Introduction

We're going to have some fun, OK? This is going to be a bit more practical. We're going to go into my favorite tool.

But I only have 15 minutes, you know? And I'm used to talking forever and ever and ever. So we're going to keep it brief today.

But I promise you this. Well, I don't know the level of every single person in this room. That's impossible to know. I'll say it anyway, I'll promise you this.

At the end of these 15 minutes, you'll at least have an idea of what people who obsess over this stuff and spend, I don't know, six hours a day playing with AI tools just for the heck of it, AKA power users or educators, do with frontier tools that are available to consumers today, okay? Concretely, I'm talking about this little button here, Deep Research. You might have heard about it. It's behind a $200 paywall.

The Worth of the Tool

A little ridiculous, right? Probably not worth it to most people. But the first five minutes, or maybe three minutes, I want to spend talking about the worth of this tool. And then we're going to show you some demos. And finally, I'm going to show you one technique that we came up with yesterday with some friends. And it's one that tackles the biggest question relating to this tool.

Trustworthiness and Utilization

And that's the question of, can I trust it? What do I do with the site? Can I cite it? Or how could I cross-reference all the data? OK, so we're going to tackle that in the end. But let me just back out a little bit to get everybody on the same page.

Comparison with ChatGPT

What does this do as opposed to ChatGPT? I'll kind of assume that everybody in here has experienced ChatGPT. At this point, you went to an AI meetup. I think that's a fair assumption. But this tool came out, I think it's two, maybe three weeks ago by OpenAI. It's gated behind their pro plan, their $200 a month plan.

Now there's some copies, they're not nearly as good, so we can touch on them in the end. But basically, if ChatGPT just goes ahead and gives you an answer like this, like hey, here's a report. Then deep research actually spends 20, 30 minutes scouring the internet, looking for 20, 30, 40 links in some cases, reading all of that, and then taking all of that context, so much context, and running it through the most intelligent model in the world, which is O3, which is not accessible in any other way, shape, or form.

So we have things like ROC3 that released last week. I'm sure a lot of you follow all of this madness, frankly. But none of the models that are out there, not even the ones that you can access from the model switcher there at the top, are as good as O3. It's not even close. In terms of IQ, I think it's expressed that the best model right now is somewhere at 120, and O3 is somewhere at 140 IQ. But the thing is, just practically, speaking as a person who just obsesses over this stuff all the time, all my free time also, a lot of my free time, It just works better than anything else, OK?

So step one, it goes into the internet. It finds the 40 sources, relevant articles. It takes all of them in, and then it runs it through this like giga intelligent model that compiles a report for you.

Economical Value of Deep Research

We'll be looking at a few of these reports here in a second. But before we do that, I want to explore the, I'm so sorry. He told me not to walk in front of the projector. Look at me just like pacing back and forth. Anyway. I want to explore the economical value of this, because it's gated behind a $200 plan.

That's a lot of money. And you can only run 100 of these a month. So if you do the basic calculation, I suppose it's $2 per generation of a deep research. But then you also get JetGPT, you get Operator, you get Sora, you get all the other stuff with it. But let's just assume it's $2 per run. Is it worth it? The big question, OK? Well, it kind of depends on how much your time is worth, right?

If you're working a freelance job and you're making, let's say, $20 an hour. Let's just assume that. Or 20 euros, I'm sorry. I'm always in dollars. If you're making 20 euros an hour, then this thing would need to save you 1 tenth of an hour, right? So if it costs 2 euros to run one deep research, then saving 1 tenth of your hour would already be worth it. 1 tenth of an hour is 6 minutes. If this thing saves you six minutes, it's worth it. Let me just tell you, like I'll show you, okay? You can make up your own mind if it saves more than six minutes, but that's kind of the benchmark we're going for here, okay? I would argue that even for somebody like me who's quite quick at typing and researching the web and compiling it all and using models and stuff. Is this working?

Some of the deep researches that I have run and that I pretty much run like five to 15 deep researches a week independently of like my work, like teaching this stuff, I would argue that some of them have saved me like 10 to 15 hours, legitimately, I'll show you the examples. And some of them are just in the territory of I couldn't have even done that. And at the lower end, like they save an hour or two, okay? So six minutes? Honestly, kind of a joke. So I really think there's a value arbitrage here in the market.

There's obviously not many of those left in this world. I'm not going to go down the tangent. There's a whole story there. But the point is, in this world, it's kind of hard to create value. A lot of it is just exchanging time for money. But this thing is, I think, one of the spots. And the fact that it's hidden behind a $200 paid wall deters a lot of people. So that's also sort of an opportunity.

Examples and Use Cases

All right, so let's have a look at some deep researches that I ran over the past week. Some of them are personal. Some of them are medical. And some of them are actually from other people I know. OK, so let's start with the most basic one. And this is a big comparison table. OK, look at this.

This is what it looks like. So basically, I just was looking for a new monitor for my gaming setup, a very simple prompt, as you can see right there. We're not going to go into the prompting. I want to really give you one prompt in the end that's going to solve the citation and can I trust this problem. We're going to look at that in the end. But basically, it's just, hey, I want to compare all the gaming monitors out there. Can you help me do that?

Example: Gaming Monitors

I didn't know much about gaming monitors. I just knew that you kind of want a high refresh rate. But then it goes ahead. It asks some follow-up questions. I answered them, I'm like, okay, no budget, 24 inch, 1080, blah, blah, blah. And then it went ahead, fought for 11 minutes, compiled 28 sources, which you can see over here. So, actually most of these are from ACES, huh? That's a little biased, but then look, that's actually just the sites that it looked at.

So you have Reddit, you have Walmart, you have PCmonitors.info, newegg.com, whatever the heck that is, msi.com. How long would it take you to read all these pages? I don't know, you can answer, I'm just kind of putting that out there, you can answer it for yourself. So all of these sources, okay? And then it fought over it for 11 minutes, and it came up with, let's be real, even for the tech-savvy people, how long would it take you to create this? with internet access.

You get a laptop, you get Wi-Fi, and go. 1Imagine somebody tells you, create me an overview of all the gaming monitors out there. I want to pick the best one. There's not a person on Earth who would do this in under six minutes. I think we can agree on that. And I would argue, with all this intricate detail, I mean, look at all these columns out here. I know I'm scrolling fast, but look at all these columns. We're not going to look into the details, but response time, panel type, resolution, screen size, connectivity, adaptivity.

adaptive sync tech, I don't even, I don't even, okay, I guess I've heard of those, but like color accuracy, viewing angles, brightness, I would not even think of including all of these. And I haven't. And I didn't know, you saw my prompt, it was just like, I want a gaming monitor, can you help me? And it's like, yeah, sure, these are the things to consider. These are all the different monitors, and I filled it out for you. And by the way, if that's too much, here's the recommendation of the overall best one. It actually arrived two days ago. I'm playing on it already. It's kind of nice.

So that's pretty amazing, right? And that's the most basic use case I have here. It's just researching stuff across the internet and compiling what the humanity put out there on various blogs and sites. Nice. Let's look at another one.

Example: Learning Guides

Learning stuff, okay? This is also... Okay, I gave it detailed info on my target audience, because for my business, I defined the different target audiences we have, so this is detailed. But ultimately, this is a prompt I want you to take away today, okay? And you'll memorize this one. Ultimate in-depth guide. That's all you need to know.

Ultimate in-depth guide. Because the word ultimate, it's just a very powerful word from prompt engineering already. Because ultimate means the very last, the best, the best. Final guide right so in other words create me a really big guide that covers everything I need to know On what topic how to learn generative AI? Okay, I'm sure we'll be back in a second here. I'll give some examples and then I said progress strategies curriculum beginner tips resources and Yep, it did its thing, 26 sources, eight minutes, and this would be kind of hard to go to, so I'll just scroll through it. First of all, I'll scroll quickly just so you understand what an ultimate in-depth guide looks like. Yep, we're not even halfway through, let's keep going.

This is an ultimate in-depth guide. Let's just kind of stop somewhere, anywhere. Hands-on reflection. Set up info streams. Subscribe to two to three AI newsletters or blogs, examples, using multiple AI newsletters. This is how you learn. If you go through this, how long would this take a teacher? Again, more than six minutes probably, right? So it's a real-time arbitrage if you can find ways to use this.

But again, these are kind of just templates, right? First one is research all of the internet. Create a big table for me and then make recommendations. Second one, create the ultimate in-depth guide on you name it. You insert the thing from your life that bugs you, the thing that you're not able to figure out on your own. So yeah, this keeps happening. One sec. So yeah, you can do this pretty much with everything. I did this with specific topics. I did this with super broad topics like this. Oh, OK.

I'll do my best. So yeah, this is pretty amazing. Let's look at another one, because how much more time we got? Like 10 minutes? That's nice. OK, here's another one. This one is a bit more personal. And I didn't show it off on my YouTube channel, et cetera.

Example: Ancestral Research

But this one is ancestral research. And it was one of the first ones I ran. And we also have a team account. We actually ran for two of these accounts, like 200 requests in three days. And this was one of the first ones that we kind of ran. It's a bit more personal, but I don't mind sharing it. It has the names of my parents. that's fine but basically it looks at my ancestry which I'm originally from Slovakia and my grandparents are from the Czech Republic and Hungary but I grew up in Austria whatever it's a bit of a mess all over Europe but the point is our ancestors are all over the world and I didn't even know now look at some of these sources I think this is the most impressive thing to me because if you look at these sources you'll find that it was looking into like the the church registers in the 18th century or something? I didn't even know that's on the internet. I would never Google that, right?

This does it. Slovakia genealogy, however that is pronounced. All these sources, ancestry.com, passenger. This one was interesting. Passengers who crossed over from Europe to New York between 1820 and 1957. Apparently every single one of them has been documented. Deep research found the documents, found the matches of my surname and was like, hey, There might be somebody from 130 years ago that just went to New York and now lives in Ohio. We found their data. It's all in this report. It's kind of wild, honestly. Look at that. So there's a bunch here. Tracing key relatives from my mom's side. Family location. The evidence we have shows Juraj and Maria residing in Bratislava. It's possible they were both from Bratislava.

And then it goes into kind of their past. And they found families with the same name that were in a different region. Earlier generations were probably this was the dad and this was the mother, just hypothetically, and then it tells you where they were in the 1910s, 20s, and then it goes all the way back to Trencin in 1243 to 1329, because it found this reference, whatever that is, Forum Historia SK, some Slovak register, with people that have the same name, I think you get the point. This thing is wild. It looks at all these sources that I wouldn't even consider, and it does the thinking too. That's the hard part. That's what takes the energy. That's what takes effort. Thinking is hard. I think everybody would agree. This kind of does it for you, and then you can go from there. I'm not saying it does it all for you, but it does way more than ChatGPT was doing.

If ChatGPT, like good prompts, many examples, many exceptions to this, but good prompts generally save you like... minutes, tens of minutes. I think with coding, that's kind of an exception. There, they can save you literally hours. But generally speaking, you're saving minutes, dozens of minutes. With this, I think we're always talking hours. And that's the big difference here, because not just that it found the sources, but it also put it all together for me. And it told me, yeah, where my family's from. Interesting.

Example: Child Car Seat

Here's another one. I'll quickly go over this one, because this is from our community. This is basically a lady that was looking for an amazing kinder seat. Wait, hold up. What is that? Like children's car seat. Is that right? Any German speakers in here? Children's car seat.

OK, she was looking for the very best one. And obviously, she cares. She wants her kid to be safe. Now, she included all of these details here. She included the details on her car, even the country. And she has sources that she trusts, and she wanted this to look at. Now, this is obviously a thing that a mom could spend an entire day on, because what an important decision this is. Well, a few follow-up questions later, safest car seat for a five-year-old, that is 17 kilograms. And then it doesn't come back with a big table that is overwhelming, because in the previous example I asked for that. It just comes back with, hey, this is probably the one. This is the number one seat you want, and here's the reasoning and all the sources behind it, right? It's not just random info.

It comes from this website, comes from this website, comes from this website. It goes on, gives you more, but I love that it just starts with the number one recommendation, because that's what a smart person would probably do if they were to create a report like this for you. They wouldn't be like, oh, I found 50 seats. Let me tell you about every single one of them. No, they would show up and be like, this one is probably the one you want, but I have some alternatives you want. That's what it does. Kind of nice. Yeah, so it goes into all the detail that she wanted here. She was really impressed with this one too.

Okay, I have two more that I want to show you. One of them is medical. Something that I've been dealing with. And the second one is just a ranking. And I just wanted to show you this because it's kind of a... These are very simple prompts. Five minutes? Perfect. Somebody was asking is actually get it to research every movie award out there and then come up with the ultimate criteria for movies and rank top 1 to 100 movies of all time. Like, honestly, let's just think for a second. How long would it take you to create the ultimate movie ranking list? Ranking 100 movies. Like, first of all, just coming up with 100 movies.

Even if you're a movie buff, like, that takes a bit of time. And then you kind of have to think through it, right? Is this one better? Is this one better? Like, how do I even start? That first draft is the bulk of the work. The fun part is, like, re-ranking it and being like, no, no, no, I think, like, number four would actually be number two because of this reason. This is the hard part.

I think you know what you're about to see, and if you look at this little bar on the top right, like how long this page is, you can kind of like project out that, yeah, this is the ultimate movie ranking list, because first it talks about all the awards, but then here, you legitimately get like top 100 movies of all time, as reported by ChatGPTO Free Deep Research, and it does the reasoning with it too, right? So it's not just like, hey, it's Citizen Kane, but it's Citizen Kane because... um this masterpiece is often decided being cited as the greatest film ever made uh it's topped the site and sound critics poll for 50 straight years and it's notable for these reasons one academy award for best original screenplay and influenced generations of filmmakers filmmakers okay so pretty good reasoning i would say you know i mean sure you can argue with that but it's something it's some ground to stand on And then the list goes on. And as you can imagine, yeah, there's 100 different movies all ranked. Look at that, Matrix, 37. What do we have at spot 60? Saving Private Ryan, Ben-Hur. Yeah, there you go, a massive list. And then I ran this prompt on so many different things at this point. You can rank everything.

One that I was trying for the YouTube because I wanted to click bait people with the title. I'll be honest, always honest. I was like, how can you make money online with no money and no skills? Give me a list of the 100 best things that I can do. And it did. It created a list like this. And I ended up not titling the video like that because I was like, I'm not going to sink to that level maybe one day. I started the video of it because it was really interesting. And the point here is it can do it for everything. You can rank everything, 1 to 100. You can create these big tables for any topic. You can pick between different products. You can do ancestral research, research on other things. And ultimately, I think this is the one that kind of is just hard to argue with.

Example: Medical Research

It's medical research. So if you're dealing with something specific, chances are, if you're in this room, you're into tech, and you probably didn't study for seven plus years as a doctor.

Maybe there's somebody. But generally speaking, most of us don't have medical expertise. With this, you have something that you can bring to the conversation with a doctor.

And I think that's the amazing thing, because usually you walk into that room and you're like, well, this hurts and then they like talk in Latin to you and you're like okay give me the pills and that's it right here you can run a little deep research so for me personally just super short story time I'm into kite surfing it's a part of the reason why I moved to Lisbon it's basically like surfing but you have a like a huge kite in the sky, and you fly while it's windy.

The problem with kite surfing is it really gets into your forearms. And especially if you get into jumping, you tend to get tennis elbow, a.k.a. tendinitis.

So it's not the muscles, but the tendons in your arms that kind of have problems. So I didn't know anything about it. And I was already researching this manually.

I was reading the articles on BetterHelp or MedDoctor or what all those sites are. This was the best thing I've got on tendinitis from like six months of dealing with it.

So basically, again, very simple prompt here. I have tendinitis in my forearms from kite surfing and looping the kite too much. Create the ultimate how-to guide.

Again, ultimate how-to guide. I want you to remember that. That's a really good one.

On how to recover my arms, including exercises, a recovery plan, and what to avoid. So a few follow-up questions later.

Very simple. Just the recovery plan in here, the thought that went into it, it's just incredible.

So it tells you all about it, but then these different phases, and once you start want to start reading these different paragraphs. This is really in-depth.

If you read this, you can enter a room with a doctor and actually hold a conversation and add to it. You can make propositions and then he can bounce off your ideas. Again, the model is doing the thinking for him.

So maybe he's having a hard day. And maybe on a better day, he would invest It's hard to quantify energy, right? But maybe he would invest more than two, three minutes of helping you out in between different patients.

Maybe he would invest 10 minutes. But maybe he's having a hard day and you're just unlucky. So what? You're going to have more of a medical problem just because he had a hard day? That's just the reality of life. That's how it goes sometimes.

With this, you can do the thinking for him. And you can show up with the information and be like, hey, I was thinking I wanted to do these exercises. Any other recommendations?

I think this is absolutely amazing. And it has been a real game changer for my life.

Additional Insights & Stories

I want to give you two final things as we wrap up and move into Q&A.

One of them is I've been talking to a lawyer in the US yesterday. It was a confidential conversation. But he basically shared this with me.

He's like, Igor. I've been using deep research, and I've been using Notebook LM. That's a free tool from Google where you can just throw 30 sources into it.

So what he does is he uses deep research to gather a bunch of info. He throws that into Notebook LM, and then he throws all the other Quartz documents into Notebook LM.

And he told me, Igor, this is not fair. I don't lose anymore. Like it's just like I'm AI enhanced, I can take more cases and I show up on topics that I've never heard about, more prepared than some of the experts that they invite there.

And not just that he can hold conversations, but he throws it into Notebook LM. That thing generates podcasts that simulate a conversation that could be happening. And he prompts Notebook LM to be the prosecutor opposing him.

And he tells it to be ruthless. So he's sitting at home, listening to a podcast with his feet up of the prosecutor already attacking him. And he doesn't even have to do the thinking, because the other voice in the podcast does the counterattacking.

So he goes through an entire court case in advance, in his little garden he listens to the entire discussion informed by multiple reports like this informed by all the data that the court shared and then he walks into the course without notes without preparation and he like wrecks them because he's been there he's done that with the power of ai and then there's people out there who are like oh you know like

AI tools, like, you know, like, ugh, not worth it, like, don't bother, bro. Or, like, prompt engineering, ugh, who needs that, you know? And then this guy's out there, like, his career is literally skyrocketing.

And he was just like, it sometimes just doesn't feel fair, but, like, I'm so thankful for these tools and just putting them to work.

So that's just one example. And I hope you will find your own use cases, because all of this is super specific. The power lies in the specificity.

And you got to do the work of applying some of these templates and use cases that I showed you to your very own context, to your very own problems.

Conclusion

And finally, I want to show you one more thing, because there's obviously a huge question here, which is like, OK, can I trust this info? And fair enough. Generally speaking, I would say,

This has gotten way more reliable than just ChatGPT outputs have ever been. So if you made your experience there, throw that out of the window and rediscover this if you can afford it, if it makes sense. If not, there's free alternatives, like Perplexity or Grog Free now, that quickly.

And the prompt that I wanted to show you is this one. Maybe we can send it per email or something. This has been actually shared with me yesterday.

It's a bit hard to read, but it basically says, hey, the entire deep research that you just created, can you create a detailed table listing all of the citations? And the table should include a comprehensive set of columns that not only capture all the details, but more and more. And basically, you get this table which allows you to cross-reference every fact in the report really effectively.

So as you can see, here's the summary of the fact. And then here is the direct source that it comes from, the link. the date that it's from, what kind of source type is this.

And if you export this into something like Excel or Notion, you could filter, like only give me the academic studies, only give me the review articles, whatever it might be. And this is the most efficient way to double check the sources if your life depends on it. So you can even do that.

So like, I don't know, to me that is sort of like, that shatters every question some might have or like every, how to put it, every concern with this tag. Because yeah, if you do a little extra work, you can double check the stuff.

And you can save five hours, 10 hours of work. Then you spend like 15 minutes double checking it. It's way easier than thinking to yourself, no?

So there you go. That's my little demo for today. If you don't want to use this deep research, I recommend Proplexity.

Hallucinates a bit more the reports are like one-tenth of the size, but it's free And then eventually they said within a month or two the deep research thing is gonna be an all paid plan So if you have a $20 chat shifty plan, you're gonna have it too. I hope you make your life better with it That's all I got for today