Gemini Just Got a HUGE Update! (Deep Research Visuals)

Okay, so there's a ton of AI features out there, right? But I consistently keep returning to one of them, which is deep researches that are now present. And Chad GBT, Gemini, Claude, Perplexity has a version, Brock has a version. Like basically all the big providers have a version of it, but they're not all created equal. And people know that. We've talked about it before on the channel. on you can use how Gemini Deep Research is sort of a bit more intelligent. It goes a bit more in depth. It makes sense. It's Google, the number one search provider in the world, who's making a deep search product that works with AI. That's what deep research is. If you're not familiar, it's basically like Chat GPT opening 50 tabs, doing 50 Google searches, and then pulling together the top results into one report. And this week, something really interesting happened is because Gemini actually pulled ahead and on their ultra plan, they showed us a type of deep research that I have never seen before. And in today's video, we're going to compare it. What I'm talking about is deep research with visuals, with graphs, with code embedded in it. It's really a multimedia version of what we've seen as textonly documents from all the providers this far. This feature with all of these graphs and visuals is only available on the ultra plan right now, but I expect it to trickle down over time as do all AI features. And then obviously I also expect OpenAI to kind of catch up with this soonish cuz they'll have to. We'll look at some comparisons here between OpenAI Deep Researches and Gemini Deep Researches on this expensive plan. But spoiler alert, it's so good that I think all other competitors will have to catch up to this deep research feature. Let's get into looking at the actual examples and reviewing so you can make an informed decision on which one of these tools fits your needs today. So here's the

first prompt and this is based on some tweet that I caught on Twitter a few days ago I think and it's basically the idea of if divorce rates in country relate to the sizes of their beds. So the question is like hey if you have bigger beds and then people have more space while sleeping is that going to correlate with divorce rates? I don't know just a thought and we're going to use the deep research feature to explore this thought both inside of Gemini and in chat GPT. To make it fair, I included answers to all the follow-up questions in both of these reports when I created it in chat GPT. And then I also adjusted the report in Gemini so it produces the same outcomes. And here's what we got. Let's have a look. This is the Chat GPT report. This is the Gemini report. So let's start with Gemini. Not going to read all of this obviously. Video would be an hour long. And if you want to check the report and double check sources and everything, I'll put a link to this conversation in the description below so you can check it for yourself. And what I want to go to is these summaries here. So first of all, summary of spatial provision. And it says USA usually has king beds that give you almost 100 cm per person. I know a lot of the viewers are US. There you go. 3. 1 ft. Whereas in Scandinavia it's less. It's only 90 cm per person. And in Southern Europe it's only 80. UK down to 75 and France down to 70 with Spain being last at 67 or 75. 70 cm guys. That's not a lot. That's 2. 3 ft. That's like I'm not a very wide person but I would barely fit. So if that is the average bed size, how does that compare? Okay, we can look at the marital stability index. So I guess it created this graph by itself from three different sources. Look at that. Euroativorce. com. Divorce. com. What kind of business is that? Okay. Yeah. So it aggregated the data from three different sources. This is really amazing. No. And then we get to the part that I asked about the correlation archetype A the wided high divorce model USA and Scandinavia. Okay. So high divorce rate cluster is United States, Scandinavia and the Baltics. And low divorce rate cluster would be Italy and Ireland. So not all of them are on the graph. Fair enough. This is more like European countries. Guess that's where the data came from. Yeah. Euro stat. Okay. And then the correlation. So the wide bed high divorce rate model. This cluster provides a stronger support for the hypothesis. It's actually the other way around. Spatial independence correlates with marital dissolution. Huh. It makes me think like maybe this culture of like being independent and doing your own thing isn't good for a marriage which kind of makes sense, you know. And then let's see what else we got here. United Kingdom disrupts the neat pattern of small bed equals stable marriage. And then it created a custom matrix. Ooh. Okay. So Nordic the Nordic paradox wide bets high divorce up here. Italians are an exception. Okay. So I can already see there's no clear correlation here. Okay. Just give me the final solution. What do you think? The conclusion. The relationship is not a simple straight line, but rather a U-shaped curve of stress. When the bed is too narrow, sleep disturbance is high. This leads to psychological stress, resentment, and high likelihood of sleep divorce, which correlates with moderate to high relationship instability. And then the US king-size problem is when the bed is too wide, physical distance is effortless. This facilitates emotional drift and removes the obligatory intimacy of touch. And then just right, it says the European sweet spot. the southern European model beds that are wide enough for basic sleep hygiene but narrow enough to ensure proximity. I would have never thought that is the case but I don't know makes sense. You can look at this yourself. I'm not saying this is a conclusion but what a great report. No, I know took a while for us to review that but I thought that was so interesting and I don't think you could get that out of any other deep research. Let's verify though. Let's look at the one that was created by Chat GPT. It found the same thing. United States large beds here. It analyzes other countries. It pulls a lot from Wikipedia, which I don't love. Again, more Wikipedia here. And then look at that. Rather a short, deep research with no graphs and visuals in between at all. Then it talks about the divorce rates, of course. And then it goes into examining the relationship. I really care about the conclusion here. As you can see, there's no graphs along the way, just cuz it doesn't have that capability yet. And in a research like this, it's obviously a really useful one. Okay, so it created a prompt for the scatter plot, but it can't produce it as I said, but it says there's no clear linear correlation. UK and France have smaller average beds, but moderate divorce rates, while Italy show a low divorce rate, and Sweden, very wide beds, has a high divorce rate. The US with large beds around 160 cm, also has a relatively high divorce rate. Overall, countries with bigger beds do not consistently have higher or lower divorce rates. In short, other factors clearly overwhelm any minor influence that bed size might have. And then it does a discussion which I actually kind of like. Intimacy versus personal space, sleep quality. And it said, "Yeah, smaller bed can encourage more intimacy, sense of closeness. " So that's good. That's useful stuff. I think I kind of liked the conclusion of Google more and it seemed better sourced. I mean, if we look at all the sources used, you'll see this used 18 sources did not go to statistical site like Euroat. A lot of Wikipedia references, a lot of articles here, too. Whereas this research, there's also Wikipedia involved, but just seems more well-rounded. And obviously the creation of the scatter plot is amazing. Now here's the thing. You could follow up and say use the canvas to create the scatter plot. Oops. And I had deep research still enabled. And then it creates it for you. So this is the important thing to know. You can do this here and also in Gemini through the canvas feature. Oh, looks like there's a mistake. Okay, I need to copy the bug back in. It's going to edit the code and now it should work. It ran successfully, but I cannot actually display it in here. And I think leaving chat GPT kind of goes against the point here. So yeah, big ups to Gemini Deep Researches. Okay, and by the way, if you're enjoying comparisons like this, I do this on the channel all the time. So feel free to subscribe and we can do this more often. Okay, let's move on. I'll quickly review one more. And at

this point, I have to say our prompts are a little biased here because we're specifically looking for graphs and things that would be well visually represented, but it just opens up this new realm of possibilities, right? So if I look at this prompt that we wanted to review, I want to conduct research on the progression of LLM intelligence over the past 3 years using the AI plays Pokemon benchmark. Basically, if you haven't seen these, they use new models to play Pokémon. And then I said, I need you to build a simulation that I can interact with that will help me understand in depth. To all the follow-up questions, I just said any. And then let's start with chat GPT. This time it shows the last three years of progression. It creates a little timeline. I love that. 2024, the big breakthrough. I remember that. GPT03 was the first one that beat it. But yeah, I love this timeline. This is really good. And here to illustrate the progression, here's a comparative summary table. Okay. 2022, 23, 24, and then now in 2025, Poker Champ, huh, that's the newest one really. Llama 3. 18B. I guess this is by order of how they conducted them. I really was looking for Gemini 3 Pro as I saw a story about that this week of Logan from Gemini reporting on it actually beating Pokemon at a record time. So, apparently did not find that, which is okay. That table was a good visualization, but that's pretty much everything we get here. There's a summary here in the end. Now, let's compare it to Google's report. Immediately, this graph really supports this. Again, we're using questions that require graphs, but like this is so helpful now, seeing that, okay, Gemini 2. 5 Pro was actually up here versus the other models being down here. Like, this is the progression for Pokemon, right? The you have the first bosses and then you have the whole game up here. And then what else? Actually, looking at this, I even want to skip like all the text and look at the visuals here. There's a another banana generation of how it did it. This is great. this blueprint of how it's done. But then lastly, does it have a simulation? No simulation. It didn't build it. So, it could have used code, but then I guess it decided against it. Overall, this is just a more well-rounded report. But I have to say that also this one lacks the a mention of Gemini 3 pros results in this. And to be fair, after a little bit of research, I realized this was something that was shared by Logan on a tweet, but I don't think they published the results of the Pokemon benchmark and Gemini free doing it anywhere. So yeah, this tweet from two days ago, both of them didn't find it. So it's equal on that front. But I think on everything else, I do really like having these visuals as a part of the report. So obviously that's an upside. But I do have to fault Gemini Deep Research with visuals. That's on the ultra plan. On one thing, it didn't create a simulation that is interactive that can help me understand this. It didn't do that. It tried to do it. It didn't manage to do it. I think it misunderstood me. Maybe I could prompt around it. But either way, these visuals are really powerful. And

as you can clearly see, when you're looking for that now, you can get it with the Gemini Deep Research. So, OpenAI has another thing to catch up to. Honestly, deep research is one of the core AI things that I use regularly, and I have learned to love the Gemini version of it. And my conclusion here overall would be Gemini had better deep researches already with the addition of this. Gosh, there's no reason for me to be deep researching elsewhere. This is just that good. But again, this kind of assumes that you have a ultra plan on Gemini Deep Research and a paid plan on Chat GPT. If you don't, if you're on the free version, well, even in that case, I think the Gemini Deep Researches used to be better and they still are. At the end of the day, I would recommend running it in both Gemini and Chat GPT and comparing. But if I had to pick only one deep research, it would be Gemini every time at this moment in time. All right, I hope that was helpful. It certainly inspired me to run more of these deep researches. And with that being said, my name is Igor Pagani and I hope you have a wonderful