November 19, 2025 • 27 mins
🌌 LIMITLESS HQ: LISTEN & FOLLOW HERE ⬇️
https://limitless.bankless.com/
https://x.com/LimitlessFT
------
Google launched Gemini 3.0, a groundbreaking AI model with advanced multimodal capabilities for interpreting diverse inputs.
We highlight its impressive benchmark performance, including a 37.5% on Humanity’s Last Exam, and discuss Google’s proprietary TPU infrastructure and the new tool, Google Anti-Gravity.
------
TIMESTAMPS
0:00 Gemini 3.0
5:34 Game-Changing Features
9:04 Benchmarking Breakthroughs
13:21 Cost and Quality Trade-offs
17:46 Google’s Strategic Advantage
24:51 Predictions for the Future
------
RESOURCES
Josh: https://x.com/JoshjKale
Ejaaz: https://x.com/cryptopunk7213
------
Not financial or tax advice. See our investment disclosures here:
https://www.bankless.com/disclosures
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Josh:
If I had a crown in my hands, I would place it on the heads of Google because they have done it again. (00:00):
undefined
Josh: They have the world's best AI model ever in history by a shockingly large margin. (00:04):
undefined
Josh: Gemini 3.0 just got released. (00:09):
undefined
Josh: It's available now to anybody in the world to go use it. And the benchmarks (00:11):
undefined
Josh: are kind of blowing everyone's expectations out of the water, myself included. (00:15):
undefined
Josh: And most importantly, it places another data point on the chart that shows we (00:18):
undefined
Josh: are continuing to ascend up this exponential curve towards AGI. (00:22):
undefined
Josh: And the roadmap is still intact and we are very quickly moving through it. (00:26):
undefined
Josh: This, EJs, I was just, I was going through the benchmarks before recording this (00:30):
undefined
Josh: and I, it's, it's shocking because we live in this world and yet somehow I'm (00:32):
undefined
Josh: still continually blown away by the progress that's made by these models. So let's get into it. (00:37):
undefined
Josh: Please walk everyone through, tell me, what did Gemini and the Google team just (00:41):
undefined
Josh: release with this 3.0 update? (00:45):
undefined
Ejaaz: People probably think we say the world's best model every week, (00:47):
undefined
Ejaaz: but this time we really, really mean it. (00:51):
undefined
Ejaaz: Like they have blown every single other model provider out the water. (00:53):
undefined
Ejaaz: The things that this thing can do. Well, how about how about I just show you? (00:58):
undefined
Ejaaz: How about I show you, Josh? (01:01):
undefined
Josh: Please, let's see some examples. (01:03):
undefined
Ejaaz: We have a thread here. And Sundar basically says, you can give Gemini 3 anything, (01:04):
undefined
Ejaaz: images, PDFs, scribbles on a napkin, and it'll create whatever you like. (01:09):
undefined
Ejaaz: For example, an image becomes a board game. (01:13):
undefined
Ejaaz: A napkin sketch transforms into a full website. (01:15):
undefined
Ejaaz: And a diagram could turn into an interactive lesson, right? So there's two examples (01:18):
undefined
Ejaaz: I want to show you, Josh. I want to get your opinion on this. (01:22):
undefined
Ejaaz: So number one, there's a short video of someone playing pickleball here and (01:25):
undefined
Ejaaz: he or she rather uploads it into Gemini and says, hey, can you tell me how well (01:30):
undefined
Ejaaz: I've done here and how I can improve my game? (01:36):
undefined
Ejaaz: And it analyzes the entire video. (01:38):
undefined
Ejaaz: It knows that she's wearing a knee brace. It analyzes her positions, (01:40):
undefined
Ejaaz: telling her where she can move to better position herself to score the point. (01:44):
undefined
Ejaaz: That's pretty nuts. But before I get your reaction to that, because Josh, (01:48):
undefined
Ejaaz: I know you're an athlete. I know you're very competitive when it comes to these (01:50):
undefined
Ejaaz: things. So this is a tool you could definitely use. (01:53):
undefined
Ejaaz: The second thing is probably applicable to a lot of listeners on this show. (01:54):
undefined
Ejaaz: They've embedded Gemini 3 into Google search and into new generative UI experiences. (01:59):
undefined
Ejaaz: The way I would summarize this is it basically is very intuitive, Josh. (02:05):
undefined
Ejaaz: It understands what you're asking for without you needing to really kind of explain yourself. (02:10):
undefined
Ejaaz: The example they're showing on the video here is, can you explain the three body problem to me? (02:14):
undefined
Ejaaz: And rather than just give you kind of like this simplistic text, (02:20):
undefined
Ejaaz: which explains the concept, it decides to create a video diagram from scratch (02:22):
undefined
Ejaaz: to show you a visual depiction of how this works. (02:27):
undefined
Ejaaz: Right. Give me your reaction in order from one to two. So starting with the sports. (02:30):
undefined
Josh: Okay, let's go. The first example. (02:35):
undefined
Ejaaz: Yes, sir. (02:36):
undefined
Josh: So this is really cool, the napkin example, where you can scribble something (02:37):
undefined
Josh: down on a piece of paper, it'll generate it in the real world. (02:40):
undefined
Josh: What all these examples are kind of showing me is what we always talk about (02:43):
undefined
Josh: with Google, where it has this awareness of physics, reality, (02:46):
undefined
Josh: and visuals and understanding what it's seeing. (02:51):
undefined
Josh: And all three of these examples are leaning into that. So it leads me to believe (02:53):
undefined
Josh: Gemini really is a multimodal first model, where it's meant to ingest, (02:56):
undefined
Josh: meant to understand the world around us. This example of the chessboard and (03:00):
undefined
Josh: the napkin is amazing because a lot of people oftentimes have sketches. (03:04):
undefined
Josh: You just draw it down on paper and it intuitively understands it. (03:08):
undefined
Josh: But the one that was most surprising to me is the video example. (03:10):
undefined
Josh: Because as far as I'm concerned. (03:13):
undefined
Josh: As far as I'm aware, there has never been a model that can ingest video and (03:16):
undefined
Josh: understand the video that it sees. And if it does exist, I've never tried it before. (03:20):
undefined
Josh: So the idea that you can, I mean, I play baseball growing up. (03:23):
undefined
Josh: I, if I could take a video of myself swinging and get a corrective coach to (03:26):
undefined
Josh: walk me through exactly what was wrong. (03:30):
undefined
Josh: A lot of people play golf. I'm sure who are watching this. (03:32):
undefined
Josh: If you could have a phone recording of you playing golf and it can actually (03:34):
undefined
Josh: critique it and then critique me as if you are a tiger woods, (03:38):
undefined
Josh: critique me as if you are whoever else is good that plays golf. (03:40):
undefined
Josh: I don't know, Rory McIlroy, whatever they are, but like critique Critique me (03:43):
undefined
Josh: as if you are an expert who is really good at golf and can give me some feedback (03:45):
undefined
Josh: on how I can better my swing. (03:49):
undefined
Josh: And what this offers in this one just narrow case example is now you have this (03:51):
undefined
Josh: personal tutor that can do anything. If you're dancing, if you're doing anything (03:54):
undefined
Josh: physical, if you're whatever it is, it can evaluate things for you. (03:58):
undefined
Josh: Even if you have a video, a podcast, EJS, we uploaded to Gemini 3.0, (04:01):
undefined
Josh: it could critique us. What did we do well? What did we not? (04:05):
undefined
Josh: What did the visuals look like? How can we improve them? And that awareness (04:07):
undefined
Josh: of video is like really cool. (04:10):
undefined
Ejaaz: Yeah, I just want to say, I think the closest we got to this was with GPT, (04:13):
undefined
Ejaaz: where you can upload an image, like what's under my car bonnet and say, (04:18):
undefined
Ejaaz: hey, what's wrong? My car stopped working. (04:22):
undefined
Ejaaz: And it can kind of like identify the point that you need to kind of like change, (04:24):
undefined
Ejaaz: change the oil, blah, blah, blah. (04:28):
undefined
Ejaaz: But that's just a static image. To go from that to live video and for it to (04:29):
undefined
Ejaaz: analyze all the frames in that video and then give you a response on that is (04:35):
undefined
Ejaaz: a massive leap upwards. We just haven't seen that anyway. So yeah, you're right. (04:38):
undefined
Josh: It's amazing. And every example we go through, it kind of breaks the mold of (04:42):
undefined
Josh: what I believe should be possible. (04:45):
undefined
Josh: And I find that it's going to be difficult to use Gemini 3.0 because there are (04:47):
undefined
Josh: so many possibilities now that have not existed previously. You kind of need (04:50):
undefined
Josh: to relearn how to engage with AI because it's so capable. (04:54):
undefined
Josh: And there's a fourth example here that I just want to touch on briefly, (04:57):
undefined
Josh: which was also cool, is that it works just as well for the other things. (05:00):
undefined
Josh: The example is a trip planning one where it starts to plan a trip and a vacation. (05:03):
undefined
Josh: And it shows you a full list that is fully interactive of all the places broken up day by day. (05:07):
undefined
Josh: And there's an option that you could just choose visual layout. (05:12):
undefined
Josh: And you see on the screen here, it'll take every single day of your trip, (05:15):
undefined
Josh: break it into images, section it out into this really nice visual grid. (05:18):
undefined
Josh: So what I'm seeing the themes here are, okay, real world understanding, (05:22):
undefined
Josh: video first, and really nice presentation, which I think a lot of models sometimes struggle on. (05:25):
undefined
Josh: Demo's out of control. I'm excited to use it. Everyone else can use it now. (05:31):
undefined
Josh: It's live. Now I want to get to benchmarks, EJS, because this is where things (05:35):
undefined
Josh: get kind of crazy, where we could actually compare one to another and see exactly (05:39):
undefined
Josh: how impressive this is relative to everybody else. So please, we have the card here. (05:43):
undefined
Josh: Walk us through what we're seeing in this model card and all the specs that we need to know. (05:47):
undefined
Ejaaz: As you guys probably know by now, benchmarks is typically how we evaluate any (05:50):
undefined
Ejaaz: typical AI model against each other. (05:54):
undefined
Ejaaz: And they're measured against a range of different benchmarks. (05:57):
undefined
Ejaaz: A benchmark can be considered as sort of like a test. (05:59):
undefined
Ejaaz: Now, right at the top, you've got humanity's last exam. This is by default, (06:01):
undefined
Ejaaz: the hardest exam that an AI model is tested against. (06:06):
undefined
Ejaaz: And it's kind of like an academic reasoning test with no tools accessible to it. (06:10):
undefined
Ejaaz: It scored a very impressive 37.5%, which is more than I think is about a 15% (06:14):
undefined
Ejaaz: increase from its previous model. (06:21):
undefined
Ejaaz: Very, very impressive. But what really blew my mind was the second stat listed (06:23):
undefined
Ejaaz: here, which is the ARK AGI 2 benchmark. (06:28):
undefined
Ejaaz: Josh, when I say this 2x'd the previous state-of-the-art model, I absolutely mean it. (06:31):
undefined
Ejaaz: In fact, let me just show you this chart here. (06:38):
undefined
Ejaaz: Now, you may notice a couple of bustly specs here. (06:41):
undefined
Ejaaz: GPT-5 Pro, Grok 4 Thinking. And then can you see that outlier right at the top (06:45):
undefined
Ejaaz: right. Do you see that, Josh? (06:51):
undefined
Josh: That's insane. The two outliers. (06:53):
undefined
Ejaaz: The two outliers. So these are the Gemini 3 Pro and the Gemini 3 deep thinking model. (06:55):
undefined
Ejaaz: Deep thinking being like, you know, a large number that can like kind of give (07:02):
undefined
Ejaaz: you a more research response. They are a... (07:04):
undefined
Ejaaz: Stand out from every single other model. And the reason why this is so crazy, (07:07):
undefined
Ejaaz: well, there's a few reasons. (07:12):
undefined
Ejaaz: Number one, all the other model progressions, as you can see over time, (07:13):
undefined
Ejaaz: has been kind of impressive, but kind of small. (07:18):
undefined
Ejaaz: Like they've been a good jump. It's been impressive, but it hasn't been as impressive (07:23):
undefined
Ejaaz: to be like, oh, you know, another model provider couldn't catch up. (07:28):
undefined
Ejaaz: These results from Google literally put it miles ahead of every other model. (07:31):
undefined
Ejaaz: So when I sat at this chart, I think, wow, Google probably has the lead for (07:36):
undefined
Ejaaz: another six months and in six months time, they're going to have a more impressive model by then. (07:40):
undefined
Ejaaz: So at this point, I'm kind of thinking, can anyone catch up to Google? (07:46):
undefined
Ejaaz: Josh, do you have any reactions to this benchmark? (07:49):
undefined
Josh: This is the chart that like the first thing I said to myself when I saw this (07:51):
undefined
Josh: is like, oh, my God, there is no wall there. (07:55):
undefined
Josh: We are not going to stop scaling. The scaling will still apply because these (07:57):
undefined
Josh: two new data points that we have blow everything else out of the water. (08:01):
undefined
Josh: And this is how exponential growth happens. (08:03):
undefined
Josh: It seems like a really small cluster down there in the bottom, (08:05):
undefined
Josh: but the reality is that was the top just a couple hours ago. (08:08):
undefined
Josh: And Gemini kind of refactored this entire chart to make it seem like it's so (08:12):
undefined
Josh: small because the progress is so high. (08:17):
undefined
Josh: And although Gemini 3.0 thinking is seemingly the most impressive, (08:18):
undefined
Josh: the really anomaly chart is the Gemini 3.0 Pro, which is basically a vertical (08:23):
undefined
Josh: line up from these other models, where the score is higher, but the cost is (08:27):
undefined
Josh: actually slightly lower. (08:32):
undefined
Josh: And if you connect a dot between these averages, you start to see literal vertical (08:33):
undefined
Josh: line in terms of improvement and acceleration in these models. (08:37):
undefined
Josh: And that to me shows that there is no scaling wall that we're hitting. (08:40):
undefined
Josh: Like we can continue to scale resources, energy, compute, and we could continue (08:45):
undefined
Josh: along this path towards AGI in a world where some people were saying, (08:48):
undefined
Josh: we don't know if it continues. The answer to me is very clearly, it continues. (08:51):
undefined
Josh: This is a step much closer to AGI. (08:55):
undefined
Josh: And again, that real world understanding makes it feel much closer to AGI than it ever has before. (08:58):
undefined
Josh: Because now it really like intuitively understands the world through video, (09:04):
undefined
Josh: through photo, through audio, through basically every sensory input we have (09:08):
undefined
Josh: outside of what taste and feel. (09:11):
undefined
Josh: So this to me, I saw this chart. I was like, oh, my God, Gemini, (09:13):
undefined
Josh: you really outdid yourselves. (09:17):
undefined
Ejaaz: I'm just going to be honest. I think over the last couple of months, (09:18):
undefined
Ejaaz: I've been getting a little bored with the models that have been released by other model providers. (09:22):
undefined
Ejaaz: And it led me to think that we're not going to make many breakthroughs until (09:27):
undefined
Ejaaz: some model provider figures out a new, unique way to train their model. (09:32):
undefined
Ejaaz: Gemini or Google has convinced me otherwise with this release. (09:37):
undefined
Ejaaz: But I know you guys are probably like fed up with listening to us hop on about benchmarks. (09:41):
undefined
Ejaaz: So how about I materialize that for you in a much more easy to understand way, right? (09:45):
undefined
Ejaaz: So here are the four big takeaways that you need to learn about Gemini 3. (09:50):
undefined
Ejaaz: Number one, for its intelligence, for the intelligence that you're getting, (09:55):
undefined
Ejaaz: it is not that super expensive. (10:00):
undefined
Ejaaz: Google trained this from scratch, as this tweet says, using their own GPU infrastructure. (10:03):
undefined
Ejaaz: And it used this kind of like layout called a mixture of experts, (10:07):
undefined
Ejaaz: which basically means that whenever you prompt the model, it's not going to use the entire model. (10:11):
undefined
Ejaaz: So it actually ends up being cheaper than what it could eventually become. (10:16):
undefined
Ejaaz: One million token context input, 64K token output. We'll get to the costs in (10:19):
undefined
Ejaaz: a bit of a second, But the point that I'm making here is that it's not as expensive (10:24):
undefined
Ejaaz: as you would expect for the intelligence that you're getting. (10:27):
undefined
Ejaaz: Now, if you compared Gemini 3 to GPT 5.1 from OpenAI, on a relative basis, (10:29):
undefined
Ejaaz: it is more expensive. But for the jump in intelligence that you're getting, (10:36):
undefined
Ejaaz: it's way better. So it's, in my opinion, worth it. (10:39):
undefined
Ejaaz: Number two, when it comes to computer use, so that means letting the AI model (10:42):
undefined
Ejaaz: control your computer and do tasks for you whilst you go do something else. (10:47):
undefined
Ejaaz: It is state of the art. It is the best here. They measured it against a benchmark (10:52):
undefined
Ejaaz: called ScreenSpot Pro, which is a benchmark which kind of like analyzes its (10:56):
undefined
Ejaaz: ability to understand images and visuals on a desktop. It just absolutely crushes it. (11:00):
undefined
Ejaaz: Number three, it is the best AI for math by far. (11:06):
undefined
Ejaaz: So again, the point I'm making here or the theme that we're seeing here is it's (11:09):
undefined
Ejaaz: not just good at one thing, it's good at many things, which makes it the best (11:13):
undefined
Ejaaz: generalist AI model in the world right now, by far. (11:17):
undefined
Ejaaz: And the final thing, Josh, and this is where it might slip up. (11:21):
undefined
Ejaaz: I'm curious to get your take on this. (11:24):
undefined
Ejaaz: It is insanely good at coding, but we don't quite know if it is the best at coding yet. (11:26):
undefined
Ejaaz: What I mean by that is it completely crushed everyone else on one coding benchmark, (11:32):
undefined
Ejaaz: but the coding benchmark that matters, which is the software engineering, (11:37):
undefined
Ejaaz: SWE, it didn't do as well as its competitor, Claude 4.5 from Anthropic. (11:40):
undefined
Ejaaz: So those are the four main takeaways. (11:45):
undefined
Josh: I would much prefer a model that understands the world than understands how to code. (11:47):
undefined
Josh: And I think we're starting to see these subset niches where if Anthropic has (11:51):
undefined
Josh: the best coding model, that's great. (11:55):
undefined
Josh: Let them focus on code. Let them narrowly make that the best model. (11:56):
undefined
Josh: Let Google handle everything else. And I think that's what Gemini is focusing (12:00):
undefined
Josh: on. So the code thing doesn't really bother me because I don't care to use Gemini for code. (12:03):
undefined
Josh: I'm happy to be in Claude Camp for code and then use Gemini for everything else. (12:07):
undefined
Josh: And then one of the points earlier that you mentioned on the pricing is. (12:11):
undefined
Josh: I find it a little interesting because it's a little bit more than just a little bit more expensive. (12:14):
undefined
Josh: The pricing, I was looking through it and it's for over 200,000 tokens. (12:18):
undefined
Josh: They're charging $4 for inputs and $18 for outputs. (12:22):
undefined
Josh: Now, relative to GPT 5.1, which just got released, they're charging for a million (12:27):
undefined
Josh: tokens, $1.25 in, $10 out. (12:32):
undefined
Josh: So you're talking about, what is that? that's about $20 versus $1.25 on inputs. (12:35):
undefined
Josh: And that is a fairly significant margin that you're paying for this quality. (12:42):
undefined
Josh: So we're starting to see the trade-offs happening on that Pareto curve that (12:45):
undefined
Josh: we talked about in a few episodes earlier, where there are trade-offs coming (12:48):
undefined
Josh: in terms of cost and quality. (12:52):
undefined
Josh: And it's clear that while OpenAI may have optimized for cost, (12:53):
undefined
Josh: Google is kind of optimizing for a little further up the cost curve in exchange for super high quality. (12:57):
undefined
Josh: And it seems like this is kind of a balanced data point for now because unless (13:02):
undefined
Josh: you are using this via API and you're requiring a ton of tokens, (13:06):
undefined
Josh: a $20 a month Google membership will get you all of the use that you need. And that is just fine. (13:09):
undefined
Josh: So in terms of a usability perspective, I think that's okay. (13:14):
undefined
Josh: But it's just an interesting thing to know is that this is a better model. (13:18):
undefined
Josh: It is also more expensive. (13:21):
undefined
Josh: And that is a trade-off that was made. And in the case that OpenAI decides to (13:22):
undefined
Josh: make this trade-off with GPT-6 or Grok decides to make this with Grok-5, (13:25):
undefined
Josh: or Grok 6. I'm losing track of all these models now. I think we're going to (13:30):
undefined
Josh: start to see the dynamic shift in terms of that Pareto curve and what model (13:34):
undefined
Josh: architects decide to remove and add. (13:37):
undefined
Josh: And in this case, it looks like Google added quality, but they also did add (13:39):
undefined
Josh: quite a significant cost increase. (13:42):
undefined
Ejaaz: I personally don't think it matters. I think it's a nothing burger. (13:44):
undefined
Ejaaz: I think that if Google wanted to make it affordable for everyone, (13:46):
undefined
Ejaaz: including the developers that want to get API access, that think it might be (13:50):
undefined
Ejaaz: too expensive, they could subsidize it. (13:52):
undefined
Ejaaz: They are a cash flow giant. They have enough money to do that. (13:54):
undefined
Ejaaz: OpenAI has been doing that for so long now that it doesn't even matter. (13:57):
undefined
Ejaaz: I don't see any reason why Google couldn't do that. (14:00):
undefined
Ejaaz: The other reason is Google just released their latest TPU, which is their GPU (14:02):
undefined
Ejaaz: that they use to train their models and inference their models. (14:08):
undefined
Ejaaz: And typically with every generation, we get a much cheaper cost of inference. (14:11):
undefined
Ejaaz: I think by the time that they release their next generation model, (14:15):
undefined
Ejaaz: which might be, you know, Gemini 3.1, we're going to see a considerable reduction (14:18):
undefined
Ejaaz: in the cost for using Gemini 3 Pro and Gemini Pro deep research. (14:24):
undefined
Ejaaz: So I'm not too worried about that. I think it's kind of like a short-term problem (14:29):
undefined
Ejaaz: and not a long-term problem. (14:32):
undefined
Ejaaz: But kind of speaking of TPUs, I just want to take a moment to really kind of (14:34):
undefined
Ejaaz: belay the point that using their own TPUs to train a state-of-the-art model (14:39):
undefined
Ejaaz: that is 2x better than the previous state-of-the-art model and probably puts (14:44):
undefined
Ejaaz: them in a six-month lead after Google started off on the back foot, (14:48):
undefined
Ejaaz: creating probably the worst model I've ever seen and (14:53):
undefined
Ejaaz: changing that all around in what's it under two years is (14:56):
undefined
Ejaaz: nothing short of insanity tpus is (14:59):
undefined
Ejaaz: uh google's kind of version of the gpu gpu is (15:02):
undefined
Ejaaz: kind of like what nvidia controls the monopoly over this is the hardware that (15:05):
undefined
Ejaaz: you use to train your ai and inference your ai the uh unique part here is that (15:09):
undefined
Ejaaz: google's never used an nvidia gpu uh or in any considerable way to train their (15:13):
undefined
Ejaaz: models they've always trained it in-house and that's such a difficult and tricky (15:19):
undefined
Ejaaz: thing to do because designing and building these (15:23):
undefined
Ejaaz: TPUs at scale, these GPUs at scale is a super hard and complex thing. (15:25):
undefined
Ejaaz: You need so much talent, you need so much expertise and insight to be able to do that. (15:29):
undefined
Ejaaz: The unique thing about Google's TPUs, well, there's two main takeaways. (15:33):
undefined
Ejaaz: Number one, it's cheaper to train the same amount of intelligence that an NVIDIA (15:36):
undefined
Ejaaz: GPU is. So it's more cost efficient. (15:41):
undefined
Ejaaz: And the second way is, and this is their secret sauce, you can stack those TPUs (15:43):
undefined
Ejaaz: on top of each other in a really scalable way that you can start training really, really large models. (15:47):
undefined
Ejaaz: If you wanted to train the same size model with nvidia (15:53):
undefined
Ejaaz: gpus it would cost way more and it would take way (15:57):
undefined
Ejaaz: longer so google made a really risky and big bet about a decade ago saying we're (16:00):
undefined
Ejaaz: going to build our infrastructure in-house and we're not going to rely on nvidia (16:05):
undefined
Ejaaz: and we're going to benefit from the full stack experience and this model is (16:07):
undefined
Ejaaz: a prime example of that bet paying off so i just want to call them out like (16:11):
undefined
Ejaaz: it's not like google has gotten lucky here they've been planning it for a while now The (16:15):
undefined
Josh: Interesting thing to me is that this is the first number one model in the world (16:19):
undefined
Josh: built on something other than an NVIDIA GPU. (16:24):
undefined
Josh: And that's fairly significant because every company in the world is trying, (16:27):
undefined
Josh: but this is proof that it's actually possible. (16:31):
undefined
Josh: And I think when we talk about Tesla and AI5 and the XAI team, (16:32):
undefined
Josh: when we talk about OpenAI working with whoever they're working with to build (16:37):
undefined
Josh: their own in-house GPUs, I think this sets a precedent that it is possible. (16:41):
undefined
Josh: And I suspect that will result in more companies putting their foot on the gas. (16:45):
undefined
Josh: When it comes to kind of destructing part of NVIDIA's monopoly that it holds (16:49):
undefined
Josh: over GPUs. So that to me is the interesting takeaway of this. (16:53):
undefined
Josh: And hearing that it was fully done, trained on these TPUs, that's very high (16:56):
undefined
Josh: signal to me saying, okay, there is an architecture chips happening. (17:00):
undefined
Josh: There is a real benefit to vertical integration if you could figure out manufacturing (17:03):
undefined
Josh: these compute units at scale. (17:06):
undefined
Josh: And now the race is on for everyone to do this. (17:08):
undefined
Josh: Because again, using the Apple example, the M series chips, unbelievable, (17:12):
undefined
Josh: and they unlocked the best computers in the world. (17:15):
undefined
Josh: And if companies can really start to refine this vertical integration of their (17:16):
undefined
Josh: own chips, you're going to see that exponential curve go vertical times 10. (17:20):
undefined
Josh: Like it is going to, I suspect that is very... (17:24):
undefined
Josh: Obviously now how we reach AGI faster than people previously thought, (17:28):
undefined
Josh: because the efficiency improvements from those vertical integrations, (17:32):
undefined
Josh: once they're able to manufacture these at scale, are going to be unbelievable. (17:35):
undefined
Josh: And I'm so excited for that to happen in the near future. Google has a big head (17:39):
undefined
Josh: start, but let me tell you, the other companies are not far behind. (17:43):
undefined
Ejaaz: Well, let me introduce you to another big advantage of being the big dog, Google. (17:46):
undefined
Ejaaz: You thought you were going to come on to this episode and listen to us hopping (17:52):
undefined
Ejaaz: on about a generalized model? No. (17:55):
undefined
Ejaaz: You're forgetting Google has many other products in their arsenal, (17:57):
undefined
Ejaaz: and you're forgetting that they can plug in their new state-of-the-art model into all of them. (18:00):
undefined
Ejaaz: So Google not only today announced Gemini 3, but they also announced a different product. (18:05):
undefined
Ejaaz: It's called Google Anti-Gravity, which is basically a new software environment (18:10):
undefined
Ejaaz: for you to code up AI agents, except this time these AI agents are going to (18:15):
undefined
Ejaaz: be super, super smart because they get plugged in with Gemini 3. (18:20):
undefined
Ejaaz: Now, if you remember earlier, I mentioned that one of the cool benchmarks that (18:24):
undefined
Ejaaz: this new model sets is in computer use, which means that it can control your (18:27):
undefined
Ejaaz: computer, it can do things autonomously for you. (18:31):
undefined
Ejaaz: Now, typically, the reason why we haven't really spoken about that on the show (18:34):
undefined
Ejaaz: is that they've been kind of lame. (18:36):
undefined
Ejaaz: Like they can book you a dinner reservation and do different kinds of stuff. (18:38):
undefined
Ejaaz: With this model, it's way more intuitive. It can do way more intelligent tasks (18:42):
undefined
Ejaaz: and it can take a lot more complex work off of your hands such that the value (18:46):
undefined
Ejaaz: that it produces to you over like the eight hours that you take to sleep overnight (18:50):
undefined
Ejaaz: would be considerable for you to actually be serious to use in your enterprise, (18:54):
undefined
Ejaaz: in your business, or just at your at-home lifestyle, right? (18:58):
undefined
Ejaaz: So the point I want to make around here is Google's moat is not just its intelligence (19:01):
undefined
Ejaaz: or ability to create new models. It's not its TPUs. (19:07):
undefined
Ejaaz: It's its distribution. It's the entire product suite that it has that regular (19:10):
undefined
Ejaaz: users like you and I that use Gmail, that use Google Suite can now kind of benefit (19:15):
undefined
Ejaaz: from simply by plugging in that model. (19:19):
undefined
Ejaaz: And I think like products like this, anti-gravity, I bet you, (19:22):
undefined
Ejaaz: Josh, we're going to see a slew of new Google product releases over the next (19:25):
undefined
Ejaaz: couple of weeks simply because they created this model. (19:28):
undefined
Josh: I hope so. I guess the contrarian take is like, okay, how many people are actually (19:32):
undefined
Josh: going to want to use them? (19:36):
undefined
Josh: We just spoke about how Claude is the superior code model. Everyone loves Cursor. (19:37):
undefined
Josh: No one really uses the mobile applications of these. A lot of people are engaging (19:42):
undefined
Josh: with AI on their phone. So maybe it works for the right type of person. (19:46):
undefined
Josh: But Google still does have that product problem where they kind of have a tough time. (19:49):
undefined
Josh: They have the amazing intelligence. They just have a tough time using it. (19:54):
undefined
Josh: I mean, I don't have the Gemini app on my phone. I mostly use Grok and ChatGPT. (19:56):
undefined
Josh: And there is this bar that they (20:00):
undefined
Josh: still need to cross that I think they're trying with Google AI Studio. (20:01):
undefined
Josh: And we had Logan Kilpatrick on, who was the head of that, to talk about it when Nano Banana came out. (20:05):
undefined
Josh: But there is still a bit of a long shot for them to get good at products to actually develop this. (20:08):
undefined
Josh: But what we saw this week is that there was a resounding, overwhelming amount (20:14):
undefined
Josh: of support, to your point, you guys, where the market just believes in Google. (20:17):
undefined
Josh: And in a week where all of the stocks, all of the Mag7 was down, (20:20):
undefined
Josh: Google was the one anomaly. Google was up this week. (20:24):
undefined
Josh: And I think it's because the market is starting to realize, one, (20:26):
undefined
Josh: vertical integration through these TPUs is a huge deal. (20:29):
undefined
Josh: Two, Google has an existing business that is not reliant on AI. (20:33):
undefined
Josh: And sure, AI places a huge hand on that scale, but it is not everything. (20:36):
undefined
Josh: And they are cashflow positive in the absence of AI. So all of this innovation (20:41):
undefined
Josh: that they're doing is really just applying later fluid on top of an already (20:45):
undefined
Josh: great business. And the market's starting to evaluate that properly. (20:48):
undefined
Josh: So Google is positioned very strongly. (20:51):
undefined
Josh: They have very high intelligence. Gemini 3 rocks. (20:53):
undefined
Josh: And I mean, again, we continue on the bull train for Google. (20:56):
undefined
Josh: I am a believer. I am a supporter. (20:59):
undefined
Josh: I am stoked that they have the crown. I assumed it was only a matter of time. (21:01):
undefined
Josh: And now the question is, who's next? (21:04):
undefined
Josh: Who is the next competitor? Who's going to set the next plot on that chart and (21:06):
undefined
Josh: set the vertical trajectory on the exponential curve we're on? (21:10):
undefined
Josh: Do you have any guesses who you think it's going to be? (21:14):
undefined
Ejaaz: Yeah, well, I don't because I don't think it's going to be anyone for a while. (21:16):
undefined
Ejaaz: I said this earlier in the show and I'm going to say it again. I think there's going (21:20):
undefined
Ejaaz: to be a six month period now where either the other model providers don't release (21:23):
undefined
Ejaaz: the model because it's not as good as Google's or they just kind of release (21:29):
undefined
Ejaaz: these kind of mediocre kind of like consumer products that kind of maybe benefit (21:33):
undefined
Ejaaz: certain consumers in one way or another, (21:38):
undefined
Ejaaz: but doesn't really kind of break the generalized model (21:41):
undefined
Ejaaz: standard that Google has just set. (21:44):
undefined
Ejaaz: Just a last point on the kind of Google bull case thesis, they may not just (21:47):
undefined
Ejaaz: play in the same ring as cursor does. (21:51):
undefined
Ejaaz: Like I was critiquing Microsoft on another episode, Josh, do you remember? (21:53):
undefined
Ejaaz: And then I got off that episode and I was just like, Microsoft like dominates (21:57):
undefined
Ejaaz: the enterprise environment. (22:02):
undefined
Ejaaz: All the boomer companies and institutions love Microsoft and they have all their data and memory. (22:04):
undefined
Ejaaz: And just because you and I don't use it or just I'll speak for myself, (22:09):
undefined
Ejaaz: just because I don't use it and I think it's doesn't mean that they're not absolutely crushing. (22:12):
undefined
Ejaaz: Google just came off a hundred billion dollar quarter of revenue. (22:15):
undefined
Ejaaz: That's like the highest they've ever had. So I don't want to be too hasty to (22:19):
undefined
Ejaaz: say that like Google's not going to make it because they can't make a sick consumer (22:23):
undefined
Ejaaz: product like OpenAI can maybe. (22:26):
undefined
Ejaaz: I just think they're maybe playing in different fields. (22:28):
undefined
Ejaaz: But to the point around like I don't think anyone else is going to catch up. (22:31):
undefined
Ejaaz: Look at these comments. I want to show you two comments all right one is from sam altman (22:34):
undefined
Ejaaz: He goes, congrats to Google and Gemini 3. This looks like a great model. (22:38):
undefined
Ejaaz: The other is from the almighty being, Elon Musk, saying, I can't wait to try this out. (22:42):
undefined
Ejaaz: And this is just one of a series of tweets that he's been putting out this week (22:47):
undefined
Ejaaz: saying, can you guys just drop Gemini 3? Because I need to see how good this thing is. (22:49):
undefined
Ejaaz: And the reason why I bring up these two people is both Sam Altman and Elon Musk (22:53):
undefined
Ejaaz: have released new versions of their models, GPT and Grok respectively, (22:57):
undefined
Ejaaz: but it's been the 0.1 upgrade. (23:02):
undefined
Ejaaz: It's GPT 5.1. It is Grok 4.1. and they are almost identical updates. (23:04):
undefined
Ejaaz: You want to know what the biggest and coolest thing about their model updates were? (23:10):
undefined
Ejaaz: Personality traits, which don't get me wrong, is cool. Like I would like my (23:13):
undefined
Ejaaz: model to kind of respond in a very intuitive manner and get me, (23:16):
undefined
Ejaaz: but it's nowhere near the state of the art standard that we've just seen broken by Gemini. (23:19):
undefined
Ejaaz: So the point I'm making is, I think these two companies might have run out of fuel for the near term. (23:23):
undefined
Josh: Grok is going to be next. You think? They're the next one. By the end of Q1, (23:28):
undefined
Josh: Grok will have the crown. (23:32):
undefined
Josh: Why? And I assume by a fairly large margin. But I assume it will be a different type of crown. (23:33):
undefined
Josh: And this is where I'm really excited to see how these models progress. (23:39):
undefined
Josh: We spoke a little bit earlier about how Cursor is kind of the coding model. (23:42):
undefined
Josh: Google has a very deep understanding of the real world and physics and video (23:46):
undefined
Josh: and understanding how that works. (23:50):
undefined
Josh: Grok and the XAI team are very focused on the pursuit of truth and information. (23:51):
undefined
Josh: And I think that's kind of the alley that we'll see them going down. (23:56):
undefined
Josh: So they have the real-time data with X. They have the pursuit of truth. (23:59):
undefined
Josh: And where Google and OpenAI and all these other companies are trained on an (24:03):
undefined
Josh: existing data set, the XAI team and the Grok team are developing an entirely (24:06):
undefined
Josh: new synthetic data set that is maximally truth-seeking. (24:10):
undefined
Josh: And we saw that early version with Grokopedia that should provide the most accurate (24:15):
undefined
Josh: and I guess thoughtful information. It should be the best of thinking because (24:19):
undefined
Josh: it's the closest to source truth. So while I think, (24:23):
undefined
Josh: Gemini will probably be better at physics and video and understanding the real (24:26):
undefined
Josh: world for quite some time. (24:30):
undefined
Josh: I suspect Grok will be really good at just communicating via text. (24:31):
undefined
Josh: If text is a modality in which we interface from, Grok should be really good. (24:36):
undefined
Josh: And again, the rate of acceleration, Grok has been around for the least amount (24:40):
undefined
Josh: of time. They're accelerating the fastest. (24:42):
undefined
Josh: And I'm very, very, very excited for a Grok 6, Grok 5, whatever we're at announcement, (24:44):
undefined
Josh: hopefully early next year. So that's the predictions. (24:49):
undefined
Josh: That's the episode. That's Gemini 3.1. It is an unbelievable new model. (24:52):
undefined
Josh: Everyone could try it out. So here's how you try it out. (24:56):
undefined
Josh: I believe you need to be a Google premium plus subscriber, whatever it's called. It's $20 a month. (24:58):
undefined
Josh: You can go on the Gemini website and it's just a text box and you can play around (25:04):
undefined
Josh: with it. They also have a mobile application. (25:07):
undefined
Josh: It's very easy to download on your phone, play around with it. (25:09):
undefined
Josh: I'd love to see examples of cool things because I think one of the problems (25:12):
undefined
Josh: for me and one of the things I'd love help with from anyone who's listening (25:15):
undefined
Josh: is how do you use this thing to test it? (25:17):
undefined
Josh: What do I ask it? And how are you interfacing with it to get the maximum amount of results from it? (25:20):
undefined
Josh: Because intuitively, I would never think to record myself and ask for feedback, (25:25):
undefined
Josh: but that's a new possibility. (25:28):
undefined
Josh: So I guess the challenge to anyone who's listening is figuring out how to get (25:29):
undefined
Josh: the most out of these new models as these new features get released. (25:33):
undefined
Josh: And Gemini 3 has just opened up the gates to a gazillion new use cases. (25:36):
undefined
Ejaaz: Yeah, I mean, this is a super cool release for Google. (25:41):
undefined
Ejaaz: And weirdly enough, it's not the only release over the last week. (25:46):
undefined
Ejaaz: I mean, I've got a list pulled up here. They've released new Android iOS updates. (25:49):
undefined
Ejaaz: They've got a new search AI mode. They've released anti-gravity that we mentioned earlier. (25:53):
undefined
Ejaaz: We've got CIMA2 research, which we demoed on a previous episode. (25:57):
undefined
Ejaaz: You should definitely go check that out. (26:00):
undefined
Ejaaz: I mean, they are just not stopping and they're a force to reckon with. (26:02):
undefined
Ejaaz: And kind of similar to them, Josh, just to kind of round this episode out and (26:07):
undefined
Ejaaz: thank you guys for listening. (26:11):
undefined
Ejaaz: We are here in Argentina, in (26:13):
undefined
Ejaaz: Buenos Aires. We are kind of meeting some of the fans that are out here. (26:15):
undefined
Ejaaz: And we spoke to one just this afternoon, Josh. And you know what he said to me? Have a guess. (26:18):
undefined
Josh: What's that? (26:23):
undefined
Ejaaz: He said, your podcast Limitless is like the state-of-the-art AI podcast. (26:24):
undefined
Ejaaz: In fact, it is 2X better than any other AI podcast that I've ever heard. And you know what? (26:30):
undefined
Ejaaz: That sounds very similar to (26:35):
undefined
Ejaaz: the Gemini 3. So you could potentially call us the Gemini 3 of AI podcast. (26:36):
undefined
Ejaaz: And so if you're a listener to this, if you are a non-subscriber on our YouTube, (26:41):
undefined
Ejaaz: you should probably click that subscriber button. (26:46):
undefined
Ejaaz: You should probably click that notification button because guess what? (26:48):
undefined
Ejaaz: We've got more episodes coming this week. And guess what? the five star ratings help us out massively. (26:51):
undefined
Ejaaz: So if you enjoyed this episode and if you want to hear more episodes of this (26:56):
undefined
Ejaaz: nature and of cutting edge news in AI, you should give us a follow and we will see (26:59):
undefined
undefined
Josh: They have the world's best AI model ever in history by a shockingly large margin. (00:04):
undefined
Josh: Gemini 3.0 just got released. (00:09):
undefined
Josh: It's available now to anybody in the world to go use it. And the benchmarks (00:11):
undefined
Josh: are kind of blowing everyone's expectations out of the water, myself included. (00:15):
undefined
Josh: And most importantly, it places another data point on the chart that shows we (00:18):
undefined
Josh: are continuing to ascend up this exponential curve towards AGI. (00:22):
undefined
Josh: And the roadmap is still intact and we are very quickly moving through it. (00:26):
undefined
Josh: This, EJs, I was just, I was going through the benchmarks before recording this (00:30):
undefined
Josh: and I, it's, it's shocking because we live in this world and yet somehow I'm (00:32):
undefined
Josh: still continually blown away by the progress that's made by these models. So let's get into it. (00:37):
undefined
Josh: Please walk everyone through, tell me, what did Gemini and the Google team just (00:41):
undefined
Josh: release with this 3.0 update? (00:45):
undefined
Ejaaz: People probably think we say the world's best model every week, (00:47):
undefined
Ejaaz: but this time we really, really mean it. (00:51):
undefined
Ejaaz: Like they have blown every single other model provider out the water. (00:53):
undefined
Ejaaz: The things that this thing can do. Well, how about how about I just show you? (00:58):
undefined
Ejaaz: How about I show you, Josh? (01:01):
undefined
Josh: Please, let's see some examples. (01:03):
undefined
Ejaaz: We have a thread here. And Sundar basically says, you can give Gemini 3 anything, (01:04):
undefined
Ejaaz: images, PDFs, scribbles on a napkin, and it'll create whatever you like. (01:09):
undefined
Ejaaz: For example, an image becomes a board game. (01:13):
undefined
Ejaaz: A napkin sketch transforms into a full website. (01:15):
undefined
Ejaaz: And a diagram could turn into an interactive lesson, right? So there's two examples (01:18):
undefined
Ejaaz: I want to show you, Josh. I want to get your opinion on this. (01:22):
undefined
Ejaaz: So number one, there's a short video of someone playing pickleball here and (01:25):
undefined
Ejaaz: he or she rather uploads it into Gemini and says, hey, can you tell me how well (01:30):
undefined
Ejaaz: I've done here and how I can improve my game? (01:36):
undefined
Ejaaz: And it analyzes the entire video. (01:38):
undefined
Ejaaz: It knows that she's wearing a knee brace. It analyzes her positions, (01:40):
undefined
Ejaaz: telling her where she can move to better position herself to score the point. (01:44):
undefined
Ejaaz: That's pretty nuts. But before I get your reaction to that, because Josh, (01:48):
undefined
Ejaaz: I know you're an athlete. I know you're very competitive when it comes to these (01:50):
undefined
Ejaaz: things. So this is a tool you could definitely use. (01:53):
undefined
Ejaaz: The second thing is probably applicable to a lot of listeners on this show. (01:54):
undefined
Ejaaz: They've embedded Gemini 3 into Google search and into new generative UI experiences. (01:59):
undefined
Ejaaz: The way I would summarize this is it basically is very intuitive, Josh. (02:05):
undefined
Ejaaz: It understands what you're asking for without you needing to really kind of explain yourself. (02:10):
undefined
Ejaaz: The example they're showing on the video here is, can you explain the three body problem to me? (02:14):
undefined
Ejaaz: And rather than just give you kind of like this simplistic text, (02:20):
undefined
Ejaaz: which explains the concept, it decides to create a video diagram from scratch (02:22):
undefined
Ejaaz: to show you a visual depiction of how this works. (02:27):
undefined
Ejaaz: Right. Give me your reaction in order from one to two. So starting with the sports. (02:30):
undefined
Josh: Okay, let's go. The first example. (02:35):
undefined
Ejaaz: Yes, sir. (02:36):
undefined
Josh: So this is really cool, the napkin example, where you can scribble something (02:37):
undefined
Josh: down on a piece of paper, it'll generate it in the real world. (02:40):
undefined
Josh: What all these examples are kind of showing me is what we always talk about (02:43):
undefined
Josh: with Google, where it has this awareness of physics, reality, (02:46):
undefined
Josh: and visuals and understanding what it's seeing. (02:51):
undefined
Josh: And all three of these examples are leaning into that. So it leads me to believe (02:53):
undefined
Josh: Gemini really is a multimodal first model, where it's meant to ingest, (02:56):
undefined
Josh: meant to understand the world around us. This example of the chessboard and (03:00):
undefined
Josh: the napkin is amazing because a lot of people oftentimes have sketches. (03:04):
undefined
Josh: You just draw it down on paper and it intuitively understands it. (03:08):
undefined
Josh: But the one that was most surprising to me is the video example. (03:10):
undefined
Josh: Because as far as I'm concerned. (03:13):
undefined
Josh: As far as I'm aware, there has never been a model that can ingest video and (03:16):
undefined
Josh: understand the video that it sees. And if it does exist, I've never tried it before. (03:20):
undefined
Josh: So the idea that you can, I mean, I play baseball growing up. (03:23):
undefined
Josh: I, if I could take a video of myself swinging and get a corrective coach to (03:26):
undefined
Josh: walk me through exactly what was wrong. (03:30):
undefined
Josh: A lot of people play golf. I'm sure who are watching this. (03:32):
undefined
Josh: If you could have a phone recording of you playing golf and it can actually (03:34):
undefined
Josh: critique it and then critique me as if you are a tiger woods, (03:38):
undefined
Josh: critique me as if you are whoever else is good that plays golf. (03:40):
undefined
Josh: I don't know, Rory McIlroy, whatever they are, but like critique Critique me (03:43):
undefined
Josh: as if you are an expert who is really good at golf and can give me some feedback (03:45):
undefined
Josh: on how I can better my swing. (03:49):
undefined
Josh: And what this offers in this one just narrow case example is now you have this (03:51):
undefined
Josh: personal tutor that can do anything. If you're dancing, if you're doing anything (03:54):
undefined
Josh: physical, if you're whatever it is, it can evaluate things for you. (03:58):
undefined
Josh: Even if you have a video, a podcast, EJS, we uploaded to Gemini 3.0, (04:01):
undefined
Josh: it could critique us. What did we do well? What did we not? (04:05):
undefined
Josh: What did the visuals look like? How can we improve them? And that awareness (04:07):
undefined
Josh: of video is like really cool. (04:10):
undefined
Ejaaz: Yeah, I just want to say, I think the closest we got to this was with GPT, (04:13):
undefined
Ejaaz: where you can upload an image, like what's under my car bonnet and say, (04:18):
undefined
Ejaaz: hey, what's wrong? My car stopped working. (04:22):
undefined
Ejaaz: And it can kind of like identify the point that you need to kind of like change, (04:24):
undefined
Ejaaz: change the oil, blah, blah, blah. (04:28):
undefined
Ejaaz: But that's just a static image. To go from that to live video and for it to (04:29):
undefined
Ejaaz: analyze all the frames in that video and then give you a response on that is (04:35):
undefined
Ejaaz: a massive leap upwards. We just haven't seen that anyway. So yeah, you're right. (04:38):
undefined
Josh: It's amazing. And every example we go through, it kind of breaks the mold of (04:42):
undefined
Josh: what I believe should be possible. (04:45):
undefined
Josh: And I find that it's going to be difficult to use Gemini 3.0 because there are (04:47):
undefined
Josh: so many possibilities now that have not existed previously. You kind of need (04:50):
undefined
Josh: to relearn how to engage with AI because it's so capable. (04:54):
undefined
Josh: And there's a fourth example here that I just want to touch on briefly, (04:57):
undefined
Josh: which was also cool, is that it works just as well for the other things. (05:00):
undefined
Josh: The example is a trip planning one where it starts to plan a trip and a vacation. (05:03):
undefined
Josh: And it shows you a full list that is fully interactive of all the places broken up day by day. (05:07):
undefined
Josh: And there's an option that you could just choose visual layout. (05:12):
undefined
Josh: And you see on the screen here, it'll take every single day of your trip, (05:15):
undefined
Josh: break it into images, section it out into this really nice visual grid. (05:18):
undefined
Josh: So what I'm seeing the themes here are, okay, real world understanding, (05:22):
undefined
Josh: video first, and really nice presentation, which I think a lot of models sometimes struggle on. (05:25):
undefined
Josh: Demo's out of control. I'm excited to use it. Everyone else can use it now. (05:31):
undefined
Josh: It's live. Now I want to get to benchmarks, EJS, because this is where things (05:35):
undefined
Josh: get kind of crazy, where we could actually compare one to another and see exactly (05:39):
undefined
Josh: how impressive this is relative to everybody else. So please, we have the card here. (05:43):
undefined
Josh: Walk us through what we're seeing in this model card and all the specs that we need to know. (05:47):
undefined
Ejaaz: As you guys probably know by now, benchmarks is typically how we evaluate any (05:50):
undefined
Ejaaz: typical AI model against each other. (05:54):
undefined
Ejaaz: And they're measured against a range of different benchmarks. (05:57):
undefined
Ejaaz: A benchmark can be considered as sort of like a test. (05:59):
undefined
Ejaaz: Now, right at the top, you've got humanity's last exam. This is by default, (06:01):
undefined
Ejaaz: the hardest exam that an AI model is tested against. (06:06):
undefined
Ejaaz: And it's kind of like an academic reasoning test with no tools accessible to it. (06:10):
undefined
Ejaaz: It scored a very impressive 37.5%, which is more than I think is about a 15% (06:14):
undefined
Ejaaz: increase from its previous model. (06:21):
undefined
Ejaaz: Very, very impressive. But what really blew my mind was the second stat listed (06:23):
undefined
Ejaaz: here, which is the ARK AGI 2 benchmark. (06:28):
undefined
Ejaaz: Josh, when I say this 2x'd the previous state-of-the-art model, I absolutely mean it. (06:31):
undefined
Ejaaz: In fact, let me just show you this chart here. (06:38):
undefined
Ejaaz: Now, you may notice a couple of bustly specs here. (06:41):
undefined
Ejaaz: GPT-5 Pro, Grok 4 Thinking. And then can you see that outlier right at the top (06:45):
undefined
Ejaaz: right. Do you see that, Josh? (06:51):
undefined
Josh: That's insane. The two outliers. (06:53):
undefined
Ejaaz: The two outliers. So these are the Gemini 3 Pro and the Gemini 3 deep thinking model. (06:55):
undefined
Ejaaz: Deep thinking being like, you know, a large number that can like kind of give (07:02):
undefined
Ejaaz: you a more research response. They are a... (07:04):
undefined
Ejaaz: Stand out from every single other model. And the reason why this is so crazy, (07:07):
undefined
Ejaaz: well, there's a few reasons. (07:12):
undefined
Ejaaz: Number one, all the other model progressions, as you can see over time, (07:13):
undefined
Ejaaz: has been kind of impressive, but kind of small. (07:18):
undefined
Ejaaz: Like they've been a good jump. It's been impressive, but it hasn't been as impressive (07:23):
undefined
Ejaaz: to be like, oh, you know, another model provider couldn't catch up. (07:28):
undefined
Ejaaz: These results from Google literally put it miles ahead of every other model. (07:31):
undefined
Ejaaz: So when I sat at this chart, I think, wow, Google probably has the lead for (07:36):
undefined
Ejaaz: another six months and in six months time, they're going to have a more impressive model by then. (07:40):
undefined
Ejaaz: So at this point, I'm kind of thinking, can anyone catch up to Google? (07:46):
undefined
Ejaaz: Josh, do you have any reactions to this benchmark? (07:49):
undefined
Josh: This is the chart that like the first thing I said to myself when I saw this (07:51):
undefined
Josh: is like, oh, my God, there is no wall there. (07:55):
undefined
Josh: We are not going to stop scaling. The scaling will still apply because these (07:57):
undefined
Josh: two new data points that we have blow everything else out of the water. (08:01):
undefined
Josh: And this is how exponential growth happens. (08:03):
undefined
Josh: It seems like a really small cluster down there in the bottom, (08:05):
undefined
Josh: but the reality is that was the top just a couple hours ago. (08:08):
undefined
Josh: And Gemini kind of refactored this entire chart to make it seem like it's so (08:12):
undefined
Josh: small because the progress is so high. (08:17):
undefined
Josh: And although Gemini 3.0 thinking is seemingly the most impressive, (08:18):
undefined
Josh: the really anomaly chart is the Gemini 3.0 Pro, which is basically a vertical (08:23):
undefined
Josh: line up from these other models, where the score is higher, but the cost is (08:27):
undefined
Josh: actually slightly lower. (08:32):
undefined
Josh: And if you connect a dot between these averages, you start to see literal vertical (08:33):
undefined
Josh: line in terms of improvement and acceleration in these models. (08:37):
undefined
Josh: And that to me shows that there is no scaling wall that we're hitting. (08:40):
undefined
Josh: Like we can continue to scale resources, energy, compute, and we could continue (08:45):
undefined
Josh: along this path towards AGI in a world where some people were saying, (08:48):
undefined
Josh: we don't know if it continues. The answer to me is very clearly, it continues. (08:51):
undefined
Josh: This is a step much closer to AGI. (08:55):
undefined
Josh: And again, that real world understanding makes it feel much closer to AGI than it ever has before. (08:58):
undefined
Josh: Because now it really like intuitively understands the world through video, (09:04):
undefined
Josh: through photo, through audio, through basically every sensory input we have (09:08):
undefined
Josh: outside of what taste and feel. (09:11):
undefined
Josh: So this to me, I saw this chart. I was like, oh, my God, Gemini, (09:13):
undefined
Josh: you really outdid yourselves. (09:17):
undefined
Ejaaz: I'm just going to be honest. I think over the last couple of months, (09:18):
undefined
Ejaaz: I've been getting a little bored with the models that have been released by other model providers. (09:22):
undefined
Ejaaz: And it led me to think that we're not going to make many breakthroughs until (09:27):
undefined
Ejaaz: some model provider figures out a new, unique way to train their model. (09:32):
undefined
Ejaaz: Gemini or Google has convinced me otherwise with this release. (09:37):
undefined
Ejaaz: But I know you guys are probably like fed up with listening to us hop on about benchmarks. (09:41):
undefined
Ejaaz: So how about I materialize that for you in a much more easy to understand way, right? (09:45):
undefined
Ejaaz: So here are the four big takeaways that you need to learn about Gemini 3. (09:50):
undefined
Ejaaz: Number one, for its intelligence, for the intelligence that you're getting, (09:55):
undefined
Ejaaz: it is not that super expensive. (10:00):
undefined
Ejaaz: Google trained this from scratch, as this tweet says, using their own GPU infrastructure. (10:03):
undefined
Ejaaz: And it used this kind of like layout called a mixture of experts, (10:07):
undefined
Ejaaz: which basically means that whenever you prompt the model, it's not going to use the entire model. (10:11):
undefined
Ejaaz: So it actually ends up being cheaper than what it could eventually become. (10:16):
undefined
Ejaaz: One million token context input, 64K token output. We'll get to the costs in (10:19):
undefined
Ejaaz: a bit of a second, But the point that I'm making here is that it's not as expensive (10:24):
undefined
Ejaaz: as you would expect for the intelligence that you're getting. (10:27):
undefined
Ejaaz: Now, if you compared Gemini 3 to GPT 5.1 from OpenAI, on a relative basis, (10:29):
undefined
Ejaaz: it is more expensive. But for the jump in intelligence that you're getting, (10:36):
undefined
Ejaaz: it's way better. So it's, in my opinion, worth it. (10:39):
undefined
Ejaaz: Number two, when it comes to computer use, so that means letting the AI model (10:42):
undefined
Ejaaz: control your computer and do tasks for you whilst you go do something else. (10:47):
undefined
Ejaaz: It is state of the art. It is the best here. They measured it against a benchmark (10:52):
undefined
Ejaaz: called ScreenSpot Pro, which is a benchmark which kind of like analyzes its (10:56):
undefined
Ejaaz: ability to understand images and visuals on a desktop. It just absolutely crushes it. (11:00):
undefined
Ejaaz: Number three, it is the best AI for math by far. (11:06):
undefined
Ejaaz: So again, the point I'm making here or the theme that we're seeing here is it's (11:09):
undefined
Ejaaz: not just good at one thing, it's good at many things, which makes it the best (11:13):
undefined
Ejaaz: generalist AI model in the world right now, by far. (11:17):
undefined
Ejaaz: And the final thing, Josh, and this is where it might slip up. (11:21):
undefined
Ejaaz: I'm curious to get your take on this. (11:24):
undefined
Ejaaz: It is insanely good at coding, but we don't quite know if it is the best at coding yet. (11:26):
undefined
Ejaaz: What I mean by that is it completely crushed everyone else on one coding benchmark, (11:32):
undefined
Ejaaz: but the coding benchmark that matters, which is the software engineering, (11:37):
undefined
Ejaaz: SWE, it didn't do as well as its competitor, Claude 4.5 from Anthropic. (11:40):
undefined
Ejaaz: So those are the four main takeaways. (11:45):
undefined
Josh: I would much prefer a model that understands the world than understands how to code. (11:47):
undefined
Josh: And I think we're starting to see these subset niches where if Anthropic has (11:51):
undefined
Josh: the best coding model, that's great. (11:55):
undefined
Josh: Let them focus on code. Let them narrowly make that the best model. (11:56):
undefined
Josh: Let Google handle everything else. And I think that's what Gemini is focusing (12:00):
undefined
Josh: on. So the code thing doesn't really bother me because I don't care to use Gemini for code. (12:03):
undefined
Josh: I'm happy to be in Claude Camp for code and then use Gemini for everything else. (12:07):
undefined
Josh: And then one of the points earlier that you mentioned on the pricing is. (12:11):
undefined
Josh: I find it a little interesting because it's a little bit more than just a little bit more expensive. (12:14):
undefined
Josh: The pricing, I was looking through it and it's for over 200,000 tokens. (12:18):
undefined
Josh: They're charging $4 for inputs and $18 for outputs. (12:22):
undefined
Josh: Now, relative to GPT 5.1, which just got released, they're charging for a million (12:27):
undefined
Josh: tokens, $1.25 in, $10 out. (12:32):
undefined
Josh: So you're talking about, what is that? that's about $20 versus $1.25 on inputs. (12:35):
undefined
Josh: And that is a fairly significant margin that you're paying for this quality. (12:42):
undefined
Josh: So we're starting to see the trade-offs happening on that Pareto curve that (12:45):
undefined
Josh: we talked about in a few episodes earlier, where there are trade-offs coming (12:48):
undefined
Josh: in terms of cost and quality. (12:52):
undefined
Josh: And it's clear that while OpenAI may have optimized for cost, (12:53):
undefined
Josh: Google is kind of optimizing for a little further up the cost curve in exchange for super high quality. (12:57):
undefined
Josh: And it seems like this is kind of a balanced data point for now because unless (13:02):
undefined
Josh: you are using this via API and you're requiring a ton of tokens, (13:06):
undefined
Josh: a $20 a month Google membership will get you all of the use that you need. And that is just fine. (13:09):
undefined
Josh: So in terms of a usability perspective, I think that's okay. (13:14):
undefined
Josh: But it's just an interesting thing to know is that this is a better model. (13:18):
undefined
Josh: It is also more expensive. (13:21):
undefined
Josh: And that is a trade-off that was made. And in the case that OpenAI decides to (13:22):
undefined
Josh: make this trade-off with GPT-6 or Grok decides to make this with Grok-5, (13:25):
undefined
Josh: or Grok 6. I'm losing track of all these models now. I think we're going to (13:30):
undefined
Josh: start to see the dynamic shift in terms of that Pareto curve and what model (13:34):
undefined
Josh: architects decide to remove and add. (13:37):
undefined
Josh: And in this case, it looks like Google added quality, but they also did add (13:39):
undefined
Josh: quite a significant cost increase. (13:42):
undefined
Ejaaz: I personally don't think it matters. I think it's a nothing burger. (13:44):
undefined
Ejaaz: I think that if Google wanted to make it affordable for everyone, (13:46):
undefined
Ejaaz: including the developers that want to get API access, that think it might be (13:50):
undefined
Ejaaz: too expensive, they could subsidize it. (13:52):
undefined
Ejaaz: They are a cash flow giant. They have enough money to do that. (13:54):
undefined
Ejaaz: OpenAI has been doing that for so long now that it doesn't even matter. (13:57):
undefined
Ejaaz: I don't see any reason why Google couldn't do that. (14:00):
undefined
Ejaaz: The other reason is Google just released their latest TPU, which is their GPU (14:02):
undefined
Ejaaz: that they use to train their models and inference their models. (14:08):
undefined
Ejaaz: And typically with every generation, we get a much cheaper cost of inference. (14:11):
undefined
Ejaaz: I think by the time that they release their next generation model, (14:15):
undefined
Ejaaz: which might be, you know, Gemini 3.1, we're going to see a considerable reduction (14:18):
undefined
Ejaaz: in the cost for using Gemini 3 Pro and Gemini Pro deep research. (14:24):
undefined
Ejaaz: So I'm not too worried about that. I think it's kind of like a short-term problem (14:29):
undefined
Ejaaz: and not a long-term problem. (14:32):
undefined
Ejaaz: But kind of speaking of TPUs, I just want to take a moment to really kind of (14:34):
undefined
Ejaaz: belay the point that using their own TPUs to train a state-of-the-art model (14:39):
undefined
Ejaaz: that is 2x better than the previous state-of-the-art model and probably puts (14:44):
undefined
Ejaaz: them in a six-month lead after Google started off on the back foot, (14:48):
undefined
Ejaaz: creating probably the worst model I've ever seen and (14:53):
undefined
Ejaaz: changing that all around in what's it under two years is (14:56):
undefined
Ejaaz: nothing short of insanity tpus is (14:59):
undefined
Ejaaz: uh google's kind of version of the gpu gpu is (15:02):
undefined
Ejaaz: kind of like what nvidia controls the monopoly over this is the hardware that (15:05):
undefined
Ejaaz: you use to train your ai and inference your ai the uh unique part here is that (15:09):
undefined
Ejaaz: google's never used an nvidia gpu uh or in any considerable way to train their (15:13):
undefined
Ejaaz: models they've always trained it in-house and that's such a difficult and tricky (15:19):
undefined
Ejaaz: thing to do because designing and building these (15:23):
undefined
Ejaaz: TPUs at scale, these GPUs at scale is a super hard and complex thing. (15:25):
undefined
Ejaaz: You need so much talent, you need so much expertise and insight to be able to do that. (15:29):
undefined
Ejaaz: The unique thing about Google's TPUs, well, there's two main takeaways. (15:33):
undefined
Ejaaz: Number one, it's cheaper to train the same amount of intelligence that an NVIDIA (15:36):
undefined
Ejaaz: GPU is. So it's more cost efficient. (15:41):
undefined
Ejaaz: And the second way is, and this is their secret sauce, you can stack those TPUs (15:43):
undefined
Ejaaz: on top of each other in a really scalable way that you can start training really, really large models. (15:47):
undefined
Ejaaz: If you wanted to train the same size model with nvidia (15:53):
undefined
Ejaaz: gpus it would cost way more and it would take way (15:57):
undefined
Ejaaz: longer so google made a really risky and big bet about a decade ago saying we're (16:00):
undefined
Ejaaz: going to build our infrastructure in-house and we're not going to rely on nvidia (16:05):
undefined
Ejaaz: and we're going to benefit from the full stack experience and this model is (16:07):
undefined
Ejaaz: a prime example of that bet paying off so i just want to call them out like (16:11):
undefined
Ejaaz: it's not like google has gotten lucky here they've been planning it for a while now The (16:15):
undefined
Josh: Interesting thing to me is that this is the first number one model in the world (16:19):
undefined
Josh: built on something other than an NVIDIA GPU. (16:24):
undefined
Josh: And that's fairly significant because every company in the world is trying, (16:27):
undefined
Josh: but this is proof that it's actually possible. (16:31):
undefined
Josh: And I think when we talk about Tesla and AI5 and the XAI team, (16:32):
undefined
Josh: when we talk about OpenAI working with whoever they're working with to build (16:37):
undefined
Josh: their own in-house GPUs, I think this sets a precedent that it is possible. (16:41):
undefined
Josh: And I suspect that will result in more companies putting their foot on the gas. (16:45):
undefined
Josh: When it comes to kind of destructing part of NVIDIA's monopoly that it holds (16:49):
undefined
Josh: over GPUs. So that to me is the interesting takeaway of this. (16:53):
undefined
Josh: And hearing that it was fully done, trained on these TPUs, that's very high (16:56):
undefined
Josh: signal to me saying, okay, there is an architecture chips happening. (17:00):
undefined
Josh: There is a real benefit to vertical integration if you could figure out manufacturing (17:03):
undefined
Josh: these compute units at scale. (17:06):
undefined
Josh: And now the race is on for everyone to do this. (17:08):
undefined
Josh: Because again, using the Apple example, the M series chips, unbelievable, (17:12):
undefined
Josh: and they unlocked the best computers in the world. (17:15):
undefined
Josh: And if companies can really start to refine this vertical integration of their (17:16):
undefined
Josh: own chips, you're going to see that exponential curve go vertical times 10. (17:20):
undefined
Josh: Like it is going to, I suspect that is very... (17:24):
undefined
Josh: Obviously now how we reach AGI faster than people previously thought, (17:28):
undefined
Josh: because the efficiency improvements from those vertical integrations, (17:32):
undefined
Josh: once they're able to manufacture these at scale, are going to be unbelievable. (17:35):
undefined
Josh: And I'm so excited for that to happen in the near future. Google has a big head (17:39):
undefined
Josh: start, but let me tell you, the other companies are not far behind. (17:43):
undefined
Ejaaz: Well, let me introduce you to another big advantage of being the big dog, Google. (17:46):
undefined
Ejaaz: You thought you were going to come on to this episode and listen to us hopping (17:52):
undefined
Ejaaz: on about a generalized model? No. (17:55):
undefined
Ejaaz: You're forgetting Google has many other products in their arsenal, (17:57):
undefined
Ejaaz: and you're forgetting that they can plug in their new state-of-the-art model into all of them. (18:00):
undefined
Ejaaz: So Google not only today announced Gemini 3, but they also announced a different product. (18:05):
undefined
Ejaaz: It's called Google Anti-Gravity, which is basically a new software environment (18:10):
undefined
Ejaaz: for you to code up AI agents, except this time these AI agents are going to (18:15):
undefined
Ejaaz: be super, super smart because they get plugged in with Gemini 3. (18:20):
undefined
Ejaaz: Now, if you remember earlier, I mentioned that one of the cool benchmarks that (18:24):
undefined
Ejaaz: this new model sets is in computer use, which means that it can control your (18:27):
undefined
Ejaaz: computer, it can do things autonomously for you. (18:31):
undefined
Ejaaz: Now, typically, the reason why we haven't really spoken about that on the show (18:34):
undefined
Ejaaz: is that they've been kind of lame. (18:36):
undefined
Ejaaz: Like they can book you a dinner reservation and do different kinds of stuff. (18:38):
undefined
Ejaaz: With this model, it's way more intuitive. It can do way more intelligent tasks (18:42):
undefined
Ejaaz: and it can take a lot more complex work off of your hands such that the value (18:46):
undefined
Ejaaz: that it produces to you over like the eight hours that you take to sleep overnight (18:50):
undefined
Ejaaz: would be considerable for you to actually be serious to use in your enterprise, (18:54):
undefined
Ejaaz: in your business, or just at your at-home lifestyle, right? (18:58):
undefined
Ejaaz: So the point I want to make around here is Google's moat is not just its intelligence (19:01):
undefined
Ejaaz: or ability to create new models. It's not its TPUs. (19:07):
undefined
Ejaaz: It's its distribution. It's the entire product suite that it has that regular (19:10):
undefined
Ejaaz: users like you and I that use Gmail, that use Google Suite can now kind of benefit (19:15):
undefined
Ejaaz: from simply by plugging in that model. (19:19):
undefined
Ejaaz: And I think like products like this, anti-gravity, I bet you, (19:22):
undefined
Ejaaz: Josh, we're going to see a slew of new Google product releases over the next (19:25):
undefined
Ejaaz: couple of weeks simply because they created this model. (19:28):
undefined
Josh: I hope so. I guess the contrarian take is like, okay, how many people are actually (19:32):
undefined
Josh: going to want to use them? (19:36):
undefined
Josh: We just spoke about how Claude is the superior code model. Everyone loves Cursor. (19:37):
undefined
Josh: No one really uses the mobile applications of these. A lot of people are engaging (19:42):
undefined
Josh: with AI on their phone. So maybe it works for the right type of person. (19:46):
undefined
Josh: But Google still does have that product problem where they kind of have a tough time. (19:49):
undefined
Josh: They have the amazing intelligence. They just have a tough time using it. (19:54):
undefined
Josh: I mean, I don't have the Gemini app on my phone. I mostly use Grok and ChatGPT. (19:56):
undefined
Josh: And there is this bar that they (20:00):
undefined
Josh: still need to cross that I think they're trying with Google AI Studio. (20:01):
undefined
Josh: And we had Logan Kilpatrick on, who was the head of that, to talk about it when Nano Banana came out. (20:05):
undefined
Josh: But there is still a bit of a long shot for them to get good at products to actually develop this. (20:08):
undefined
Josh: But what we saw this week is that there was a resounding, overwhelming amount (20:14):
undefined
Josh: of support, to your point, you guys, where the market just believes in Google. (20:17):
undefined
Josh: And in a week where all of the stocks, all of the Mag7 was down, (20:20):
undefined
Josh: Google was the one anomaly. Google was up this week. (20:24):
undefined
Josh: And I think it's because the market is starting to realize, one, (20:26):
undefined
Josh: vertical integration through these TPUs is a huge deal. (20:29):
undefined
Josh: Two, Google has an existing business that is not reliant on AI. (20:33):
undefined
Josh: And sure, AI places a huge hand on that scale, but it is not everything. (20:36):
undefined
Josh: And they are cashflow positive in the absence of AI. So all of this innovation (20:41):
undefined
Josh: that they're doing is really just applying later fluid on top of an already (20:45):
undefined
Josh: great business. And the market's starting to evaluate that properly. (20:48):
undefined
Josh: So Google is positioned very strongly. (20:51):
undefined
Josh: They have very high intelligence. Gemini 3 rocks. (20:53):
undefined
Josh: And I mean, again, we continue on the bull train for Google. (20:56):
undefined
Josh: I am a believer. I am a supporter. (20:59):
undefined
Josh: I am stoked that they have the crown. I assumed it was only a matter of time. (21:01):
undefined
Josh: And now the question is, who's next? (21:04):
undefined
Josh: Who is the next competitor? Who's going to set the next plot on that chart and (21:06):
undefined
Josh: set the vertical trajectory on the exponential curve we're on? (21:10):
undefined
Josh: Do you have any guesses who you think it's going to be? (21:14):
undefined
Ejaaz: Yeah, well, I don't because I don't think it's going to be anyone for a while. (21:16):
undefined
Ejaaz: I said this earlier in the show and I'm going to say it again. I think there's going (21:20):
undefined
Ejaaz: to be a six month period now where either the other model providers don't release (21:23):
undefined
Ejaaz: the model because it's not as good as Google's or they just kind of release (21:29):
undefined
Ejaaz: these kind of mediocre kind of like consumer products that kind of maybe benefit (21:33):
undefined
Ejaaz: certain consumers in one way or another, (21:38):
undefined
Ejaaz: but doesn't really kind of break the generalized model (21:41):
undefined
Ejaaz: standard that Google has just set. (21:44):
undefined
Ejaaz: Just a last point on the kind of Google bull case thesis, they may not just (21:47):
undefined
Ejaaz: play in the same ring as cursor does. (21:51):
undefined
Ejaaz: Like I was critiquing Microsoft on another episode, Josh, do you remember? (21:53):
undefined
Ejaaz: And then I got off that episode and I was just like, Microsoft like dominates (21:57):
undefined
Ejaaz: the enterprise environment. (22:02):
undefined
Ejaaz: All the boomer companies and institutions love Microsoft and they have all their data and memory. (22:04):
undefined
Ejaaz: And just because you and I don't use it or just I'll speak for myself, (22:09):
undefined
Ejaaz: just because I don't use it and I think it's doesn't mean that they're not absolutely crushing. (22:12):
undefined
Ejaaz: Google just came off a hundred billion dollar quarter of revenue. (22:15):
undefined
Ejaaz: That's like the highest they've ever had. So I don't want to be too hasty to (22:19):
undefined
Ejaaz: say that like Google's not going to make it because they can't make a sick consumer (22:23):
undefined
Ejaaz: product like OpenAI can maybe. (22:26):
undefined
Ejaaz: I just think they're maybe playing in different fields. (22:28):
undefined
Ejaaz: But to the point around like I don't think anyone else is going to catch up. (22:31):
undefined
Ejaaz: Look at these comments. I want to show you two comments all right one is from sam altman (22:34):
undefined
Ejaaz: He goes, congrats to Google and Gemini 3. This looks like a great model. (22:38):
undefined
Ejaaz: The other is from the almighty being, Elon Musk, saying, I can't wait to try this out. (22:42):
undefined
Ejaaz: And this is just one of a series of tweets that he's been putting out this week (22:47):
undefined
Ejaaz: saying, can you guys just drop Gemini 3? Because I need to see how good this thing is. (22:49):
undefined
Ejaaz: And the reason why I bring up these two people is both Sam Altman and Elon Musk (22:53):
undefined
Ejaaz: have released new versions of their models, GPT and Grok respectively, (22:57):
undefined
Ejaaz: but it's been the 0.1 upgrade. (23:02):
undefined
Ejaaz: It's GPT 5.1. It is Grok 4.1. and they are almost identical updates. (23:04):
undefined
Ejaaz: You want to know what the biggest and coolest thing about their model updates were? (23:10):
undefined
Ejaaz: Personality traits, which don't get me wrong, is cool. Like I would like my (23:13):
undefined
Ejaaz: model to kind of respond in a very intuitive manner and get me, (23:16):
undefined
Ejaaz: but it's nowhere near the state of the art standard that we've just seen broken by Gemini. (23:19):
undefined
Ejaaz: So the point I'm making is, I think these two companies might have run out of fuel for the near term. (23:23):
undefined
Josh: Grok is going to be next. You think? They're the next one. By the end of Q1, (23:28):
undefined
Josh: Grok will have the crown. (23:32):
undefined
Josh: Why? And I assume by a fairly large margin. But I assume it will be a different type of crown. (23:33):
undefined
Josh: And this is where I'm really excited to see how these models progress. (23:39):
undefined
Josh: We spoke a little bit earlier about how Cursor is kind of the coding model. (23:42):
undefined
Josh: Google has a very deep understanding of the real world and physics and video (23:46):
undefined
Josh: and understanding how that works. (23:50):
undefined
Josh: Grok and the XAI team are very focused on the pursuit of truth and information. (23:51):
undefined
Josh: And I think that's kind of the alley that we'll see them going down. (23:56):
undefined
Josh: So they have the real-time data with X. They have the pursuit of truth. (23:59):
undefined
Josh: And where Google and OpenAI and all these other companies are trained on an (24:03):
undefined
Josh: existing data set, the XAI team and the Grok team are developing an entirely (24:06):
undefined
Josh: new synthetic data set that is maximally truth-seeking. (24:10):
undefined
Josh: And we saw that early version with Grokopedia that should provide the most accurate (24:15):
undefined
Josh: and I guess thoughtful information. It should be the best of thinking because (24:19):
undefined
Josh: it's the closest to source truth. So while I think, (24:23):
undefined
Josh: Gemini will probably be better at physics and video and understanding the real (24:26):
undefined
Josh: world for quite some time. (24:30):
undefined
Josh: I suspect Grok will be really good at just communicating via text. (24:31):
undefined
Josh: If text is a modality in which we interface from, Grok should be really good. (24:36):
undefined
Josh: And again, the rate of acceleration, Grok has been around for the least amount (24:40):
undefined
Josh: of time. They're accelerating the fastest. (24:42):
undefined
Josh: And I'm very, very, very excited for a Grok 6, Grok 5, whatever we're at announcement, (24:44):
undefined
Josh: hopefully early next year. So that's the predictions. (24:49):
undefined
Josh: That's the episode. That's Gemini 3.1. It is an unbelievable new model. (24:52):
undefined
Josh: Everyone could try it out. So here's how you try it out. (24:56):
undefined
Josh: I believe you need to be a Google premium plus subscriber, whatever it's called. It's $20 a month. (24:58):
undefined
Josh: You can go on the Gemini website and it's just a text box and you can play around (25:04):
undefined
Josh: with it. They also have a mobile application. (25:07):
undefined
Josh: It's very easy to download on your phone, play around with it. (25:09):
undefined
Josh: I'd love to see examples of cool things because I think one of the problems (25:12):
undefined
Josh: for me and one of the things I'd love help with from anyone who's listening (25:15):
undefined
Josh: is how do you use this thing to test it? (25:17):
undefined
Josh: What do I ask it? And how are you interfacing with it to get the maximum amount of results from it? (25:20):
undefined
Josh: Because intuitively, I would never think to record myself and ask for feedback, (25:25):
undefined
Josh: but that's a new possibility. (25:28):
undefined
Josh: So I guess the challenge to anyone who's listening is figuring out how to get (25:29):
undefined
Josh: the most out of these new models as these new features get released. (25:33):
undefined
Josh: And Gemini 3 has just opened up the gates to a gazillion new use cases. (25:36):
undefined
Ejaaz: Yeah, I mean, this is a super cool release for Google. (25:41):
undefined
Ejaaz: And weirdly enough, it's not the only release over the last week. (25:46):
undefined
Ejaaz: I mean, I've got a list pulled up here. They've released new Android iOS updates. (25:49):
undefined
Ejaaz: They've got a new search AI mode. They've released anti-gravity that we mentioned earlier. (25:53):
undefined
Ejaaz: We've got CIMA2 research, which we demoed on a previous episode. (25:57):
undefined
Ejaaz: You should definitely go check that out. (26:00):
undefined
Ejaaz: I mean, they are just not stopping and they're a force to reckon with. (26:02):
undefined
Ejaaz: And kind of similar to them, Josh, just to kind of round this episode out and (26:07):
undefined
Ejaaz: thank you guys for listening. (26:11):
undefined
Ejaaz: We are here in Argentina, in (26:13):
undefined
Ejaaz: Buenos Aires. We are kind of meeting some of the fans that are out here. (26:15):
undefined
Ejaaz: And we spoke to one just this afternoon, Josh. And you know what he said to me? Have a guess. (26:18):
undefined
Josh: What's that? (26:23):
undefined
Ejaaz: He said, your podcast Limitless is like the state-of-the-art AI podcast. (26:24):
undefined
Ejaaz: In fact, it is 2X better than any other AI podcast that I've ever heard. And you know what? (26:30):
undefined
Ejaaz: That sounds very similar to (26:35):
undefined
Ejaaz: the Gemini 3. So you could potentially call us the Gemini 3 of AI podcast. (26:36):
undefined
Ejaaz: And so if you're a listener to this, if you are a non-subscriber on our YouTube, (26:41):
undefined
Ejaaz: you should probably click that subscriber button. (26:46):
undefined
Ejaaz: You should probably click that notification button because guess what? (26:48):
undefined
Ejaaz: We've got more episodes coming this week. And guess what? the five star ratings help us out massively. (26:51):
undefined
Ejaaz: So if you enjoyed this episode and if you want to hear more episodes of this (26:56):
undefined
Ejaaz: nature and of cutting edge news in AI, you should give us a follow and we will see (26:59):
undefined
Popular Podcasts
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.
My Favorite Murder with Karen Kilgariff and Georgia Hardstark
My Favorite Murder is a true crime comedy podcast hosted by Karen Kilgariff and Georgia Hardstark. Each week, Karen and Georgia share compelling true crimes and hometown stories from friends and listeners. Since MFM launched in January of 2016, Karen and Georgia have shared their lifelong interest in true crime and have covered stories of infamous serial killers like the Night Stalker, mysterious cold cases, captivating cults, incredible survivor stories and important events from history like the Tulsa race massacre of 1921. My Favorite Murder is part of the Exactly Right podcast network that provides a platform for bold, creative voices to bring to life provocative, entertaining and relatable stories for audiences everywhere. The Exactly Right roster of podcasts covers a variety of topics including historic true crime, comedic interviews and news, science, pop culture and more. Podcasts on the network include Buried Bones with Kate Winkler Dawson and Paul Holes, That's Messed Up: An SVU Podcast, This Podcast Will Kill You, Bananas and more.
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Follow now to get the latest episodes of Dateline NBC completely free, or subscribe to Dateline Premium for ad-free listening and exclusive bonus content: DatelinePremium.com