Full Tutorial: GPT-5 vs Claude 4.5 vs Gemini 2.5 for 10 Tasks (Oct 2025)

Hey everyone. All right. So today I want to share an updated guide on how I use chat GPT Claude and Gemini. I will also cover nine AI projects that I rely on daily and how I use deep research to make important decisions. So if you want to learn how I actually use AI to get stuff done, this video is for you. All right, so let's get started. Okay, so here's my table showing what capabilities each AI model has. I'm going to walk through each use case and then pick my favorite. All right, let's start with everyday answers. For everyday answers, GPT5 is fast and concise, which makes ChatGpt a great choice. Claw is pretty good, too, but I keep running into rate limits on the $20 a month plan. So, my choice here is Chat GPT. Now, let's talk about writing and editing here. I think the winner is still Claude. Let me go ahead and mark Claude as the winner. And that's because I think Claude is the best writer and editor if you give it examples of your best writing. You know, Chad Chbt likes bullet points a little bit too much and generalized writing is way too verbose. Claw is my go-to here. And this is what I use AI for 80% of the time.

Okay, so next for coding, it is a lot more competitive. And actually, I'm gonna mark two winners here. So, let's make two stars and let's get rid of this. And let's actually move chat GBT as one of the winners and claude as the other winner. And you know personally I prefer using cloud code but I have many engineering friends who swear by opening codeex with GPD5 high. Codeex is a little bit slower than cloud but can often fix gnarly bugs in just a few lines of code. And overall it's pretty remarkable how fast OpenAI has caught up to anthropic in this area. All right. Next, let's talk about deep research. For this one, my preference is Claude. And I think the main difference here is that Claude is only deep research tool that actually writes maybe four or five pages of report that you actually want to read instead of chat GBT and Gemini that go off and create 30 and 40 pages. I'll cover deep research a lot more later in this video. Next, let's talk about web search. So, web search, I would say uh I'm going to give the winner actually to Gemini. And I think it's pretty close here between chat GPT and Gemini, but I'm giving to Gemini because the AI mode is just a lot faster than waiting for the chat to generate in chat GPT. And when I'm looking for web searches, like looking for nearby restaurants or something quick, speed is the most important thing. Okay, so I'm going to cover the rest pretty fast. For voice chat, I think I still prefer chat GPT. I think it has the most personal and the most authentic voices. Now, that being said, a lot of my AI voice dictation is actually using other tools like Whisper Flow because I find that, you know, all three models tend to interrupt me a little bit too much when I'm trying to talk. So, yeah. So, I prefer to use other apps for AI voice. For image generation, I think uh Gemini has the best image model right now. So, let's put it here. Gemini's Nano Banana is a incredible image model. It also generates I think a little bit faster than ChatBT. And uh I have a full tutorial on how you can build nano banana apps that you can check out in the description. Okay. Now, for video generation, this is actually a pretty tight race, but I'm have to give the winner to Chad GPT. And uh in my opinion, VO 3. 1 from Gemini, which was most recently released, is actually arguably a better model, but I just really enjoy creating videos of my family and friends using Sora. Chat GPT, OpenAI, they hired a ton of people from Meta, so they definitely get social a lot better than Gemini and Google. And that's why I think Sora is still my favorite choice. All right, for live camera, to be honest with you, I haven't used it too much, so I'm not going to give it a winner. And for computer use, I have tried computer use across all three different providers. In my opinion, I think computer use kind of sucks. Like, you're basically watching the AI click around a mouse and take screenshots and it just takes forever. And yeah, I just feel like I haven't really found the right use case for it yet. you could just leave it alone and you know ask it to help you book a restaurant or buy some flowers or something but I also don't trust AI enough to do that. So for me personally I don't think AI is really good for computer use right now. All right and

let's quickly talk about something else. So OpenAI also recently launched a new web browser in Atlas. So let me actually open it up and this is actually pretty amazing. OpenAI seems to be taking a path of launching MVP product and using their massive user base to iterate with users as quickly as possible. But you know just using Atlas I think um you know let's say I search for uh my favorite good restaurants nearby and it just basically calls chat GPT and the time it takes for this thing to generate the chat it just a little bit longer than I want. Right. So, it took like what, one or two seconds to generate. But if I go back to regular Chrome and search favorite restaurants nearby, you know, it's like near instant. And of course, it comes up with much better results. These are actually near where I live versus catchy looking up places like San Jose. So, I think there's still a uh a bunch of room for improvement in Atlas, but I'm sure OpenAI is going to improve it quickly. Now, you may see here that I have Grock, right? I have Grock here and I haven't given a single award to Grock and that is because going to next slide. I just feel like Grock may be a great model but honestly I cannot take the product seriously because it markets AI companions so aggressively. These AI companions, some of them are waifuss, weird characters. I don't know. I I just feel like this kind of stuff doesn't actually appeal to the regular user, regular consumer. And personally, I think XAI isn't making a big branding mistake by pushing these companions so hard. And I think they need to hire some real marketing people to the company. Okay, so there you have it. This is kind of my takeaway on my favorite models. And again, I have a post in the description where I've linked this table, but let's keep going, right? Let's talk about some something else. Okay, so let's talk about my go-to AI

feature for all three models, which is projects. And you know, maybe I'm old-fashioned, but I think projects is far more useful of a feature than AI agents, AI browsers, or whatever is hot in AI right now. And all three big labs have this feature, but I prefer clawed projects because it lets me paste in simple text files, lets me attach Google Docs, and it even has memory. So, let me kind of show you how things works. Okay, so first let me talk about kind of three quick tips that you might not know about to get the most out of projects. So, tip number one is to attach Google Docs directly. This ensures that your project is always aware of the latest updates in your Google Docs and other files. I also like to add plain text files and deep research outputs as context. Okay, tip number two is to include your output style in the prompt. So, this ensures that whatever the AI spits out will always be in the style and format that you want. And for me, my preferred style is to use short bold stems followed by two or three short sentences so that the entire output is easy to skim. All right. And last but not least, and this is more of a pro tip, is to get AI to iterate on your prompt. So let's imagine you have a long conversation with AI in your project and you finally get to the output that you want. You should follow by asking, can you update the project prompt to produce this output in one shot the next time? And then you should save the updated prompt so that you can get to the desired output faster the next time you talk to the project. So let me kind of make this really practical by walking through a real example. Right? So as a product manager, let's be honest, we spend a lot of time editing documents, editing strategy documents, editing PRDs and so on. And here's an example from Clarvo on just how much internal iteration it takes to produce a typical strategy document. Right? First you have to think really hard. You got to talk to customers. Then you got to get feedback from stakeholders. You get a flood of Google comments on the right side. Then you got like go back and iterate and think hard again. And then you do another round of feedback. And then you know this whole process can take probably a couple of weeks. And you know you spend your day just kind of editing these Google documents, right? So here's how a project can help you with this process. Okay. So I'm going to use a kind of like hypothetical example. So let's say we are uh working on boat which is an AI coding startup and AI coding is an incredibly competitive space and we want to create a strategy document for boat. Okay. So first I'm going to create a project for boat. So I have a project right here and uh I've uploaded three different files. So the first file I've uploaded is the working backwards template from Amazon. I like this template because it starts with the customer and it kind of like expresses the strategy from the customer's point of view. But the point here is you want to upload some sort of template or some sort of example of what a good output is. All right, the next document is I actually did some deep research on both strategy and here's the deep research document, right? So there's a pretty long document. I think I use Gemini for this and then I'm just going to attach this document to the project too. And last but not least, I have my own document on uh how I think each AI coding tool will play out. Uh right, there's a bunch of stuff here and doesn't have to be like this. It can just be like your voice dictated notes and your overall thoughts on the strategy. But basically, it's like a very rough first draft of the strategy that you just kind of spit out in your head. Now that I have all three of these documents, right? So now I can actually ask AI to go ahead and draft the PRD. And before I do that, let me just kind of show you the prompt. Uh the prompt is basically talking and giving some context about three different documents that I've uploaded to the project and then asking it for help to synthesize these materials into a compelling strategy. Let's look at uh one of our past conversations here. Right? So let's go up here and uh basically I asked it to create Amazon style PR FAQ document for boat. I want you to be my thought partner and figure this out. And first I asked it to try to draft some answers to Amazon's working backwards questions. And that's exactly what it did. Who is the customer? And so on so forth. And then I asked to make the PR FAQ and it drafted it right here. Right. And this is pretty good. And then I gave it some feedback. And I can use like a voice dictation tool or something to give it feedback. But basically my feedback is like personally I think where the money is in the enterprise space and I think uh you know Bo needs to figure out how to compete with Figma make and some of these other tools. So given all this can you come up with some more strategies and it did and you know we can keep going back and forth like this. So you know this is the key behavior change that I want you to remember today right? So basically create projects for any kind of important work or thing that you're working on and then give AI the right context and then use voice dictation use your feedback to work with AI as your thought partner to refine the output. You know you can go through many iterations with AI to make your document much better before you start sharing with stakeholders and other customers. Okay, it just makes the iteration loop for editing documents much faster. All right. So, here are my top three AI projects for product creator and personal work. And I already showed you how I create strategy documents and PRDs. So, let's talk about this show notes project that I have. So, this is a project that I use after doing a podcast interview to create a bunch of other artifacts. So basically all I have to do is pasting a raw transcript from the interview and then this project turns it into a list of top quotes, moments to cut, YouTube titles and thumbnails and even social posts. And honestly, this project alone probably saves me 3 to five hours a week. And is pretty involved here. You see this? So the prompt is very long. Uh it has a bunch of stuff about top quotes that you want to pull out, top moments to cut, right? It has a bunch of thumbnail and title output formats using best practices. It has a bunch of social post examples down here. It is a very long prompt and uh you can find this prompt in the description where I've linked my newsletter post. Uh however, uh I'd have to say it is only available to pay subscribers of my newsletter post, but it it's worth it. If you're making a podcast, it's worth it to get this prompt. But anyway, let me show you how this prompt works. So, let's take one of these files. So, let's take an interview that I did with Zapier CEO Wade recently. And here's kind of like the full raw transcript, right? And based on this raw transcript, what has done is created the top quotes from our intro reel. It has created the moments to cut where we had some audio issues or repetitive chatter. And uh most importantly is created a bunch of thumbnail patterns from myself and uh other YouTubers that I respect. A thumbnail and title copy which you know normally would take me a very long time to come up with done that. Uh if I keep scrolling down it's created show notes with the right timestamps and it's created takeaways from the interview. It has even created the social preview and social posts. Now this stuff isn't perfect in one shot, right? So the next prompt that I asked it is, let's work on social post. Can you make it more interesting and engaging and here's some feedback? But the point is like it created like a really solid first draft for all of these artifacts which is honestly just saves me a ton of time and then I can iterate with this project to make these assets actually good. Okay, so that is the shown project and now let's talk about the family trip project. So this one I actually created in uh Chat GPT just to show you that you can create projects in Chat GPT and Gemini 2. And here the project is basically a bunch of instructions for things that me and my family likes. Right? So for example, we want to create memorable family experiences. Here are some favorite places we've been to in the past. Montter Bay Aquarium. So on so forth. And there's different types of trips that I want you to plan out. A half day trip, a day trip, so on and so forth. and also kind of when we are available. So for example, my daughter has taekwond do practice Saturday morning. So we're actually not available then. So we're available Saturday afternoon and on Sunday and use this project, I can basically ask it, you know, what kind of familyfriendly events this weekend, right? And just ask it that. And because it has all that context, it can search everything and it can find a bunch of really cool family-friendly events that are topical. not this headless forcement might be a little bit too scary for my kids but yeah it can find all these awesome events because it has a context okay so that is on projects and if you want a deep dive on all nine projects again check out the newsletter post that I've linked in the video description for more right I got one more topic which is deep

research now deep research is the real extended thinking mode for AI and it is honestly useful for much more than just market research you should think about using it whenever you need to make a difficult decision, okay? Or try to evaluate two different trade-offs. So, uh, every provider now offers deep research. So, I tested each model with this prompt, right? Research the AI coding market and share both strategy to compete. And here's kind of like a sampling of the results. And the point I want to make is that Claude produced a seven-page report. You see this here? It's like a pretty short report. and pretty concise while both chat GBT and Gemini produced a much longer report like this is like a lot to read. So because of this again I think I prefer Claus deep research because it's actually able to synthesize information to something concise instead of just dumping a bunch of information on you. But honestly like all three deep research are totally usable and they're also pretty good. All right. So now, how do you get the most out of deep research? Most people just type a single question into deep research and then they wonder why the output sucks. So here's five steps that you should follow instead to get much better out. So first start with a clear goal. Tell it the specific decision you're trying to make or the problem that you're trying to solve. And then very importantly, share relevant context. And this is probably the biggest lever. So share a full page of context including any documentation or any kind of voice dictated notes that you have. The more context you provide up front, the better the research output becomes and the more personalized it becomes. Okay. Number three is specify the sources it should rely on. So for example, if you're trying to decide what kind of product to purchase, ask it to research both professional reviews and user reviews on websites like Reddit, right? point it to the right sources that you want and always ask it to ask you questions first before it starts his research. So just include in your initial prompt ask me questions before you begin your research and this way the AI can clarify things before it gets started. And finally um deep research it naturally integrates with projects in cloud you can easily one tap to add an entire research report to the project. All right so let's make this super practical. Let me show you examples of how I use deep research. So, here's the first example. So, I'm learning how to play the piano and I want to find a great app to help me read sheet music and practice the piano. And note how detailed my prompt here is, right? I told it that I have a kauawaii piano. I'm advanced beginner. I've been playing my favorite songs using YouTube videos that have floating notes. and I want to learn how to read sheet music more and learn proper fingering and pedal technique. So, give me the right app and look up uh recent reviews and ask me any questions before you get started. And therefore, it asked me a bunch of questions about my budget and so on and so forth. Okay, so based on this, it did a bunch of research and here is the research report. Now, I'm only aware of certain piano apps like Simply Piano, but the recommendation that I came up with, I actually had no idea about. The recommendation is called Playground SE sessions. And Playground Sessions is an app that has audio feedback, is a lot more professional. It teaches the right lessons more than simply piano. And uh it just has a bunch of uh really good content here. and it's kind of explained how it actually meets my requirements because I've given it all the personalized context up front. So that is a playground sessions and it has a bunch of runner ups and you know it's pulled together this like pretty awesome report right this is like something that maybe someone who is an expert in piano apps would give you. So again the more personalized you can be in your context the better the deep research is. Okay let's keep going. Let's cover another quick example. I'm going to skip the creator business one. Uh you can read about it in my newsletter post. Let's talk about the family trip example, right? So here's uh some deep research that I did for a family trip. So I'm looking for a place to go in December with my parents and my kids. So it's going to be a multi-generational family trip. And in this project is what I showed you before, which has a bunch of personal context on what kind of trips my kids like, what kind of places we like to go in the past, and so on and so forth. So, because it has all this context, I'm asking it, "What are some great places to go that are under $10,000 total trip expense? " Okay, so it's asking me a bunch of questions and I answered them and then it went off and did deep research and the top recommendations for December trips are New Orleans, Mexico's Riviera, Maya, Costa Rica, and some other places. Okay. So again, we just asked in a random chat window to do deep research on family trips. It might not have come up with these answers, right? It came up with these answers because I have a project that has a bunch of personal context on what kind of preferences that I like. And then I can go off and go back and forth with this chat to figure out exactly where we're going. And uh it turns out that we are going to Mexico based on my chat with uh Claude. Okay

there you have it. We covered a lot in this video and I want to make it as practical as possible. So let's do a quick recap. So number one, use different models for different use cases. Use chat GPT for everyday questions. Use claude if you spend a lot of time writing and editing. And if you're already into AI coding, either use claw code or open AI codeex. Gemini I think is great for images, any kind of multimodal stuff and I think Gemini 3 is coming out fairly soon too. All right, so that's uh number one. Number two, use projects for any kind of major work stream that you're doing because projects let you upload relevant context so you don't have to do it each time you chat with AI and it's just incredibly useful. So I use it for uh PM work, creative personal stuff, too. And number three is uh use deep research whenever you need to make an important decision. Again, give it a bunch of personal context. Ideally, all this context is already available in your project. Get it to ask you a few questions before it gets started. And uh once you get the deep research report, add it to your project to give your project even more context. If you enjoy this video, please like and subscribe to this channel and I will make more practical AI tutorials for busy people exactly like this. All right, until next time.