EP 588: ChatGPT’s Updated Canvas Mode in GPT-5: What’s new and how to make it work for you

Topics Covered in This Episode:

1. GPT-5 Canvas Mode Overview & Features
2. How to Activate Canvas Mode in ChatGPT
3. Canvas Mode: Inline Editing Capabilities
4. Canvas Mode: Live Code Writing & Rendering
5. Building Interactive Web Apps With Canvas
6. Uploading & Visualizing Data in Canvas Mode
7. Real-World Canvas Mode Business Use Cases
8. GPT-5 Coding Advancements in Canvas
9. Step-by-Step Live Canvas Mode Demonstrations
10. Troubleshooting Canvas Mode Errors & Limits
11. Comparing Canvas Mode to Gemini’s Canvas
12. Five Business Solutions Using GPT-5 Canvas

Keywords:

GPT-5, Canvas Mode, ChatGPT Canvas, GPT-5 Canvas Mode, GPT-5 update, OpenAI, ChatGPT updates, interactive editor, inline editing, code generation, code rendering, business dashboard, interactive web app, vibe coding, non-technical coding, prompt engineering, Context engineering, Retrieval Augmented Generation, RAG, interactive learning game, live coding demo, dashboard builder, data visualization, file upload, CSV analysis, resume website generator, portfolio site with Canvas, client ROI calculator, proposal generator, contract comparison tool, meeting summary builder, budget variance dashboard, enhanced code capabilities, SWE Bench Verified, software engineering benchmarks, syntactic code generation, contextual understanding, model selector, GPT-5 Pro, thinking mode, fast mode, AI-powered productivity, business use cases, workflow automation, interactive apps, visual dashboards, AI coding advancements

Podcast Transcript

Jordan Wilson [00:00:01]:
There's been a lot of talk since OpenAI's GPT-5 was announced less than a week ago. A lot of that talk has been on models that were cut off and then brought back and benchmarks and all of these sometimes more technical things, things that we might read on paper. But I think there was one low key, under the radar huge advancement in GPT-5 that no one's talking about and that is their updated in vastly improved Canvas Mode inside of ChatGPT. So on today's episode of Everyday AI, we're going to be going over what's new in GPT-5's Canvas mode, how it works, and most importantly, how to make it work for you. This is our new weekly segment of Putting AI to Work for you on Wednesdays. All right, let's get into it.

Jordan Wilson [00:01:16]:
What's going on, y'? All? My name is Jordan Wilson and welcome to Everyday AI. This is. This is your daily livestream podcast and free daily newsletter helping everyday business leaders like you and me not just keep up, but get ahead. How we can leverage all these new tools and techniques and modes and models to grow our companies and grow our careers. If that's exactly what you're trying to do. This is your new home. Maybe it's your first time listening, your first time watching. Thank you for that.

Jordan Wilson [00:01:41]:
But your cheat code, how to become the smartest person in AI in your department or your company. That's our website, your everyday AI.com, because there we have almost 600 episodes. You can go watch and listen and learn from the world's leading experts, but also you can go sign up for our free daily newsletter. So we're going to be recapping today's show with some of the highlights, maybe some things that we didn't get to in chat, GPT's new and updated Canvas mode, as well as all of the AI news that's going to be in the newsletter. All right, so like I said, I don't think people are talking about GPT-5's new canvas mode. And I'm wondering why. Yes, there's always especially, I would say the GPT-5 release was one of the most anticipated releases in AI history. You could say, I think when you looked at the original release of ChatGPT in November of 2022, let's be honest, there's very few people that even knew that that was coming or knew what generative AI and large language were.

Jordan Wilson [00:02:51]:
So I do think GPT-5 was probably the most anticipated release in AI history, which is when it was released. A lot of the conversation was based around a few things. Oh, it's, you know, according to benchmarks and user preferences, it's the most powerful model in the world. And then people were complaining that OpenAI kind of got rid of, you know, old models without warning and then there was some hiccups on their rollout and that's been dominating the conversation. But I think probably if you're someone that's just looking at this new model and you're like, okay, how can I put it to work for me? Canvas mode, like OpenAI actually did a few fairly impressive demos of what their new Canvas mode was talking about, yet I literally didn't see a single in depth news article specifically about Canvas Mode. And I was honestly expecting to. So it seems like it was kind of glossed over. But don't worry, we're going to be going over it in today's episode.

Jordan Wilson [00:03:46]:
So today we're going to talk about what's new and improved in GPT-5's canvas mode, give 5 real business use cases of how everyday business leaders can go use it today and show a few live examples of our use cases and things that I'm actually using Canvas mode for and I hope to use it for more in the future and just some fun examples. All right, so like I said, we started doing this AI at work on Wednesdays probably about two or three months ago. So this is our weekly look at how we're using, you know, at everyday AI, how we're using all these new models, all these new modes, all these new features primarily on AI at work. Wednesdays we're usually just using either chat, GPT, Google, Gemini, Anthropics, Claude, or maybe Windows co pilot. So we're really sticking to kind of the more enterprise large language models because I feel that's where a big majority of our listeners are spending their time. All right, so there's actually a ton of things that I want to get to, but let's start live, shall we? Hey, good morning Dennis. Good morning to our Live Stream crew. If you're listening on the podcast, you can always join us live.

Jordan Wilson [00:05:00]:
We go 7:30 live on LinkedIn and other social media, but this thing's unscripted, unedited, and that's why I say when we're starting live. Number one live stream audience, thanks for joining. If you have any questions, comments, get them in now, but we're going to see because there's a good chance some of these things might not work. I think a lot of times when you look at examples or demos online, they're very polished. They've been run through multiple times. I'm going to be doing really pushing canvas mode inside GPT-5 to its limits live, running some examples that I haven't run before. So I'm guessing at least one or two of these is going to break Canvas mode, and if it doesn't, I would actually be impressed. All right, so going to go through some.

Jordan Wilson [00:05:44]:
Some demo demons and some bugs, but we're doing it live. So, Brian, thanks for joining Rodrigue from dc. Appreciate you tuning in. Miguel joining from Porto, Mont. Uh, so thanks. Yeah. Uh, and have you guys been using canvas mode in GPT-5? Am I the only one? I don't understand it. All right, so let's start live.

Jordan Wilson [00:06:04]:
Live stream audience. If you could let me know if you're seeing our screen and podcast audience. This is one of those you might. You might want to follow along for. Right. So if you're just listening to the podcast side of this, you can go watch the full video because there's going to be a lot happening on the screen. I'm going to do my best to describe it to our podcast audience, but you should probably just go to your everyday AI dot com, click on episodes, and then find today's episode, and then you can watch the video or watch it on our YouTube channel. All right.

Jordan Wilson [00:06:35]:
Yeah, this one might be a little hard to follow on the podcast side, but we're going to try our best. All right, so let's go ahead and share my screen live stream audience. Sorry, I had to pull up a fresh window of ChatGPT. So if you could live stream audience, let me know once you can see my screen. I don't want to be talking about things that I'm doing here in Canvas mode and it's not actually popping up. Okay, so a few things, housekeeping, things to get out of the way. So GPT-5 was released on Thursday. So here we are less than a week later, and there's already been multiple changes to ChatGPT.

Jordan Wilson [00:07:17]:
So a lot of them are model names and limits. Okay, so one thing that's new as of just a couple of hours ago, and it does pertain to Canvas, and I'm going to tell you why is now there's different model names. Yes, confusing. All right, but now you have auto mode, fast mode, thinking, Many thinking and pro. Okay, we'll probably go into these modes a little bit more, but Auto decides how long GPT-5 will think fast gives you instant answers from GPT-5 thinking. Mini uses thinking but quickly in GPT-5 thinking normal mode thinks longer for better answers. And then you have Pro GPT-5 Pro, which is research grade intelligence, which you only have on the Pro plan. But one thing to keep in mind, Canvas mode does not work on Pro.

Jordan Wilson [00:08:14]:
I hope they change that in the future because I would really like to see what the combination of the Most powerful, powerful GPT-5 variants with canvas can do. So for some of these, I'm going to be going between Thinking Mode and Thinking Mini. Okay, so if you're following along, you're going to start by first choosing, you know, are you going fast? Are you going thinking Mini, are you going thinking? And then in Canvas mode, you have to activate it. So sometimes you can ask the model to use Canvas mode and you don't have to click on it. But I always encourage people to click Canvas mode. All right, so what you're going to do is you're going to click the plus button, which it used to be a toggle that was on screen, so now you have to actually click plus, you have to scroll down to more and then you have to activate Canvas. All right, so I'm going to show you a couple of things if you're brand new to Canvas, a couple of things that Canvas can be used for. So it actually has two very distinct purposes.

Jordan Wilson [00:09:17]:
One, one, you can use Canvas for an interactive inline editor inside of ChatGPT. Let me show you what that looks like. So I'm going to say, you know, write a sales page for everyday AI podcast by Jordan Wilson, promoting a new GPT-5 prompting course. Should we update it? All right, so I'm using the Thinking Mini mode and I have Canvas mode enabled. Okay. So the first kind of feature, and for the most part this isn't new, but a lot of people don't know how Canvas works. And it's actually even this right here, this inline editing version of Canvas is extremely underrated. So you have your inline editing kind of variation where it's almost like you are working with ChatGPT in a live document as a thought partner.

Jordan Wilson [00:10:12]:
But that's one part of Canvas. The other part of Canvas is it can run and render different code even if you don't know how to code. I don't really know how to code. I know the basics, but I can't code by hand. So the second big use case of Canvas is It can run in render code. So you can have it write code and then it can run it as if, you know, you were setting up an entire website and you had everything technically in order. Okay, so on the first side here. So I did a very simple prompt, I said, right, Sales page copy for Everyday AI by Jordan Wilson, promoting a new GPT-5 prompting course.

Jordan Wilson [00:10:48]:
And you'll see what happens on the right hand side, it created some landing page. All right, not the best, but this is just for an example here. Okay, so now it says, you know, by Jordan, you know, Everyday AI presents GPD5 prompting mastery by Jordan, host of the Everyday AI podcast. Right. So what I'm going to do right away, just to show you an example of how Canvas mode works in this one use case, I'm going to highlight this first section here. Okay. And then when I highlight it, there's a. Essentially popup.

Jordan Wilson [00:11:24]:
So it's kind of like you're working in like Google Docs or Microsoft Word, right? Which if you're you, if you're not using G Chat, ChatGPT and Canvas mode, you might be like, wait, you can do this? Yes. So when you highlight something, you have some formatting options, but you ask, you also have this thing to say, ask ChatGPT. All right, so I'm going to say, you know, don't mention Jordan's name. This is for a landing page. I'm going to say, you know, show, don't tell or something like that. I don't know, we'll see what, what happens. This is more just for an explanation because what you'll see, you know, podcast audience, when I do this, it's just going to change and update the section that I highlighted. Right.

Jordan Wilson [00:12:06]:
I probably should have switched the mode for this. Out of thinking, but it actually just updated it in like five seconds. So there you go. And it worked right away. Okay, so now, so not only can you highlight something like that and have ChatGPT change something in line and nothing else changes, you can also go down and write something yourself. Right? So I can interact in the same document as ChatGPT. So without, you know, let's say only the middle needs to be changed, right. You don't have to, you know, go back and forth and oh my gosh, I only wanted the middle couple paragraphs changed or modified.

Jordan Wilson [00:12:49]:
But ChatGPT changed the whole thing, Right. So now you can just highlight certain sections or you can go in and start chatting yourself. So that's one quick, you know, one half of what Canvas mode is. But I don't think that's where the, the true new power and capability is coming from. All right, so now let's look at the second half of the use cases, which is where we're going to spend the majority of our time today, which is looking at writing and rendering code, even if you don't know how to code. I think there's so many great business use cases here. All right, so let's go ahead and try this example. All right, again, we're doing this live.

Jordan Wilson [00:13:28]:
I'm going to keep it on thinking mini. All right. Because I really want this one to work. So I'm going to start a prompt and let me make sure. I should have made sure. All right, I think. Thank you. Thank you, Joe and Angie for saying you can see it.

Jordan Wilson [00:13:48]:
Sorry, sorry, Shane. I do a lot of chit chat here. All right, so now let's go ahead and run this in Canvas mode. So one thing as an example, and I did this last week during our AI AI at work on Wednesdays, an example of learning the difference and similarities of context engineering, which is kind of a new terminology or thought on large language models and RAG retrieval augmented generation. So what I said in this prompt, and you'll see for our live stream audience, it is writing code live. So I said in the area of large language models, I'm trying to learn more about the differences and similarities of context engineering and rag. I said focus on 2025 info only. Then create an informative and interactive learning game in Canvas mode.

Jordan Wilson [00:14:45]:
Make it fully interactive. Make it a fully interactive web app that renders in canvas. Make it interactive, informative, and some sort of element that makes it either competitive or scoring with progress. All right, so then you'll see it finished running the code. Great. All right, so here's what we're going to do. Sorry, we're going to be juggling around. So before we look at that, I'm going to try and start another one.

Jordan Wilson [00:15:11]:
So I'm going to start another chat and I'm going to use the full thinking mode with Canvas. All right, so I'm going to explain what this one is later, but I'm going to show you live because I think this is one of the ones that's going to break it because I'm going to upload some pretty big files. So that's another thing that you can do that a lot of people don't know is you can upload different files. So I, I will let everyone know what these files are, but I want this one to go live because I Think it's going to break it? One of the CSV files here is 66 megabytes. It is gigantic. All right, so I don't even know if it's going to upload. We're going to see. It looks like it's going to upload.

Jordan Wilson [00:15:54]:
Okay, interesting. So all the files uploaded. So I'm essentially uploading a bunch of my podcast and newsletter stats, and I'm asking ChatGPT thinking mode to make a dashboard. All right, so this one, I assume, is probably going to take 10 minutes. So it's like a cooking show. I'm putting that one in the oven. And then we're going to go back to our context engineering verse rag game. Okay, perfect.

Jordan Wilson [00:16:22]:
It looks like cake number two is starting to bake, baby. All right, so now we have our code that is done. So this is our kind of interactive learning environment. That's the difference between context engineering in large language models and rag. So it ran the code. So I have a bunch of code here, and I'm looking at it, and I know the basics of what's happening on the right side, but I actually know nothing else here. And I think that's important for people to understand. When working with Canvas, you don't have to know anything about coding.

Jordan Wilson [00:16:53]:
Right. You might hear this term, vibe coding. What's it mean, what's it do? Right. Throw a problem at a large language model and especially GPT-5's new canvas mode, and just have it build you something that you can use to learn better, to be better at your job, etc. All right, so now in the upper right hand corner, once the code is done, I can click run code and it says run code in Canvas. So. All right. Phew, it worked.

Jordan Wilson [00:17:19]:
That was a close one. I would have hated for it not to work. Okay, so we'll see here. And you can resize it a little bit here. And then you can also share this code. You can download it and you can copy it, because people are always like, okay, well, what would you do with this? All right, if you have someone on the IT side, you can essentially send this code to them, and you really wouldn't need anything else. You could host something like this on your website, on your company's intranet. Right? So you can build a lot of different things.

Jordan Wilson [00:17:47]:
If you're uploading files, you know, you have to make sure to upload those on your server and make sure that your essentially web environment has any libraries that Canvas mode is. Is using. Right. But you can also say, here's my Website. Here's our tech stack. Can you build this in a way that it will work with us just copying and pasting this file on our server essentially? So, yeah, a lot of people are always like, well, what can you do? Well, I can share this with anyone and they can use it in the ChatGPT environment or I can copy this code and host it on my website. All right, but you'll see here we got a working. Looks like it's working a working interactive game.

Jordan Wilson [00:18:25]:
So the top it says context versus rag Interactive learning game. So it says start, open the tutorial. All right, this is cool. It's like it's giving me an order. So it says to start, open the tutorial, then try the interactive modules in any order. Your score in the top right, updates as you complete items. Progress persists in your browser. All right, and then we have a little why it matters.

Jordan Wilson [00:18:47]:
All right, in 2025. And then a suggested path. So it says read the tutorial for five to eight minutes, play the matching game. All right, and then run the rag simulator and take a timed quiz to test retention. I'm obviously not going to go through this all, but we're going to click around and we're going to see if it works. So here in the tutorial, I clicked it. It works. So it's kind of like an interactive tab website with some bright different colors.

Jordan Wilson [00:19:12]:
So some examples here. Then I can go to the matching game. All right, so let's see what it says. Re ranker. All right, so I'm not exactly sure how this game works. All right, but you can see here it's interactive and it looks like I have to rank different things. I'm not going to go through and read the five to seven minute tutorial. But it looks like it works in the matching game.

Jordan Wilson [00:19:35]:
And then there's a rag simulator, which is pretty cool. So you can change your knowledge base size with a little slide slider. You can change your chunk size with it, which is tokens. There's a re ranker. And then you can run a simulator. Very cool. All right, it did it. And then there's a timed quiz.

Jordan Wilson [00:19:56]:
All right, so question one, which best describe describes context engineering. Designing what the model sees, a specific retrieval algorithm for vector databases or training a new large language model from scratch. So one of those is rag. One of those is pre training. One of those is context engineering. The first one. Oh, I like this. This is a timed quiz too.

Jordan Wilson [00:20:13]:
So I know the answer for this. Luckily it's the first one. All right, I'm not going to go through all the questions to see which ones are right and wrong, but there you go. A fully working interactive game right there. Right. I didn't have to write a single line of code. It didn't take long at all. All right, so I'm assuming my other one is still going to be working.

Jordan Wilson [00:20:36]:
So we're going to start a new chat here inside GPT-5. We're going to go thinking mini because I think this one is not going to be too difficult. All right, so here's what I'm going to do for this next one. So I have a copy of my resume. I took out all my contact information and I'm going to copy and paste this. My resume hasn't been updated and a long time. And then I'm going to go back into ChatGPT I'm going to paste this prompt in. All right, I'm going to select Canvas Mode and I'm going to make sure at the top that I tell it to also use Canvas Mode.

Jordan Wilson [00:21:14]:
Please create a. I'm going to say using Canvas Mode. All right, here we go. So here's what I put in this prompt. So I said, this is my resume using Canvas Mode. Please create a beautiful interactive website that I could use as a personal portfolio. If you need to, please browse the web to find out more info about me or context of my resume. And that's really it.

Jordan Wilson [00:21:42]:
Right? So maybe you're trying to get a promotion within your company, right? And here's a cool way to do this. Maybe you're a small business owner, a freelancer, right? And you're just looking at your resume and you're like, man, there has to be a way that I can spruce in this up and do something creative. And there you go. Let's see how long it took. It took six seconds and it looks like it's already done. So I don't think this is going to be super great, but I could be wrong. All right, let's find out. Live stream audience, let's see what's, what's in the oven here.

Jordan Wilson [00:22:20]:
All right, so now I clicked run code. All right, it's actually. So I need to zoom out here. I'm a little too zoomed in and things are running together. All right, this is actually better than I thought. Okay, interesting. It's, it's interactive. Some elements of it are interactive.

Jordan Wilson [00:22:45]:
So you see here it created almost like a one page website with different tabs. So right now by default it's on the about tab. Then I can click Experience. And it looks like it should scroll down, but it didn't. There's a speaking thing. It didn't work. And then a contact button. The content scrolls down to the bottom and it looks like there's somewhere that someone can send an inquiry.

Jordan Wilson [00:23:09]:
So pretty good. But the good thing is, is like when you're working in Canvas mode, just like any large language model, you don't have to know anything about the COVID code to improve it. So I'm going to say. I'm going to say the buttons don't always work. Aside from the contact button, which works, I'm going to say fix that. I'm going to say make it more interactive and cleaner and more visual. And I'm going to say make it dark mode. All right, so I essentially gave it three simple text commands.

Jordan Wilson [00:23:49]:
And you can do this, obviously, with anything. I could have done this with the. The context engineering versus rag simulation. So you only. Let's see here. So one thing I will say in GPT-5, and this could be maybe because of some Chrome extensions that I have going, sometimes I have to actually manually refresh the page for it to work. All right, so you'll see here it said failed to generate. All right, so we got our first error.

Jordan Wilson [00:24:20]:
So I'm going to make sure that Canvas mode is selected and I'm going to say at the end, I'm going to say use Canvas mode to update this. All right, so here we go. Take two again. I should have first disabled some of the. Some of the Chrome extensions that I had going because I know that sometimes those can cause some issues rendering the code. I actually know for a fact because I've had to troubleshoot this before. All right, we're going to give this one a minute. Live stream audience.

Jordan Wilson [00:24:50]:
If you have any questions, I would love to hear them. So Daniel said, what happened to Canvas mode in the ChatGPT app? I don't see the option for it anymore. Yeah, I saw that as well. I'm guessing that they're going to fix this. All right, Michelle says love the extra tools it offers, plus the coding. Yeah, same. All right, so here we go. Let's see if it updated our portfolio website.

Jordan Wilson [00:25:15]:
We're going to run the code. All right, what do you think? Live stream audience Second version. Little better, right? A little better. I like this. So it gave us dark mode. That works. Everything else looks similar. It does look a little cleaner, at least for my taste.

Jordan Wilson [00:25:37]:
Let's see if the kind of the buttons work. So. All right, the about button kind of scrolled it. The experience button. Now it's scrolled it in. So if you've ever used a one page website and there's little buttons and then you click the button and it scrolls you to, that's what's happening here. So it essentially turned my resume into an interactive website. So now those buttons are working.

Jordan Wilson [00:26:01]:
That's great. It looks like there's a toggle for light mode versus dark mode which didn't work. But again, if I really wanted to to do this, I could. There's also pretty cool. I didn't know this. There's also some little pop up modals so I can click speaking and then there's a little, there's a little kind of pop up that comes up about different talks that I gave and then you can email or request a booking. So pretty cool. I didn't even ask it for that, that functionality, but it went ahead and built it, which I think is pretty cool.

Jordan Wilson [00:26:36]:
All right, so we have a couple versions. I don't know if our other one is going to fully work. I'm going to refresh it here so our live stream audience can see it going. We might have to check in on it at the very end. But I do have one other use case that I kind of like pre built and this is something that I've been using. Right. So I don't know if that one is going to work. Like I said, I'm using the highest level of thinking and I threw a ton of data at it.

Jordan Wilson [00:27:10]:
So one thing in Canvas mode, you can see right away, it put the content of the spreadsheets in the actual kind of view. Right. So I can click to expand this table. So these are kind of our buzz sprout stats. All the. My gosh, these are all the different cities. Apparently that our podcast has downloads in more than 27,000 cities. I didn't even know there's that many cities in the world.

Jordan Wilson [00:27:36]:
All right, so you can see here it instantly converted all of my spreadsheets and all of my data into interactive tables inside of ChatGPT. All right, this would be a bummer. So it thought for 12 minutes and it looks like it failed to generate. All right, I'm gonna say please regenerate. So again I thought, actually it looks like it's still working. So we'll have to check back in. We'll have to check back in on this one. All right, I'm going to say, you know, please update.

Jordan Wilson [00:28:19]:
All right, so we broke one. We broke one. I knew that we would probably break one. Doing that live, you know, less than a week after ChatGPT came out also, or after GPT-5 came out also. One thing to keep in mind, GPT-5 just a few hours ago did update to this new kind of model selector, model picker. So I'm guessing a lot of people are going in to test out the different modes. So probably not the best time to really push ChatGPT to the limits. But hey, I broke it once.

Jordan Wilson [00:28:47]:
All right, so one other use case that I actually use this all the time, right? And if you listen to our podcast, I said in 2025, non technical people are going to be essentially vibe coding different little apps for themselves. So as an example, this is an app that I use all the time. All right? I don't know why there's no great podcast revenue calculators online, right? I found one like three months ago and I didn't save it and now I can't find it. So you know, for me, right, we have a lot of companies who like to advertise with a podcast, right? We recently had Square on as an advertiser. We've obviously had Google on, we've had Microsoft on, right? And a lot of times I'm trying to figure out what is the right price, right? So it's hard. You know, our stats are always changing, right? So I'm always like, wait, I'm not good at this math. There's a formula with a cpm and I'm always trying to find the right calculator online and I just can't find it because it's a very niche use case, right? And a lot of these calculators online are actually for the media buyers, right? Because it's like put in your budget. It's not for the actual, you know, people with the podcast who are trying to find the right price for ads.

Jordan Wilson [00:30:12]:
So I just literally built, built myself a simple podcast revenue calculator where you can put in the CPM, right? So we might usually do 50. All right? And then I can sort it by per episode, daily, weekly, etc, but then I can go, you know, I can say, let's see, daily. I don't know, we probably have, I'd have to look probably about 11,000 downloads a day. We have three ad spots and if we fill them all a hundred percent. So there you go. So I can go and very easily say, hey, according to, you know, fair market prices out there, here's about what I should be charging, what we should be charging advertisers. And then we can also explain it Right. We didn't pick up a random number.

Jordan Wilson [00:31:02]:
Nope. Here's, you know, industry averages for CPMs for the type of audience that we have, etc. So this is literally a way that I'm already using GPT-5, their new canvas mode, and I'm probably going to make some updates to it as well, but very simple way that you can use it. All right, so let me. Let me double check. Yeah, I think we broke the one. I'll check in on it at the very end. But I do want to, before we wrap, if you do have any questions, get them in now.

Jordan Wilson [00:31:37]:
Yeah, a lot of people. So, Robert said, still impressive results with decent response times. Yeah, I agree. I'm. I'm loving canvas mode in GPT-5. All right, so let's go back here and I want to talk a little bit of the details. Right? We started out live, you know, took about 15, 20 minutes to go over some actual use cases that we're using here at Everyday AI. But I want to talk now specifically some of the details in the coding advancements, because I would say if you look at GPT-4O and GPT-5, some of the biggest jumps are in coding, right? So you pair that with this new and improved Canvas editor, and I think this is Vibe Coding 101, right? If you don't want to go, you learn cursor or pay for tools like lovable, etc.

Jordan Wilson [00:32:31]:
For the 700 million weekly active users, you have a Vibe coding basic platform, Right. I wouldn't recommend you try to build an app, right, an MVP for a SaaS or something like that inside Canvas Mode, but. But you can at least get something started. But let's talk about some of the coding advancements in GPT-5. So their latest model demonstrates significant improvements in coding capabilities, surpassing previous iterations and competing models on various benchmarks. One of the big ones was SWE Bench Verified, which is essentially a software engineering benchmark. And it achieved a huge increase, about a 75% accuracy, whereas GPT-4O was 30%. Right.

Jordan Wilson [00:33:14]:
So if you were previously using canvas mode with GPT-4O and, you know, trying to run and render some code, and you're like, this isn't that great? Well, yeah, it should be now more than twice as good. Which is, again, why I'm like, scratching my head and I'm like, why aren't more people talking about Canvas Mode inside of GPT-5? You know, if you've been listening to the show, I've been bonkers for Gemini's version of this. Gemini's Canvas mode since it came out. And they even had a low key update where you can use Gemini's AI features in the kind of AI apps that you build. I would love for GPT-5 Canvas to offer that, but right now they don't. So you have enhanced code generation because GPT-5 demonstrates stronger capabilities in generating functional and sin synthetically correct code, syntactically correct code snippets in response to user queries, and there's improved contextual understanding. So as the base model improves in understanding your queries, combining that with better coding under the hood and a improved canvas interface. Right.

Jordan Wilson [00:34:27]:
It's really just helping you better have an iterative conversation to build something that you can use. So the example that OpenAI gave in their GPT-5 live stream is this engineer at OpenAI essentially created a gamified version to learn French, right. And there was kind of like flashcards and quizzes and also a little game, right. So you can really direct it. You can have an open ended prompt and just say, hey, here's a bunch of data, kind of like the one that I broke GPT-5 with. Or you can say, hey, here's exactly what I want, which I like. So normally when I'm, you know, building like a business dashboard and interactive kind of strategy guide based on my data, I'll usually just let GPT-5 canvas do what it wants to at first and I'll see what it creates and then I'll either iterate that or, or I'll start a brand new canvas based on what it built me and then ask it for specifics. Right.

Jordan Wilson [00:35:33]:
So again, seeing examples and what people have built with it and iterating on the original is going to give you the best results. Speaking of use cases, all right, so I'm going to give you five different use cases for the new GPT-5 canvas mode. So one is a contract comparison tool. Right. So again, what are these things that you think I gave you an example of? Some, some little things that take up my time. Right. When I'm trying to learn something in a way that is actually ingrained. Right.

Jordan Wilson [00:36:09]:
I love using NotebookLM their new video overviews, but I'm using Canvas mode, right. I'm a very, you know, competitive person. I like to gamify things. So before I do a podcast on something or if I'm interviewing an expert, I might go into GPT-5 canvas now and create an interactive learning guide so I can really, when I'm interviewing someone, I have a better understanding. So think of what it is that you struggle with on a day to day basis or with the software that you have. It's something clunky, right? So you could create a contract comparison tool. This is Use case number one. So again you have to make sure that you have the, the permission to upload any documents.

Jordan Wilson [00:36:46]:
I'm going to put that out before I go over these five different use cases. But what you can do is you can upload two different vendor contracts or service agreements and then have a side by side dashboard that looks at the different liability clauses, payment terms, potential red flags, etc. And this can just save you hours of manual reviews. Right? So that's use case number one, Use case number two. Build a meeting summary builder. Right? There's all these apps out there, right? Some of them have okay, free plans, some of them have, you know, paid plans. But on the desktop app on Mac, ChatGPT actually has a, like a meeting function, right, where you can kind of record meetings. Obviously you should be getting consent of the people on those meetings, right? But then you can just literally use Canvas mode and have a meeting summary app so you can upload those transcripts that ChatGPT can record.

Jordan Wilson [00:37:44]:
And you can build an interactive action item tracker with assigned owners, deadlines, priority levels and automated follow up reminders. Right? Pretty cool. Yeah. You might already have systems in place for this, but if you don't and if you, maybe you're a small team or your team is heavy ChatGPT users, there's a way to do it right away. So instead of just a follow up email that you send to everyone, remember again, you can share these creations in canvas mode. Right? So you can create instead of a follow up email that's kind of boring and dry and you know, not fun to work with. You can create an interactive visual dashboard. Right.

Jordan Wilson [00:38:24]:
If you have one big important meeting every single week, I think that's a great way to look at it. All right, Number three, use Case, a client ROI calculator so you can upload your service pricing and your client's current metrics. And then you can have GPT-5 Canvas build an interactive calculator where prospects input their number to see input their numbers to see projected savings, revenue gains and a payback period if they use one of your certain services. Right? That's another thing. You have all these y' all like in the early 2000s, there's all these essentially online calculators that became like multi million dollar businesses over the course of multiple decades just because there's niche use cases that people Always want calculators, right? Like sometimes it's like, okay, I need to upload this document and count, you know, the word frequency or something like that. Or you know, a comparison tool. Right? Here's a great one. You can build this for your company, put the different, you know, whether it's, it's hourly by the project, you can put in all your prices and then you can create a front end calculator for your clients or potential clients.

Jordan Wilson [00:39:32]:
Because as an example, if you know that one of your USP as a business is there's a certain product or service that you provide a better value versus your clients, you can literally show people that with an ROI calculator that you could host on your website or you could send it to clients. All right, Another one, proposal generator. Literally you can just build software, right? There used to be software, I remember paying, I don't know like couple hundred dollars a month for a certain piece of software that was just for proposals. Right. Obviously you're not going to be able to have a full fledged piece of software with one prompt. But if you want something that's better than just sending, you know, a bullet point email or a PDF Word document, that's a proposal. You can create something interactive, you can create something highly designed and with a great UI ux. Right.

Jordan Wilson [00:40:24]:
So you could upload your service, catalog past winning proposals, client requirements and GPT-5 can autogenerate and build you proposals with customized pricing, timelines, deliverables and terms based on your clients needs. All right, so yeah, in that you can cut down proposal time from maybe many hours, multiple days of back and forth. Right. You can even have a form where people fill that out, pass that information to GPT-5 canvas or you know, use it within GPTs. That's another one, right? GPTs, custom GPTs. You can have canvas mode enabled in them. Right. So that's another kind of huge hack.

Jordan Wilson [00:41:04]:
I didn't even probably should have started with that. All right, and then use case number five, a budget variance dashboard. So you can upload monthly or quarterly budget versus actual spending and have GPT-5 in canvas mode, build a visual dashboard showing variances by department category and timeline with drill down expectations and forecasting. So in this case you can spot budget issues early, make data driven decisions and improve financial planning accuracy. All right, so those are our five use cases. You know whoever said someone on LinkedIn, I can't see who said that said, I so appreciate your bravery in doing this live, Jordan. Yeah, doing live demos is not easy. Right? But this, I mean this is one of the reasons why I do everyday AI the way I do it.

Jordan Wilson [00:41:55]:
I think so much out there. It's pre polished, it's, it's scripted, it's edited. Right? But that's not how generative AI works. Sometimes, sometimes you ask it to do something and you run into some hiccups, right? Which is what we ran into when I try to get it to do the, my business dashboard with all my data. But you know what? I'm going to rerun that later and I'm not going to use the, the full thinking mode. I'm going to use thinking Mini. So. All right, if you want to see how that one turned out, I'm going to share it in the newsletter and you can go look at that interactive dashboard that I built.

Jordan Wilson [00:42:34]:
All right, so that's a wrap, y'. All. So to recap, because like I said, I think a lot of the recent attention and buzz around GPT-5 is on benchmarks, is on all this drama of they got rid of the old models and now they're back, right? People wanted the, you know, these models that just agreed with them, like GPT-4O, even though GPT-5 is better, right? And then there was all these issues that OpenAI had. Rolling it out and hidden in all of that is I think, the best feature for everyday business professionals. And it is turning your time consuming manual, at times monotonous decision making projects and tasks into something a little more interactive, into things that are going to save you time and ultimately save your business or your department money. Right? So obviously there were huge gains in GPT-5O when it came to coding, when it came to software development. And you might look at that and say, okay, well that's great for the people who build software. And I say, no, right? You saw an example.

Jordan Wilson [00:43:54]:
I'm building little pieces of software almost every single day, right? And I've really enjoyed, although I love using Google Gemini's Canvas, I've really enjoyed these updates in GPT-5 canvas. I think this gives a new set of tools to hundreds of millions of ChatGPT users, but so few people know about it. All right, so now you know how it works. And you can put the updated canvas mode in GPT-5 to work for you this Wednesday and every day. All right, I hope this was helpful, y'. All. If so, don't be a jerk. Let someone know about this.

Jordan Wilson [00:44:39]:
All right, if you're listening on the live stream, if you found this helpful, people are always like, oh, Jordan, everyday AI has been my cheat code. And I'm like, well, tell someone about it then, right? I think there's so much garbage out there in the world of AI smokes and mirrors, marketing people just trying to sell you crap. I want to make ongoing, updated, factual AI education for the masses free. And it only works. It only works if you tell someone about it. So if you're listening live, please repost this, tell someone, send this out to your company. You know when we send out today's newsletter for this to someone that needs to hear it. If you're listening on the podcast, appreciate your support.

Jordan Wilson [00:45:18]:
I know this was a more visual episode, so like I said, if you want to see the video version, you can go check out our YouTube channel or go to our website at your everyday AI dot com. Or while you're there, make sure you sign up for the free daily newsletter. We're going to be recapping the highlights of today's show. If we get that dashboard to finish, I'll share that out there as well as you can see everything we got going on under the hood. So thank you for tuning in today. Hope to see you back tomorrow and every day for more Everyday AI. Thanks, y'all.