecture. Right? It's a new model architecture optimized for reasoning over long context. In Elon's announcement, he did say that Grock 4 is going to be ready for release by the end of 2025. So, I think that would be a huge step up. And obviously, AI is just changing so much that by that point, I don't even know how good AI is going to be. It's going to be insane. But, I think right now, like Grock 3. 5 beta launching next week, it's probably going to just provide insights into development trajectory towards Grock 4. And also this is just based on announcements and industry reports rather than confirmed specifications as well. So what we know beta release next week exclusively for super grock subscribers that's about $50 a month. You got focus on technical domain. So Elon specifically mentioned rocket engines and electrochemistry. Additionally you got claimed first principles reasoning. So this is not going to be use just like internet trained data but we'll see. And this is obviously part of a broader road map. So, Gro 4 is released and planned for release by end of 2025. So, that's pretty much it. Just

stay updated. Make sure you subscribe to this channel and you will find the latest updates on Gro 3 and what's going to happen next. So, thanks so much for watching. If you want to get access to all my best workflows, tips, templates, tutorials, plus an amazing community of people focused on making money and growing their business with AI, feel free to get the AI profit boardroom. This is an amazing community of 645 members focused on growing and building stuff with AI. Additionally, inside here, you'll get access to all my best courses, workflows, AI agents. And if we go inside the SAP section over here, you'll see all of these different SAPs for new tools. This is just from April. It's absolutely wild what's happened over the last month. For example, like Luchra, Quent 3, Suna, Trey AI, etc. If you just want to learn and if you want actionable use cases for using AI, then feel free to get that. Additionally, you can post inside the AI profit boredom community if you have any questions and you're struggling with anything. So, thanks so much for watching and additionally, if you want to get a free onetoone SEO strategy session, feel free to get that link in the comments description. We'll show you how we take websites from zero to 145,000 business month and generate hundreds of thousands of dollars in sales on autopilot. On this free link building acceleration session, you'll get everything you need as a customtailored game plan to grow your business with SEO. And that's basically it for the Notebook LM and G and Gro 3. 5 updates. Thanks for watching. Appreciate it. Bye-bye.