Everyday AI Podcast – An AI and ChatGPT Podcast - EP 607: How To Use Gems, GPTs & Projects for Real Business Results

Episode Date: September 10, 2025

You're probably opening a fresh chat in ChatGPT, Claude, or Gemini—and that might be sabotaging your productivity.Randomly starting conversations with no plan wastes time, fragments context, and lowers the quality of what you get from LLMs.Learn the essentials of Gemini Gem, GPTs, and Projects—how they differ, when to use each, and how to structure prompts and workflows so AI actually speeds you up instead of slowing you down.Newsletter: Sign up for our free daily newsletterMore on this Episode: Episode PageJoin the discussion: Thoughts on this? Join the convo and connect with other AI leaders on LinkedIn.Upcoming Episodes: Check out the upcoming Everyday AI Livestream lineupWebsite: YourEverydayAI.comEmail The Show: [e-mail] with Jordan on LinkedInTopics Covered in This Episode:Harnessing Custom GPTs for EfficiencyGoogle Gems vs. Custom GPTs ReviewChatGPT Projects: Features & UpdatesClaude Projects Integration & BenefitsEffective AI Chatbot Usage TechniquesLeveraging AI for Business GrowthDeep Research in ChatGPT ProjectsGoogle Apps Integration in GemsTimestamps:00:00 AI Chatbot Efficiency Tips04:12 "Putting AI to Work Wednesdays"08:39 "Optimizing ChatGPT Usage"11:28 Similar Functions, Different Categories15:41 Beyond Basic Folder Structures16:25 ChatGPT Project Update22:01 Email Archive and Albacross Software24:34 Optimize AI with Contextual Data27:49 "Improving Process Through Meta Analysis"30:53 Data File Access Issue33:27 File Handling Bug in New GPT36:12 Continuous Improvement Encouragement41:16 AI Selection Tool Website43:34 Google Ecosystem AI Assistant45:46 "Optimize AI Usage for Projects"Keywords:Custom GPTs, Google's gems, Claude's projects, OpenAI ChatGPT, AI chatbots, Large Language Models, AI systems, Google Workspace, productivity tools, GPT-3.5, GPT-4, AI updates, API actions, reasoning models, ChatGPT projects, AI assistant, file uploads, project management, AI integrations, Google Calendar, Gmail, Google Drive, context window, AI usage, AI-powered insights, Gemini 2.5 pro, Claude Opus, Claude Sonnet, AI consultation, ChatGPT Canvas, Claude artifacts, generative AI, AI strategy partner, AI brainstorming partner.Send Everyday AI and Jordan a text message. (We can't reply back unless you leave contact info) Start Here ▶️Not sure where to start when it comes to AI? Start with our Start Here Series. You can listen to the first drop -- Episode 691 -- or get free access to our Inner Cricle community and all episodes: StartHereSeries.com Also, here's a link to the entire series on a Spotify playlist. 

This is the Everyday AI Show, the everyday podcast where we simplify AI and bring its power to your fingertips. Listen daily for practical advice to boost your career, business, and everyday life. Meet Firefly AI Assistant, now live in Adobe Firefly, the All In One Creative AI Studio. Just describe what you want to create and the assistant handles the rest, orchestrating multi-step workflows across Photoshop, Premiere Express, and more in one conversational interface. You direct the outcome. The assistant accelerates execution. You probably think that you're saving yourself a lot of time.

Whenever you go into your favorite AI chat button, you click that new chat button, and then you spend some time getting it to respond just how you want it, sharing your files, and you get an output that you're happy with. You saved yourself time, right? Maybe. Or maybe you just wasted a ton of time by not knowing how to use things like Google, GEMS, OpenAI, chat GPTs, GPTs, and Claude's projects and Open AI's projects. You see, these things have changed so much in the last week or two, whether you're a new person

learning the basics of these AI chatbots or you were kind of a seasoned vet using them for hours a day. There's been some under-the-radar changes to these projects, GPs, and gems that I think change how we should be using these AI chatbots. Because if you're just going in there every single day in clicking that new chat button, you're actually probably not saving yourself very much time. All right,

we're going to be talking about that today and a lot more on everyday AI. What's going on, y'all? My name's Jordan Wilson and welcome to Everyday AI. This is your daily live stream podcast and free dealing newsletter, helping us all not just learn AI, but how we can leverage it to grow our companies and our careers. If that's what you're trying to do, you're in the right place. It starts here with the unedited, unscripted, live stream, and podcast.

But where you actually can leverage what you are learning is on our website at your everyday AI.com. So there, yes, in our free daily newsletter, you can sign up for it. We keep you up to date with everything you need to know that's happening in the world of AI. But maybe more importantly, we pull out the most important insights or things that we didn't even have time to get to from today's episode. So if this is helpful, you're really going to want to go not to be.

just sign up for today's newsletter, but read it. All right, because that's how we actually can become the smartest people in AI, in our companies or in our departments. All right. So if you're here for the AI news, just go make sure to go check that out in today's newsletter. But I really want to talk about these recent changes to custom GPs from OpenAI, some updates to OpenAI's projects feature, Google Gems and Claude's projects. Because if you are just going to click that new chat every single time you open your favorite AI chatbot, probably not saving as much time as you think. You may actually be wasting a lot of time.

This is difficult. Let's be honest, because to learn a new skill, it takes a lot of practice and repetition. The problem is, these AI systems that we use so much, and some of us have become reliant on almost for, you know, our day-to-day work in business, they change too quickly to actually establish those good routines and those good habits. So unless you're spending hours a day like me to understand these things, you're probably actually maybe wasting some time. Is using an AI chatbot in this way faster than doing things manually?

Absolutely. But the bar is continuing to rise. So you and your skill set, you have to do the same. You can't become complacent just because you're using AI, right? You have to always stay up to date and take advantage of the biggest updates. And that's why we've started this new segment calling putting AI to work on Wednesdays, right? So hopefully you like this. These shows have actually, even though this is only our third, I think it takes a lot of planning.

So if you want to see more of this, let me know. And this is what this ultimately has stemmed from. A lot of people are saying like, hey, Jordan, I listen to your show all the time. But, you know, I get maybe once a week or, you know, once every other week, I get some important insights. And people ultimately are still asking, like, how are you using AI? So that's what I'm doing. I'm actually giving you all a little look into, you know, everyday AI, how we operate, things we're working on. Our stats, right?

I'm literally just, you know, pulling the cover off here and allowing you all to look underneath, but also to see how we're using these, you know, front facing AI chatbots as well. So hopefully you can learn. But as we go along, I'm going to be giving you examples of what we're doing. But you need to, before we start on these Wednesdays, you first need to get into the frame of reference of, okay, get your use case first. All right. So keep that in mind.

We're going to be going through my use case. But this is going to be applicable on Wednesdays. it's going to be applicable to just about anyone. You just have to first think and say, okay, even hit pause if you need to and say, okay, what's my use case for custom GPs, for gems, for projects, et cetera. All right. So we're going to do this one a little different. We're actually going to start live.

All right. This isn't edited. This isn't scripted. This is one of the reasons why I do these things live. I know sometimes I ramble on a little bit, but generative AI is generative. Sometimes things work well. Sometimes they don't. And I want you all to see the realness.

because if you just see these polished finished products like everyone else does, you don't see sometimes where things go wrong or how long they take. So we're going to start live. So also for our podcast audience, thank you for tuning in. This might be one of those. Check out the show notes. You might want to watch the video on this,

although I'm going to do my best to describe exactly what's going on if you are choosing to do this one audio only. All right. So I'm going to share my screen right now. And all I'm going to do, I have a Google gem, a chat GPT, custom GPT, a GPT project, and a Claude project. They're set up the same way. They have the same instructions and access to the same file.

So more on that later. So all I'm going to do right now is I'm going to drop the same exact prompt in each of these. So we're going to see them all work live. All right. So first, I am going into my Google gem that I created. And I even name these the same. They're just called Everyday AI AllStats, June 2025. All right.

I'm going to put this prompt in. We're going to talk about it later. All right. So now my Google gem is off and running. Now I am in my chat GPT custom GBT. And I'm using the O3 model. Some more on that later. Or sorry, O3 Pro.

All right. So we'll see how long that takes. It's probably going to take a minute. Now I'm going into my chat GPT project. Same thing. I'm using O3 Pro. Also, FYI, the Google Gem is using Gemini 2.5 Pro. And then I'm going into Clause.

And I'm actually, I was going to use Sonnet, but let's go ahead and use Claude Opus for, which is their most powerful model. And I do have the extended thinking toggle on for Claude. So essentially for each of these, I'm using the most powerful model that they have access to. And the reason being is because I'm asking for a lot. All right. So more on that. And we're going to get back to that probably in like probably 15 minutes.

All right. I do expect some of them may take up to eight to 12 minutes. But we're doing this live. It could blow up in my face. We'll see how it goes. All right. So let's talk a little bit about the problem. Just using these AI chatbots, whether you're using, you know, Gemini,

Claude, chat GPT, co-pilot, meta, mistrope, right? Whatever you're using. If you're using something on the web, I would say most people, they just go in and they click that new chat, which is actually dangerous. Because one of the biggest mistakes people make is, overlooking things like context window, organization, right? I'm even doing this show as a reminder to myself because I'm right there with you. Sometimes it's human nature when you're using AI now.

It's it's almost become like a default mode for me. I always just go to that new chat button as well. And I'm constantly having to remind myself, like take advantage of the most powerful technology that we have available, which is not just going in there and clicking that new chat. So number one, it becomes disorganized and you have all these chats, right? Think of how you use it for me. I use chat GPT all the time for everyday AI.

I use it for other clients, other projects, multiple things in my personal life. So what happens if I'm not using, you know, whether we're talking about projects that have more of this folder in this organizational structure or if I'm not using GPTs, what happens is I can never go back and find this. So, you know, generally, if you've taken our free prime polish, course, which I know, we got to kick the next round off, guys. I know. Thank you all for being patient. You know, we've had, I think, more than 11,000 people take our live prompt engineering course. But one of the biggest problems is we aren't taking advantage of the context window. And then we're also not taking advantage of the true capabilities that these large language models have. And that's one of the things is being able to take advantage of context window and the not.

that you can share with it. Right. So sometimes you're still going through, but you're having to redo that each and every time. So think, maybe you want chat GPT or Claude or Gemini to respond in a certain way. Maybe you want it to be your brand voice for your company, right? Maybe you're updating a blog post or creating a new SOP. And it needs to be in your company speaks. So you're spending that time to get it to respond exactly how you want to.

And then you're sharing some files, right? But then where did that chat go? Maybe you go in and you search for it. But most people, I would say, are just wasting time. You're wasting time by being more productive, but not in an organized way, which I know sounds backwards. But this is where using Google gems, custom GPTs from chat GPT, projects from chat GPT, and projects from Clawd can help erase this problem because the solution is literally right in front of us. So here's what we're going to do. We're going to talk about why GPT gems and projects are essential for any business leader,

large language models. We're going to quickly show the pros and the cons of each. And then we're going to do this live demo. The live demo is already underway. We started it. We're going to get back to it. Don't worry. All right. So, uh, podcast audience, I got a nice little chart here, a little feature comparison matrix. And hopefully this will make a little bit more sense as we go along. But we have our Google gems, our custom GPTs are projects from chat GPD and projects from Clawton. And where this gets confusing is the core functionality is kind of the same, right? And I would separate these into one category is your GPTs slash gems.

And your other category is projects. Okay. But ultimately, they kind of do the same thing. You give them a set of custom instructions and you give them a project files. And then depending on how your instructions are written, right, usually anytime you go in chat with the jammer GPT, number one or number two, anytime you go chat or create a new chat within that project folder, it is going to look at the custom instructions and refer to the files, the project files that you upload. So on the surface, the two very different setups have a lot of, uh, crossover in terms of utility, right? That's the base utility. But from there, they do get a little bit different. So I want to talk about a little, some of the things that differentiate each. So when

we talk about Google gems, I would say one of the biggest differentiators is number one, it uses the Gemini 2.5 Pro, which until we get full benchmarks for O3 Pro, it's the best model in the world. The other big advantage that you have using GBT's is you have all of their apps integration. So if you are using a workspace account, which I know many of us are in, you know, Google. So if it's, you know, like our email is info at your everyday AI.com. So, you know, it's Google workspace. So if I go in there, I have access to all my workspace apps, which is great. Google Drive, Google Docs, Gmail, calendar, etc.

It's all in there, which is great. Actually, if you're using your personal Gmail, though, you have more. access to different apps. So I hope that eventually Google allows that. So even when you're using gems, you can still click the app button and you can add mention some of these apps, which is good. But like I said, your personal Gmail, you actually have access to more. You have access to things like YouTube, YouTube music, Google flights, you know,

just some things that you don't have access to if you're using a workspace plan. All right. So Google Gem, again, think of it like a personalized, customized version of the big model and you can use at mention Google workspace apps. All right, let's go to GPTs. So GPTs were just updated a couple of days ago. And this came under the radar. Open AI actually didn't really announce it at first.

They updated some projects, which we're going to talk about here in a second, but they enabled anyone to use any version of their model for GPTs. And this is why it's huge. You know, opening eye kind of started this, this phase. of the, you know, GPTs and we thought, oh, it's going to be like the app store, right, for AI. And it never really took off. One of the reasons I think is because you can never choose which model to use. So, you know, as we got these new models that can think in reason like Gemini 2.5 Pro or like, you know, Claude Sonnet 4 or Claude Opus 4, we were still stuck using GPTs, these non-reasoning models

until a couple of days ago. So now you can use the most powerful models. including, you know, 03 and 03 Pro. So didn't even announce it at first, right? So it was kind of this under the radar update. So it's extremely exciting. So some differences with the GPTs. Well, one big one is you can actually connect via kind of external APIs or actions, which is pretty unique in terms of chat GBT, what it can do in GPTs.

You can't really work with not directly third party APIs any other way. an indirect way here in a minute. All right. So then we look at chat projects. Okay, so first we went over in GPs and gems. That's kind of their own category. I like to think of those as just small, little personal,

you know, versions of the big model that follow you around. Similarly, projects are a little different projects. I think originally people thought of them as just a folder structure because it's a folder structure, right? Whereas GPs and gems,

not really. So people think projects are more of just a folder to organize. But it's much more than that. Because like I said, not only, yes, if you go into a project and start a new chat, it will be found and it will live there that helps you be more organized. But like I said, any chat that you start in there has access to the custom instructions and the files. So similarly, how a GPT or gem would.

The difference with these projects is a lot of times you get other capabilities that you don't have in these gems or GPs. The big one in chat GPD projects would be using things like Canvas mode, would be using things like deep research. So that's a new update. You can actually in chat GPD projects now, you can do deep research using just the information in the project. Brand new update. No one else has that. So that's a very unique thing for projects inside chat GPD.

But in general, the projects are good because you can still use the full capabilities of that model. Whereas on GPTs and gems in Google and OpenAI, you can't use things like deep research. You can't use things like Canvas. But you can inside projects. You can use Chad GBT Canvas and you can use Claude's artifacts, which are kind of their version of projects. So a big benefit there. One last one to talk about is integrations. So again, there's some things available in chat GPT projects that are not available in the custom GPTs.

So deep research canvas, but also having that memory across the projects. On Anthropic project side, you have access to your Gmail. So just the different integration. So you have Gmail, Google Calendar, Google Drive, but then you have MCP, their model context protocol. So this is something, again, a very underlooked aspect that you can use the MCP protocol in anything inside of Cloud projects. So that's kind of like a version of API, but for the web. All right.

So I know that was a lot. And hopefully this will make a little bit more sense as we dive in and take a look. But that's just kind of an overview. So let's go ahead and actually, we'll take a quick break here. I've got to take a drink. And also quickly shout out our sponsors at Google. This podcast is supported by Google. Hey everyone. David here, one of the product leads for Google Gemini. Check out V03, our state-of-the-art AI video generation model in the Gemini app, which lets you create high-quality, eight-second videos with native audio generation.

Try it with a Google AI pro plan or get the highest access with the ultra plan. Sign up at Gemini.com to get started and show us what you create. Adobe just introduced an entirely new way to create, bringing the power and precision of its creative suite. into one conversational experience. Meet Firefly AI Assistant, now live in the Adobe Firefly app, the all-in-one creative AI studio. Powered by Adobe's creative agent, Firefly AI Assistant lets you start with your vision, just describe what you want, and shape the outcome as it takes form with the Assistant. The Assistant orchestrates multi-step workflows, drawing on 60-plus pro-grade tools across Adobe

Creative Cloud apps, including Photoshop, Illustrator, Premiere, Lightroom Express, and more to help bring your ideas to life. You can also get started with creative skills, a growing library of pre-built workflows for common creative tasks like batch editing photos, creating mood boards, portrait retouching, and creating social variations. Every step the assistant takes is visible so you can refine, redirect, or take over at any time. You stay in the driver's seat as the creative director.

Adobe Firefly AI assistant now in public beta. See it today at Firefly. adobe.com. All right. So let's first go check in now on the prompts that we started to begin with. So like I said, we're doing the exact same thing and it's the exact same setup in Google GEMs, chat GPTs, GPTs, chat GPTs projects and Claude's projects. So let me first explain exactly what we did. So I did the same prompt for each of them. I said, please carefully analyze my files for everyday AI seen at your everyday. A.I.com. Do not do additional research on the web when, or sorry, I said you can do additional research on the web when needed. Do not make anything up. Take your time. Be detailed, blah, blah,

blah. So here's essentially what I'm asking it for. I'm saying, give me 10 of the most impactful stats or trends that are obvious. Okay. So first, I want to make sure that it can find the obvious things in all of my files. Then I'm saying, I'm asking for 10 of the most impactful stats or trends that may not be obvious. So I'm testing its ability to really, connect common threads and I'm going to tell you what files I'm using. Then I'm going to say 10 problems or opportunities. I should start working on ASAP. Then I'm going to say 10 of the biggest blind spots that I'm not aware of based on this data.

And then last but not least, 10 extremely specific or creative ideas you have based on my data in your research. So ultimately, I'm asking for 50 different things. Now, this is a lot. All right. Now let me quickly, let me just open up. Actually, I'll just tell you. you what we have access to here. So I shared a lot of different things. All right. I actually am just going to open a new tab here. All right. So you'll see here I'm sharing my Claude project file.

So I have a lot of different files in here. I have 10 in total. Unfortunately, Claude would only accept nine because it has a file limit. All right. So anything over 30 megabytes, you cannot upload. So I had a spreadsheet that was like 60 megabob. or no, it was like 42 megabytes or something like that. And Claude couldn't take it, whereas Google gems and GPTs could. All right. So instant kind of bummer for Claude, but, you know, it is what it is. So here's the different files that I uploaded.

Now think this is again where I want you to think of your use case. So I updated. I would have to go and look, but I uploaded hundreds of thousands of rows of data, a lot of data. All right. So here's what I uploaded. I uploaded Google Search Council data for different terms that are bringing traffic to our website. I uploaded Google Search Council page data, so different pages and how they're appearing on the web. I uploaded two different, actually three different kind of files from Google Analytics,

so everything that's happening on the Your Everyday AI website. I uploaded our BuzzSpout, so we use BuzzSpout for our podcast hosting. So just literally all of our stats, the history, every single episode, you know, downloads in the first couple of days, locations, et cetera. And then I also uploaded, let's see, did I get my, oh, oh, no, this is the file that wasn't, that Claude wasn't able to upload. So this is a bummer of clot. The big file was the one from Beehive. So that's what I use for our email newsletter provider. So it's every single email that we've ever sent, which is, that's why it was a huge.

file because we've sent, uh, let me do the math, almost a thousand emails. Maybe not quite, but like eight like 800 emails, I believe, um, or maybe like 700 emails, uh, but it has the content of all of those emails as well. So Claude couldn't handle it, but we have that as well. And then I have one other thing, which I think is kind of cool. We have this, uh, Albucross, uh, file, which this is a piece of software on our website. It's kind of like B2B lead ID. All right.

So, uh, a lot of data. I dumped this all. it would take I obviously know this data because I've been looking at it for a long time. If I were to hire a consultant, it would take them forever just to understand what this data is about. So notice all I did when telling, oh, I actually didn't even tell what the data is. I probably should have done that. Oh, wait, hold up. Sorry, I'm getting distracted here.

In the custom instructions. Okay, this is good. We talk about the custom instructions. So in the custom instructions, I did tell them. what each of these things are. So I said any file labeled GE or sorry, GA is Google Analytics. Anything labeled GSC is Google search counsel.

Oh, YouTube. I uploaded some YouTube data there as well. I think I missed that. Beehive is our email newsletter. Buzz Sprout is our podcast distribution. And Albucross is a B2B lead ID software. All right. So all the models have all of those files,

except Claude couldn't take one. And then they have that set of custom instructions. And I'm asking them for 50 different insights that are going to be, extremely valuable for me and my company and hopefully our growth. So will this can like will this replace like hiring a big consultancy? Maybe. Let's be honest. I couldn't afford to go hire a six figure consultant.

But I think right here we have the makings of it. Right. And again, in this case, I use the most powerful model. What's your use case? Right. And a lot of times I think people are like, okay, Jordan, well, I'm not a business owner. I don't have access to all that data. Well, you probably have access to a lot of data that you're not just thinking about, right?

So what of your company docs can you put in there? What about industry white papers? What about information about your competitors? What about information about your brand voice? What about the last, you know, 50 projects you've worked on and completed? What about spreadsheets that aren't, you know, private? So, you know, obviously only put data that you have the, you know, access to put in there. But think of all the different contexts that you can put in.

there and then every time you go in now think instead of go clicking new chat now all of a sudden if you just unload your brain unload everything about your position put your job description in there who you are who you work with meeting transcripts right you can fit so much information in these gems gpts and projects that all that time that you would normally spend trying to get the most and explain things and feed all this data to a large language model if you just set it up correctly and then use it every single time, think of not just how much time you're saving, but how much better the outputs are going to be. Now, even for me, even if I don't necessarily have a data question for everyday AI,

I should probably still just use this, right? Because again, now this gives these models context, a huge amount of context. Even if I'm not asking for specific stats, if I'm just using, you know, any of these AI chat bots as a strategy partner, as a brainstorming partner to help me plan things. Why wouldn't I want to give it access to all of this information? This is my brain. This is my brain and then some. This is my brain and my computer's brain, right?

I would probably go through here at a later time and, you know, upload a bunch of meeting transcripts. I would probably go through and do a couple like one hour recordings of myself, just word vomiting, right? Hey, here's what's going right with our website. Here's what's going right with the podcast. here's my biggest opportunities. Here's my challenges, right?

How would probably go through and upload those and sort those as well. So think you should be doing the same thing, right? Even taking that unstructured data, your thought process, your domain expertise, talk it into a microphone, transcribe it and dump it into Google gems, chat GPT, GPTs, or folders. All right. So enough and do. Let's just go and check and see how they did. Well, let's first see if they're done.

All right. hopefully they're done. All right. So I'm sharing my screen now. I'm not going to go through all of these. And I don't know. If you want to see these 50 things for all four of these, I don't know. Go repost the episode.

I'll put them all in the document if you care that much. And I'll share them with you, right? If you really want to see the juicy details, go repost this on LinkedIn or Twitter. And I'll send it to you. How about that? All right. So I'm first going to see, did all of these properly go through and do the things I asked. All right. So let's look at the Google gem first. All right. So 10 of the most

impactful stats or trends. The obvious ones got me all 10. The not so obvious stats and trends. Let's see. Got me all 10. 10 problems or opportunities I should be working on. Got me all 10. 10. 10 of the biggest blind spots. Got me all 10. Good. And the creative ideas got me all 10. So Google on their gems side, pass the test. And the good thing is, right, I can also go in here and click this show thinking, which is good. It looks like Google, uh, which is really cool to see Google Gemini, uh, 2.5 pro, uh, which you can use the most powerful model in your gem, which was a huge advantage, uh, for gems until a couple of days ago when, uh, chat, GPT allowed you to have the most

powerful model because before that gems had a big advantage, uh, in that department, but I can click here, uh, it looks like I can see, I can see the code or hide the code in the chain of thought, but I can go through here and see the chain of thought, see how it thought, see how it thought. So actually, you want to get super meta. Probably what I'm going to do is I'm going to upload a transcript of this episode and I'm going to dump all this into all of these files. And I'm going to copy and paste the chain of thought as well and be like, how can I improve this process, right? If you want to get super meta and into my brain on how I work, there you go. But like I tell you all every time, you need to be reading this chain of thought. I'm not going to read it right now, but this tells you because we're using these either reasoning models or hybrid models that go between reasoning and

non-reasoning. But the quality of my prompt was very low, right? The custom instructions were not very great. So when you don't spend a lot of time before you go and set these very powerful models on motion, you don't really know how they're going to respond. Generative AI is generative. It's going to be different every single time. So by reading this chain of thought, it's going to tell me how I should improve my inputs to get a better output. I should probably be more descriptive in certain areas. I probably might need to, you know, use some, some more data to begin with, some better custom instructions, a better prompt. I just wanted to do something basic that I could read out loud in the podcast and not get too into the weeds. But going over here, looking at what Google gems did, it looks

like it did a pretty good job. It found some errors. It looks like it's, you know, it says diagnosing column mismatch. So it looks like it did a pretty good job. It found some issues with my spreadsheets. that's the other thing. There's literally hundreds of thousands of rows of data in there. And they probably weren't all super clean, super correct. Sometimes when you export something, you have to do a little bit of cleanup. I didn't do any of that. I just clicked export all on all of those and just dumped it in there and said,

yo, here's what these files are, you know, figure it out, right? Figure it out. All right. So Gemini and their gems at least pass the test. All right. So now I am going into the GPT. believe, yeah. So the problem is I wish that chat GPT would change back some of their user interface. It was a little easier to see before they made this update, but that's okay. So the first one,

I'm using the gem or sorry, the GPT. So you'll see here, I can go and now choose whatever model I want. So I use the O3 Pro in this GPT. And let's kind of scroll and see how it did. So it actually didn't take long. I thought it was going to take much longer than this. It only reasoned, for four minutes and 14 seconds. So unfortunately, with chat, GBT, you only get a summarized, a very summarized kind of chain of thought. So not a ton of value right here. I can't really see what went right and what went wrong, but I can very quickly look at

the results. I like how these results are formatted. This is pretty nice. So let's see. Okay, interesting. So it said, I'm unable to open any of the data files you uploaded. The Python environment that normally lets me read and analyze spreadsheets and CSVs is returning empty results for every file access attempt. Because I cannot inspect the underlying numbers, I would have to guess, which would violate your request and not make anything up.

Okay, interesting. So in my original testing, and I'll probably just want to open up a new tab and see if I can grab my one that I did when I was planning this show because I thought it didn't happen. It didn't happen the last time, but I'm going to have to go through and double check here. Love doing these things live. So let's see if I can find it here. All right. Okay. It looks like the same thing might have happened twice.

Okay. So for whatever reason, I might have to diagnose this a little bit more. It looks like the GPT was not able to access on the files. going to go into edit GPT, make sure I put them all in there. Yeah, I put them all in there. Okay. So unfortunately, it looks like we had kind of a little bit of a misstep here on the GPT. So I'm trying to see if it actually gave any ideas. So it says, you know, please reshare data in a way I can access. The sooner I can open the files, the more. Okay. So what I'm actually going to do is I'm going to manually upload in these in the body of the chat.

All right. I don't think this is going to make a difference. But let me just go ahead and do this and see if it works, although I don't think it's going to. All right. It's giving me an error. I'm trying to upload too many files at once. There's a limit of uploading 10.

So let me just upload a couple fewer. Let's see how many this is. This is one, two, three, four, five, six, seven, eight. Oh, yeah, that's way too many. All right, let's do this. All right. I'm trying one more time. So now what I'm doing in the GPT, we got the little error.

So I'm just re-uploading the files in the body of the chat, which in theory shouldn't make a difference. But for whatever reason, it is. Okay. So I just found a little bit of an issue that I hope Open AI can iron out. So it looks like it's opening. It was able to open about half of those files. when I shared them in the body of the chat, but when I uploaded them as files in the project knowledge,

it didn't work. So that's interesting here. A little bug, like I said, these new updated GPDs have only been out for a couple of days. So I'm guessing we just discovered a bug. So I'll have to reach out to the team over at OpenAI. If they're not aware of this, they might be. But it looks like now it's taking some time to reason.

So let's go through the projects in chat, GPT and see if we ran into that same issue. So I'm guessing not because I'm looking at the reasoning right now. And it says it reasoned for 13 minutes. And what's interesting here is when I look at the reasoning in a project, I get a lot more information, right? Maybe because it was actually able to go and look at all the files. And maybe I would have gotten a similar amount of information in the GPT if it would

have worked. All right. So let's see if we actually got good results here. All right. So same thing. So the, uh, the projects in chatGBT similarly, uh, said at the moment, I'm unable to open or read the spreadsheets or CSV files that were uploaded. The execution environment is blocking any attempt to list the directory or load the files. So I'm wondering what it did for, uh, that long. All right. So unfortunately, it looks like we got a couple of strikes from chat GPT, which in my original testing, I'm going back now.

And I'm going to double check when I ran this one last night. Okay. So I'm going to share this one. Interesting. Generative AI is fun, y'all. All right. So exact same thing when I ran this last night in the projects and it worked. All right.

So when I scroll through here, in the projects, I have the 10 obvious high impact stats and trends. I have the 10 less obvious insights. I have the three problems and opportunities to tackle immediately. I have the 10 blind spots. I have the 10 creative data-driven insights. So interestingly enough, ran the exact same thing twice the second time. It didn't work inside chat. TPD might be running into some issues.

I don't know. That's why we do these things. live. That's why I always do things multiple times, uh, because pretty interesting, but you need to understand this as well. Generative AI is generative. Just because it doesn't work once, doesn't mean it's broken, right? This is why I encourage companies that I work with that hire us to, you know, don't just do something once and write it off. You need to be revisiting these things, whether it's biweekly, bi-monthly, right, depending on how many people are, are using this technology, what kind of processes that you're building within your organization, you know, never try something

once and say, this is good enough, right? You don't have to go the full, like, benchmarking route of doing things, you know, 50 times and, you know, finding the mean or the median, the average. You don't have to do that, but generative AI is generative, right? I would always encourage you, especially if it's something that you're going to do over and over and over and roll it out within your organization, you should at least be testing it minimum five times, minimum, and you need something to measure. Because you saw in that case, if I did this once for Chad GPT, I probably would have given up. But because I did this multiple times, I know that it had already worked once in the project's setup. All right. Now, let's move over. We've done three of our four. Let's look

at Claude. All right. So it looks like, let's see how long it took Claude to go through this. it looks like it went through fairly quickly. A couple of seconds, not even a minute, which I don't know how, I mean, again, I'll later look at all these and let you all know which one was best. I don't know. I don't have a ton of faith in the fact that Claude went through all this in a couple of seconds. Maybe it'll still be good,

but I would have even hoped even a large, language model would spend multiple minutes on this, thinking about it critically, trying to connect dots between the different data patterns, et cetera. All right, but let's look to at least see if this quad project did the job. So number one, 10 most impactful stats are trend that are obvious. Got it. Number two, the 10 most impactful stats are trends that may not be obvious. We got all 10.

The 10 problems are opportunities. I should be working on ASAP. We got all 10. Okay, interesting here. In the middle of this, Claude went out and did some research, which is good to see. Right. So it started doing, it started doing some things without going to do research.

And then it stopped midway through and said, all right, I'm going to go research some things. So it got me the 10 biggest blind spots that I'm not aware of based on this data. Then it got me the 10 extremely specific or creative ideas based on my data and research. There we go. All right. So Claude did a pretty good job, although I'm going to have to go be the judge and see which one actually. did the best. So we saw some, some failure issues for chat GPT in certain instances on the second time I ran this, because again, I ran it last night. It worked fine on the GPT side and the project

side today, not so much. So it could just be an issue. Chat ChpT's been having a lot of downtime the past couple of days. That might be it. Or this could just be a bug because they did just update both their projects and their GPDs. So I'm probably going to have to come back to this one at a later time to ultimately be a judge on which one I'm going to use the most. You saw Gemini 2.5 gems. It looks like it did a pretty good job, right? I'm trying to see here. It didn't show me how long that it thought, but I looked at the chain of thought.

It looked like it did a pretty good and a pretty thorough job. It identified some mismatched columns and it told me about this. So again, I can't judge the accuracy of everything just yet, but it looks like uh, Gemini off the bat in Gemini in the gems. I might have the most confidence in before looking at everything, just looking at how it handled it. And it looks like Claude also did a pretty good job. I like that Claude halfway through said, yo, hold up.

I got to actually go research some more because in order for me to, uh, you know, find out this information, I'm not sure. So I like that midway through. Uh, it actually did some research on, uh, it looks like it went out and searched for AI podcast market. growth, uh, 2025, uh, and analyze some things there. All right. So, uh, overall, y'all, as we start to, uh, wrap this thing up, I want to just show you, which one should you use. All right. Uh, and I actually, uh, am sharing something on my screen now, which I think is

really cool. Uh, and, hey, just comments, uh, what should you comment? Um, comment, um, comment Give me. All right. And I'll send you this, this little link here. All right, if you want to take a look at this. Because I think this is actually pretty, pretty helpful. So the, I went in and used Google Gemini, which, y'all, the new canvas mode in Google Gemini is so freaking good. Probably one of the most unique things of any AI tool I've literally ever seen is this one little button. No one else has it. So in the canvas mode, so right, Chad Chb-T has a canvas mode. mode, Claude has their version of it, which is called artifacts. So I just dumped in a bunch of data and I said, hey, you know, build me something that helps me choose which one I should use, who, uh, Gemini gems, uh, custom GPs, uh, chat GPT projects or cloud projects. So now on my screen, I'm sharing this. And it's a super cool.

It's actually a little website. So it's actually responsive. So if I, uh, you know, move my, my browser now I'm in mobile mode and it's like a mobile responsive, uh, like website. It's interactive. There's nice animation. So it did a great job of creating a visual that can help you better understand which AI assistant to use.

But what I love about Google Gemini, no one else has it, is it has this ability to one click. And you can actually add AI access to the thing that you're building. Literally no one else has this. Maybe everyone else will copy. And I can't believe more people are talking about this. But now there's literally, Gemini is embedded in this thing. I created and anyone can go use it. So now I can go in.

And let's go to, you know, custom GVTs. So it tells me what it's best for, some of the key features, some example prompts. And then it says this thing of generate more examples. So there is literally an AI element built into this that gives it a dynamic and unique experience for anyone using it. All right. So I can click the, I might have to refresh this here. It's not working.

I was trying it earlier and it was working. All right. So for whatever reason, this isn't working, although it was working earlier. All right. But I did this one last week. If you want to see that AI feature, I actually was able to create a presentation and an actual like literal chat bot. So I'm trying to go. If you want to check that out, that was last week's work with us on Wednesday. So that was episode 544 if you want to go check this out. But let me wrap this up by actually reading what Google, Gemini Canvas mode put together because at the end, I showed you some of the pros and cons.

Some of the issues I had, you might not have them. You might, right? But let's go and wrap this up and say Gemini Gems, Open AIs, GPs, Open AIs, GPs, and Clawed projects, and Claude Projects, which one's best for what use case. So Google gems are best if you are using the Google ecosystem. It's best for Google ecosystem power users who need real-time access to their personal and professional data. So the best features are native access to Google workspace, drive Gmail, and docs.

It integrates with consumer apps if you're using the personal version such as YouTube flights and hotels. And it's powered by the most powerful model, at least the ones that's been benchmarked to date, Gemini 2.5 Pro. All right. For custom GPs, this is best for creating reusable purpose-built AI assistance for highly specific tasks and workflows. Some of the key features are they're modular and highly personalized with custom instructions.

They can be equipped with tools and API actions. And the thing that I didn't even get to talk about is you can instantly call them from anywhere inside chat, GPD. You're using a GPT model, which is a big benefit over projects. So you can be in any chat. And if you want to call on or use your custom GBT, you can just click the app button and you can start chatting with that GBT anywhere. So another unique feature for the custom GPD. from OpenAI, although I do know that Google has been testing that feature out on Google

gems for a while. They just haven't released it. So another kind of good use case there for GPDs. Then chat TVT projects is best for organizing complex work with persistent files and memory for deep research and synthesis. That's the big unique factor there in chat chp t projects, the ability to do that deep research just on the information in there. And as well as using all the models.

and capabilities available like Canvas, deep research, and this new memory across all assets within the projects. And then Claude Projects is best for deep AI work that requires strong integrations with external applications like Google Workspace, some of the good features, connects to Google Calendar, Gmail, Drive, and more, supports deep research and artifacts, and it has access to their most powerful models,

Claude, Sonnet 4 and Opus 4. All right, we went over, a ton, y'all. Let me just wrap by saying this. Don't click new chat. Don't do it. This is also a reminder to myself. It's a habit, right? Also, I use voice mode a lot. And by default, it just jumps in into outside of these modes. But if I go into, you know, a project or a GPT first or a gem first, that's what you should be doing. So spend the extra, you know, 10 to 15 seconds when you're using AI chatbots and get organized, right? Because to get organized takes one step, right? So now as an example, all of these, I'm going to update the files.

I'll put in more information, but there's no reason for me not to go use these gems, these custom GPs or these projects now. Even if I don't think I'll necessarily need them, think of all the extra context. So I essentially turned my AI usage from using a general large language model, which might know everything, it might know nothing to having a fine-tuned, literal like six-figure consultant ready there sitting. So even if I don't think I need all of this data in this context, it's there. It's only going to make all of your results exponentially better. It's going to save you so much time. So whatever you do, don't click that new chat button. Do it the right way and put AI to work for you on this Wednesday. All right, I hope this was helpful, y'all. Do you like these

segments, let me know yes or no. Literally, just type yes or no in the comments right now. If you want to keep doing these things, I hope they're helpful. But if I'm being honest, there are a lot of work to put these together. But I want to make sure that you can show up, even if it's just once a week. Maybe this is the only thing you care about. Maybe you don't care about our, you know, when we do our AI news shows on Mondays where we recap the latest news or, you know, our hot take Tuesdays where, you know, I try to come

with a hot take on something in AI. And then usually on Thursdays and Fridays, we're doing more interview styles, you know, talking to great guests. So maybe this is the only thing you care about or maybe you don't care about. Regardless, is this something that is worth investing the time? Do you like this? Let me know yes or no. Or if you're just going to be reading the email newsletter today, just respond.

Yes to AI work on Wednesdays. No to AI work on Wednesdays. It's up to you. Thank you for tuning in. I hope this is helpful. Please go to your everyday AI.com. Sign up for the free daily newsletter. I'll see you back tomorrow and every day for more everyday AI.

Thanks y'all. Meet Firefly AI Assistant. Now live in Adobe Firefly, the Allman One Creative AI Studio. Just describe what you want to create in your own words and the assistant handles the rest, orchestrating multi-step workflows across Adobe Creative Cloud apps, including Photoshop, Premiere Express, and more in one conversational interface. You direct the outcome while the assistant accelerates execution. Stand control with the ability to step in and refine at any time.

See it today at firefly.adobie.com. And that's a wrap for today's edition of Everyday AI. Thanks for joining us. If you enjoyed this episode, please subscribe and leave us a rating. It helps keep us going. For a little more AI magic, visit Your EverydayAI.com and sign up to our daily newsletter so you don't get left behind. Go break some barriers and we'll see you next time.