Everyday AI Podcast – An AI and ChatGPT Podcast - EP 585: GPT-5 Released: 7 big trends you should know

Episode Date: August 8, 2025

Obvious: ChatGPT's GPT-5 is here and it's really good.Not so obvious: the gap between novice and experts just shrunk 90%. In a short few hours, OpenAI gave even free users access to now the world's most powerful model. As the most used AI chatbot in the world by a wide margin, the quality work we all produce has also just gotten a huge bump. But there's a lot beneath the surface. Join us as we dissect what's new in GPT-5 and 7 big trends you probably don't know but should pay attention to. 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:GPT-5 Official Release OverviewGPT-5 User Rollout to 700 MillionGPT-5 Unified Hybrid Model ArchitectureAuto Model Switching and User ControlMajor Upgrade for Free ChatGPT UsersGPT-5 Benchmark and Accuracy ImprovementsGPT-5 Vibe Coding and Canvas FeaturesAdvanced Voice Mode in Custom GPTsReduced Hallucinations and SycophancyMicrosoft Copilot Instant GPT-5 UpgradeImpact on Enterprise Software and APIsGPT-5 Disruptive API Pricing StructureTrends in Corporate AI AdoptionTimestamps:00:00 "Everyday AI Insights"05:54 "Adaptive Model Response Modes"08:14 GPT4O Model Critique11:17 GPT4O Nano Upgrade Impact17:26 GPT Model Selection Simplified20:53 Canvas Code Rendering and Quick Answer Feature24:09 "GPT5 Model Routing Overview"26:44 "GPT-5: Your New Daily Driver"30:08 AI Model Advances: Game-Changing Improvements33:43 Advanced Voice Mode in GPTs37:45 Massive Microsoft Copilot Upgrade38:49 Software Access and Licensing Challenges43:09 AI Implementation Challenges in Top Companies46:37 "GPT-5 Testing and Trends"Keywords:GPT-5, GPT5, OpenAI, AI model update, Large Language Model, flagship model, hybrid model, AI technology, model auto-switching, deep thinking mode, fast response mode, model router, free AI access, paid ChatGPT users, ChatGPT free users, model selection, GPT-4O, GPT-4 Turbo, model reasoning, hallucination rate, sycophancy reduction, advanced voice mode, GPTs custom models, Canvas mode, Vibe coding, API pricing, API tokensSend 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. 700 million people are about to get access to a technology so powerful we've quite literally never seen it before.

And we don't know what it'll mean. Let me repeat that. 700 million people now have access to OpenAI's newest model GPT. And it's actually pretty big. I can't understate it. Yes, it's going to be slowly rolling out. There's going to be outages over the first couple of days. But opening I just released after a two and a half year waiting period, its next flagship model.

And I think, yes, it's obviously very impressive. For a power user like myself, I don't think the ceiling went up. exponentially, but the floor has shot up to about where the previous ceiling was. So I'm going over what I mean by that, telling you everything that's inside of the new GPT5, as well as filling you in on seven big trends that I think you need to know. I'm excited for today's show. I hope you are too. What's going on, y'all?

Welcome to Everyday AI. My name's Jordan Wilson, and this is your daily live stream. podcast and free daily newsletter helping everyday business leaders like you and me, not just keep up with AI and all these new model updates and every single day. There's an alphabet soup of new things to try, but how we can actually take advantage of that, leverage that to grow our companies and our career. So if that's what you're trying to do,

you're in the right place. It starts right here with the unedited, unscripted, going live off the hip podcast and live stream. But if you want to take it to the next level, your cheat code, write it down. your everyday AI.com. That is our free daily newsletter. You're going to want to sign up.

We're going to be recapping the highlights from today's show, everything you need to know in GPT5 that maybe we didn't get time to get to in today's live stream, but also more than 780 episodes. You can go listen to. Learn from the smartest people in the world for free on our website. And if you want today's freshest AI news, aside from GPT5, make sure to go check that out in the newsletter. But let's get straight to it. GPT5 is here. If you're a paid customer, you probably should have access to it already. And it's going to be rolling out to free customers as well.

So probably by the time you're listening to this on the podcast, you should have access for everyone else on the live stream. Yeah, it still might be a couple of hours. But the rollout should almost be complete. Like I said, for 700 million users are going to wake up today or tomorrow or later this week. So Open AI just recently reported, they have 700 million. active weekly users. Active. That's just an insane amount of active users who now have access to an AI technology.

Like I said, we've never seen it before. It is by far the best model out there. I mean, we'll see whether it's a couple of hours, a couple of days, or a couple of weeks until Google responds. We know they're going to respond. But at least right now, literally, you have a huge fraction, a huge percentage of the world that we have a technology and capabilities that we didn't have before. And I like to say capabilities because I think sometimes when even I talk about large language

models and AI every single day, I think about the technology. But we also have to shift, I think, our collective thinking on what this means that we're now capable of, us as humans, as business leaders, that we weren't capable of before. And I think it's a lot. So on today's show, here's what we're going to be going over. I'm going to tell you what's new inside of GPT5 and why free users are actually the biggest winners. I'm going to reveal why the biggest GPT5 wins won't be felt inside of chat GPT itself.

I'm going to forecast how this release will shake up the AI landscape. And I'm going to give you seven obvious and not so obvious trends and facts that are more important than you'd think with this new GPT5 update. All right. Live stream audience. Great to see you all. Happy Friday. to you as well. Big Bogey Face joining on the YouTube. Bronson, everyone else. Brian and Jose,

thanks for joining us. Happy Friday from Chicago, Cecilia. Good to see you, Douglas, Monica, everyone else. If you do have any questions on GPT5, let me know. I'll start with this. Not a lot of people in the world have had access to this for very long. There are trusted testers that get access to it maybe a week or two before it comes out. For me personally, I've only been able to play with this for a couple of hours. Our live stream audience might be able to see. I'm not my normal kind of home office studio. I'm actually traveling.

So I have even less time to play with it than I would have normally if I was back in Chicago. So I have played with this for maybe about two or three hours since it came out. And let me just say this. It's really good. But let's get into all of the details before I start throwing out my opinion to you all. So first, one of the biggest differences, and I guess overall, this is an improvement. This is something I'm not particularly a fan of, but there's just model switching now. So it's this auto switching.

So you don't go in and choose a model. You're not going to go in and I believe that there was, what, nine models previously if you were on the highest tier plan. So now if you're a free or a basic user, you're going to see. really only one or two models. And essentially, GPD-5 automatically chooses between fast responses or deep-thinking modes. So this is very similar to how Google's Gemini 2.5 Pro works. The model itself essentially decides, hey, should I give you more of a fast topical answer?

Or should I take two, three, four, ten minutes and really think about this deeply? There are ways that you can kind of manually override. So if it's taking longer than you would like it and you're like, wait, I just asked you a very simple question. Why are you thinking for three minutes? There's a little button that says, you know, essentially give me a faster answer. And then on the other side, if it's giving you a faster answer and you're like, wait, normally I would choose a model that would take much longer to figure this out. With some simple prompt engineering, you know, you can, you know, just essentially tell it like, hey, take longer. Like take a lot of time to think about this.

You should be thinking about this analytically and critically, you know, go through these, multiple iterations in your processing, right? So you can still force it to think a little bit more. So at least for me, that's one of the biggest changes. Aside from, yes, the technologies are the better, the benchmarks, the, you know, the ELO scores. Yes, all of that's better. But the biggest thing I think, the biggest change is just this, it's one model now, right? Which we've known this for a couple of months when CEO Sam Altman announced this about six months ago.

So we knew that GPT5 was going to be more of a system. And we're not going to have to go in there and, you know, necessarily select from anywhere between six to nine models, depending what plan you're on. Here's the other thing. This is free for everyone. Literally, this is free for everyone. I have, I think, six or seven different Chad GPT accounts.

I think five of my accounts, including my free account all had access, or about five of the seven had access to GPT5. most people you should have it and again 700 million weekly chat gpt users have this and this is why it's big and i'm going to touch on this a little bit later gpt4 o was not a good model right and a lot of people didn't realize that when using the free version uh you had gpt 4 oh which is a model that i literally do not use and i haven't used it for a very long time and i would advise other other people not to use it. And that's not because it was a bad model in it of itself. That's just because compared to everything else that Open AI had prior to yesterday,

it was not a good model. Right. If you needed speed, I would say, all right, well, go use 04 mini, right, a thinking model. I use thinking models 95% of the time, even when working with chat EBT. So if I needed something quick, you know, I'm not going to use 03 or 03 pro. That may take two to 10 minutes, but I would use 04 mini, right, and get something in a couple of seconds. But that's why this is a huge jump because GPD 4-0, not a great model, at least compared to everything else that was out there. So the majority of people, even paid users, right, even people who weren't novices. So many people use GPT-40 when I don't think people really should have. I've said that on the show before.

I'm like, don't use GPT-4-O, right? If you need a fast answer, like I said, there's reasoning models that are a little faster. So now that the floor is through, I think, where the previous ceiling was. So that's why I can't understate how consequential this is to everyone's capabilities. Because I'm sure, you know, I can remember back to when GPT4O was the best model that OpenAI and offered inside of ChatGPT before the kind of reasoning craze before these O series models that can. and think and plan step by step and kind of have a little bit of human logic. So now that free users have GPD5, a model that technically it is thinking, it is a hybrid model, right, it can switch back and forth.

The floor, let me repeat that, the floor for a free user that maybe doesn't really know what they're doing, I think is higher than the ceiling was for the average novice users. And like I said, what that does from a capability standpoint? 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.adobie.com. I don't think most people are going to be able to grasp what this ultimately means. It is for logged in users, FYI. So I don't think, I didn't get a chance to read the entire model card, right? Because Open AI put out a ton of information, a ton of videos. I'm still making my way through it. I didn't see anything for non-logged-in users, FYI. So logged-in users have access to GPD-5, even on a free plan.

It didn't say what non-logged-in users have access to. So it might be one of their nano-GP-T-5 nano plans. So in the same way, a lot of people don't know this on the free plan, you know, because everyone's like, oh, chat GPT stinks. Well, okay, how are you using it? Number one, you're not logged in. Okay, so that means you were actually using GPT-40 nano, which is an even worse version of a not that great model.

So again, that's why I think we're going to see such a huge bump. And even people that were had a logged in plan who were previously using GPT40, after a certain number of messages, you just get knocked down to GPT40 mini or you did previously. So like I said, this is a daily driver upgrade that is hard to comprehend. It's like, you know, if you had the original iPhone, 15 years ago, however long it's been. And then you had the iPhone 17 Pro Max plus zenith, whatever it's called, right?

That's the kind of jump we're talking about from what you had access to yesterday to what you have access to now. The other thing, co-pilot. My gosh, talk about an instant glow-up for Microsoft co-pilot users because now millions of workers using Microsoft 3.3. 365 copilot, GitHub co-pilot, teams, you get GPD5 automatically on launch day. Again, assuming your IT company or your IT department can figure it out. This is the largest enterprise AI upgrade in history, just like that. A couple other and just kind of going over kind of an overview here of what's new and noteworthy vibe coding.

It's really good. I might show you depending on how long I asked. accidentally make this show here. I just cloned Slack in like 30 seconds this morning. Just taking a sip of water, sitting here on the couch. I'm like, let me, let me test this out, right? Really good. Can't clone Slack in 30 seconds with the old version of chat GPT.

Very impressive GPT5 on the vibe coding. And I think it's going to popularize vibe coding even more. All right. A couple other things to know. It is much more accurate on a critical topic. So this is according to Open AI, they said the medical question error rate dropped from 13% to 1.6%. Let me repeat that. The medical question error rate, according to OpenAI, dropped from 13% to 1.6%.

That's huge. In overall hallucination rates decreased by 45% with GPT5 compared to GPT4. So about half the hallucinations and much more factually accurate. And on the coding side, this is a very impressive coding model. And you may be thinking like, all right, Jordan, I don't care. I don't code. Well, you're going to be coding. Right.

If you listen to our 2025 AI kind of blueprint and predictions series, I talked about in 2025 non-technical people are going to be coding apps that they're going to use themselves. This is it. Right. Yes, it's been vibe coding has been, you know, very popular for a year plus. There's plenty of great tools. I love using personally.

I love using Google Gemini's Canvas to kind of vibe code little pieces of software that I use for very niche purposes. But this is bringing it to the masses here. Bringing it inside the chat. If you don't believe me, if you have a paid plan, go ahead and toggle on the thinking mode. I'm going to talk about there's three different modes technically. So yes, you don't have a model selector, but you have three different levels of reasoning that you can put out if you're on the highest plan. If you're on the medium plan, the $20 month plan, you have two, right? Go ahead and

switch it to the, what's called GPD5 thinking and just use canvas mode and just say clone blank app, clone Microsoft Word, right? Clone Google Sheets. And sit back and you'll be impressed, right? It might not one shot it every single time because generative AI is generative, but go back and forth and iterate and you're like, wait, what? This is possible. So like I can't underestimate, again, how many people are just going to be spinning up little useful and helpful apps with chat GPT that they can go in and use that's just going to make everyone's business lives a lot easier. And then reduce synchifancy. So, you know, like one of the.

most annoying updates to chat chb t that they actually rolled back essentially chat chitp t just started agreeing with everything you said this is a problem with all large language models right they're just yes man you can be like uh you know hey large language model i have an idea to uh quit my uh well-paying job and uh go live on an island and um sell sunshine to people on an island right it's dumb like that's absolutely dumb, right? But previous models would talk you through it and pump you up and be your hypeman. Oh, yes, that's great. Yeah, Jordan, you should do that, quit your good job.

Go sell sunshine. Everyone's going to buy it. You're the smartest person. You'll figure out how to sell sunshine. Like, in AI, I would tell you, oh, yeah, that's great. You know, hype you up for no reason. So GPD5 has cut down on that. So it said it's cut down on kind of those yes-man responses for,

from 14.5% to under 6%. So it's less likely to just automatically agree with users incorrect statements. That's the biggest thing, right? If you ask an AI model sometimes, if you say, hey, I'm thinking, you know, tell me why, you know, the Chicago Cubs are the best team in baseball this year. All right. Hey, although my Cubs are a little bit better than in the years past, they're not. But most large language models wouldn't correct you.

They're going to be like, oh, the Cubs are the best. team in baseball because A, B and C, right? They're just going to mirror whatever you tell them. So the fact that we got a pretty big response rate cut from 14.5 to 6%. Again, not ideal to still be at that 6% rate. That's a huge drop-down. And obviously previous models from Chachapiti and everyone else are criticized for just being too eager to please and sometimes at the expense of accuracy, the thing that is ultimately

the most important. All right. So like I said, even though, you know, you technically, GPD5 is more of an architecture, right? And it's a hybrid model and it decides, you know, how it's going to work. You still do have model choice. So I know that's confusing. But I think ultimately this is better because before, let's look at what we had before, right? You had GPT4.

You had GPT 4.1, which was more for coders. You had GPT 4.5, which was their EQ model, great for writing. Then you had GPT40 Mini, right? Then you had 03. You had 03 Pro. You had 04 Mini. You had 04 Mini. Right.

Confusing for 99% of users. I personally loved it. Right. I personally like going into a model selector in seeing 7, 8, 9 different choices. now it's going to be different. So if you're on the plus plan, so if you're paying $20 a month, you're going to see GPT5, it's going to say flagship model, and then you're going to see GPT5 thinking. So essentially it's just like, hey, think a little more.

That's it. If you're on the pro plan, the $200 a month plan, that's what I'm on. You're going to have three options. You're going to have GPT5, GPT5 thinking, and GPT5 pro. For the most part, think of it as just low, medium and high. right and that's when you can actually control without having to prompt engineer your way to it you can control how little or how much time chat gbt takes to essentially think or to build or to strategize etc whereas if you are on the free plan you're not going to have that option

you're just going to have gpt5 and when i say just it's just really really good uh so that's important to keep in mind uh to keep in mind uh y'all i feel If you do have a question, let me know. I'll try to tackle some of the questions at the end for our live stream audience. And FYI, it helps just because, you know, there's a lot of comments over here. Just put a question or something like that in front of your question. So as I'm scrolling through, a couple dozen comments and questions here from different platforms, it's easier to see.

So for paid users, you still do have a little bit of choice, right? You still have some agency left as an AI user, not as much as we had yesterday. It is still much simpler. And I, if I'm being honest, I've only been able to do simple prompts in testing the difference between GPD5, GPD5 thinking, and GPD5 pro, right? So what OpenAI said, you know, there's benchmarks and all that and, you know, X percentage of users, you know, the majority of users preferred pro responses versus GPD5, right? But I do still personally need a lot more time to even go into depth on these three different

models. I will tell you this. If you're using Canvas mode, right? And that's kind of the mode. A lot of people think of Canvas mode as like an interactive Google Docs, which it is. You can like write and collaborate in line with ChadGGBT. But what Canvas is obviously really good at, it's,

that's how you can render code. Right. So when I said go clone Slack, it used Canvas to then run and render that code. So you can chat with it on the left. Canvas, it's like think of it as a web browser, right? you can render the code that you write. Anyways, on the pro level, so on the highest tier, Canvas mode doesn't work. So aside from that, I haven't really seen any differences except it's going to take longer.

It's going to use more, you know, it's more inference, more computes to give you hopefully better answers. But like I said, you don't have the option to, you know, now choose, oh, I want a transformer model, right, which is essentially a very smart auto-complete versus a reasoning model, which thinks. Now it's just one inside GPT5 and it decides on its own if it's going to give you a spit it out in a quarter of a second or if it's going to take 10 minutes. All right.

But like I said, if it is starting to take longer and you're like, wait, I just need a quick answer. There actually is a button that just says, you know, give me a quick answer. So if it's, you know, if you're asking it something simple and you're like, wait, I think it misunderstood me or it's taking this way too far, right? And that's, that is kind of nice, right? because previously with 03 in 03 Pro, you know, you might think that you're asking it a fairly direct question.

Then it's going to take 30 seconds, but you still want that reasoning, right? And you're like, wait, five minutes later, it's still going. And there's nothing you can do about it. Right. So that is a pretty nice upgrade there. On the developer side, a lot of other options. We're not going to get to these too much. But from a pricing standpoint, very impressive.

And I'll be excited to see the full third party benchmarks with some of these other models. specifically I'm looking at mini. So from the API side, you have GPT5. So these are for companies building on top of the OpenAIs API. So you have GPT5, you have GPT5 Mini, and then we have GPT5 Nano. And I'm going to talk a little bit on the pricing structure on our seven big trends because that's huge. All right. So now I'm going to have a little bit.

I'm just going to read a little bit of this is from OpenAI's announcement. All right. So I kind of told you my. high level overview, what you need to know, but I think some important things in here from Open AI, so reading this off their announcement post. So they said, we are introducing GPT5, our best AI system yet. GPT5 is a significant leap in intelligence over all our previous models, featuring state-of-the-art performance across coding, math, health, visual perception, and more. It is a unified system that knows when to respond quickly and when to think longer to provide expert

level responses. GPD5 is available to all users with plus subscribers getting more usage and pro subscribers getting access to GPD5 Pro, a version with extended reasoning for even more comprehensive and accurate answers. So now a little bit more about this unified system from OpenAI. So they said GPD5 is a unified system with a smart efficient model that answers most questions, deeper reasoning model, GPD5 thinking for harder problems, and a real time router that quickly decides which to use based on conversation type complexity tool needs and your explicit intent. For example, if you say, think hard about this in the prompt.

The router is continuously trained on real signals, including when users switch models, preference rate for responses and measured correctness, improving over time. Once usage limits our reach, a mini version of each model handles remaining queries. In the near future, we plan to integrate these capabilities into a single model. All right. So let me talk a little bit about what this means. So like I said, when you talk to GPT5 now, it's going to decide essentially what of these three variations, if you're on the higher plan, which one it's going to use and how much essentially compute it's going to give to your query. And the good thing about Open AI having 700 million users,

you saw it right there. This router is going to hopefully get smarter. over time. So if you are using GBT5 today, number one, keep in mind it's going to be crowded, right? There are literally hundreds of millions of people probably over the course of the first couple of days that are going to be trying this all at once. So far, open AI servers have been pretty good. I've only had a couple errors come up. So pretty surprising. But there's going to be literally hundreds of millions of people trying to test this all out at once. So give it time, but also with that time, that router is going to hopefully improve. So it will hopefully know and get a little better at, hey, when I ask a simple query,

you don't got to think about it. All right. A little bit nor from OpenAI. So they said GPT5 in ChadGBT is our next flagship model and the new default for logged in users. This part's important. So it simplifies Chad GVT to a single auto switching system that brings together the best of the previous model into a smart past model. GPD 5 is available to all chat GPD users users on the paid tiers plus pro and team have access to that model picker like I told you about,

which enables you to manually select GPD 5 or GPD thinking. And then if you're on the pro or team, all right, that one's, I didn't actually know that one until I read this post because on my team plan, that's the one out of my seven plans that hasn't been upgraded yet. So on the team, you do have access to that GPT5 thinking pro. which takes a bit longer, but delivers the accuracy you need for complex tasks. And then a little bit more. So GPD5 not only outperforms previous models on benchmarks and answers questions more quickly, but most importantly, is more useful for real world queries.

That's the big thing. You know, I think that this is going to become a daily driver model for a lot of people, whereas even before, like myself, I might go out and use Claude for one specific thing. I might go use, I mean, I'm actually using Gemini 2.5. 5 Pro a lot, but I might go to different models for different purposes. I think with GPT5 and this kind of auto router, I think it is going to become a daily driver model for people that might usually use two or three. They also said we've made significant advances in reducing hallucinations, improving instruction

following in minimizing sycifancy, while leveling up GPT5's performance in three of chat GPT's most common uses, writing, coding, and health. All right. So now you know. big picture, you know the overview, you know what's new, right? I didn't go too much into kind of the benchmarks. I'll probably do that at a later point as I've had time to go through them, but I'll say this. It is the number one model on LM Arena across all major categories, right? Which usually doesn't happen. I think Google Gemini 2.5 Pro is the only other model ever that has held

that distinction. So what that means, this has been being tested blindly in the LM Arena, which is essentially the Pepsi versus Coke blind challenge. You put in a prompt. You have two models. You get two different responses. You pick what you want best.

That leads to something called an ELO score, like what is used in chest. So essentially, GPT5 is now across, and then it measures, you know, it classifies everything in categories. So now GBT5 is the number one model in the world on LM Arena across all categories. So the actual more scientific benchmarks, it's outbenching just about everything in every single category. And at least for initial testing on real world human preference, it's winning in every single category as well. And it's not really close. It has a sizable lead. But like I said, I wouldn't be surprised if we get a Gemini drop in days or weeks. But now let's get to the biggest trends that you

need to know. Some of these, I think, are under the radar. Some of them are obvious. Some of these I kind of mentioned a little bit, but I think they're worth reiterating. So number seven, free users are the biggest winner, period. Maybe for me, I'm seeing a, you know, 20 to 50% boost in my capabilities within chat, GBT. I'll say for free beginning users, I'm not kidding, you're seeing a 5x to 10x. All right, let me say that again. For someone like me, maybe 20 to 50%, you all, everyone else,

if you're a casual user, if you're on the free plan, huge jump. And I'll say actually, even casual users, right? If you're a GPT40 user, even on a paid plan, I can't even communicate how much better this base model is because we went from GBT40, which is based, again,

GPT4, the model itself, came out two and a half years ago, right? And then we went to GPT4 Turbo and then we went to GPT4O, but these are all just very good and sizable improvements on a model that was essentially released two and a half years ago, which in AI time is like, I don't know, that's like multiple generations, right? It was still a good-ish model. But if you were primarily a GPT40 user, even on a paid plan, right, that's a model that doesn't think. It doesn't reason. It doesn't logic.

It doesn't plan ahead. The difference between a model that does all of those things versus doesn't, it's hard to measure. So that's why I think free users here are the biggest winner. So like I said, yes, the ceiling went up for everyone, including power users. but the floor shot through the previous ceiling. It's huge. I'm talking about 10xing people's capabilities overnight. If you were a casual user,

and if you take the time to actually learn the capabilities, understand them, and make a model like this, your daily driver. And if you start using GPT5 in instances where maybe you wouldn't before, maybe you only used it for content writing and coding. Now, if you start using it as a strategist, a brainstorming partner, an editor, et cetera, I think you're going to see those, that capability jump that I talked about.

And that's why I think this, you know, this model router and, you know, a hybrid model will actually do wonders for people once you start moving more of your day-to-day processes inside of GPT5. So that's number seven. Number six, GPTs are now true PhDs in your pocket. So what do I mean by that?

So a while ago, OpenAI released GBT's. And at the time, there was a lot of hype. GPD is essentially a personalized version. You can create a personalized version of the big model, Chad GPD. Like if you're like, oh, I love this except, you know, I need it to always, you know, respond in my tone of voice and I always need to upload files, right? So GPTs are custom versions of chat GPD that you can make. You can make unlimited, unlimited versions of chat, dbtbt, upload your own. files, custom instructions, you know, which is essentially a system prompt.

But they kind of flopped, right? Because they didn't really get updated for like a year and a half. And they were only had access to older models. So if you tuned in, I think it was about six weeks ago. They upgraded that and they gave GPTs the ability to use the O series models, which is huge. So now obviously, GBT's, if you're on the paid plans, they have access to GPD, GPD 5, GPT thinking. And then if you're on the pro plan, you have access to GPT thinking pro.

So that's big, right? So you have these models now inside GPTs, but that's not the big thing. Now you also have advanced voice mode. A small little nugget from yesterday's hour and a half long, I think it was about an hour and a half long live stream that I don't think a lot of people picked up on is now inside of GPTs, you can use advanced. voice mode. So here's what that is. You've probably used advanced voice mode, right? It's very neural. It sounds like talking to a human. The latest update, which I'm if you on, right?

Sometimes it like chuckles or hesitate. So it's very much starting to sound more and more like talking with a human. So number one, they did update advanced voice mode, but they didn't go into a lot of details. They just essentially said it's better and more natural. But the other thing is, Now you can use advanced voice mode inside GPTs, whereas before it would have used standard voice mode, which, you know, aside from the fact, you know, it's, it was, you know, less like talking to a human. It was just slower and not as good. Think of it like a different model, like jumping from, you know, GBT4 to GPD5, right? So now you can create a custom version of chat GPT and GPTs, upload all your information and use advanced voice mode. That is big. That is something I've been waiting for for a very long.

time and I'm like when this happens this is now going to be everyone's going to have that personal strategist right you can upload every single piece of important every document data etc into a GPT give it custom instructions so it always knows you and knows what you're working on right and then you can just talk to it hey you know why are my podcast downloads you know up 20% this week and it can go look at everything right whatever I give it access to it can have access to my emails, my Google Drive files, you know, it can query the web. It can build, it can build things with Python, advanced data analysis, right? This is huge because now GVTs, I think, are true PhDs in your pocket. Number five, on our big trends. Open AI is going to

cause some chaos with their pricing. All right. So not only the fact that this is free, all right, GPD5 free. That's going to create a lot of chaos. They released, earlier this week, and I did a show on this yesterday, their GBTOSS, right, which is open source. You can download it, fine tune it, shut off the internet, and commercialize with it. All right. So not only GPD5 is free.

That's number one. Number two, you have GPTOSS. And then number three, the API prices are bonkers cheap. All right. I have a little comparison chart for our podcast audience. And if you're wondering, you can always watch the video version on our website at Your EverydayAI.com or on our YouTube channel. I know a lot of people listening on Spotify are like, where's the video version? That's where it is.

So let me talk about the pricing here in GBT5. And let me compare it to Claude Opus 4.1, which just came out and Claude Sondit 4. So that's the models from Anthropic. And I chose Anthropic because for whatever reason, their API pricing has just been ridiculously high. And now I think Anthropics in trouble. All right. Let's look at GPD 5 and compare it to Opus 4.1. So this is for developers on the API side.

So if you're building on top of their platforms, which is millions of developers, millions of companies. Okay. So GPT5 for output, it's $10 per million token versus Claude Opus 4.1, $75. It is 7.5x more expensive. And it's not better in any benchmark. It's like, I know like at first it was kind of like trendy and cool and hipstery to be like, oh, I'm using Claude. Yeah.

Oh, no. I code with Claude. Why? What's the reason? Yes, it's great on the agenic side as well. But I mean, everything that GPD5 is showing, it does. Doesn't make sense on paper to pay a 7.5x premium for something that is on paper and according to human preferences inferior. Right.

And then you look at on the input side. Same thing. GPD 5, $1.25 per million tokens on the input. Claude Opus $4.1.15. Doesn't make sense. I think Open AI this week is creating some chaos. I think some middle tier or what I call B tier companies are going to be in some trouble. All right, trend number four.

The corporate world just got a huge boost with Microsoft co-pilot. So I already mentioned this, but here's a new stat. 100 million plus. There are 100 million plus copilot users are going to get a ridiculous upgrade because there's a good chance that they were just using GBT4-O. Even though the O series models were available in Microsoft co-pilot, I talk to, I mean, I talk to people all the time, right? We do consulting in front-end strategy for people.

big companies, big organizations. I talked to huge enterprise companies that are Microsoft co-pilot companies that didn't even know that you could have used the O models inside of Microsoft co-pilot. So many people didn't know. So everyone's just using GPT4O, which again is not the greatest model. So to go from a non-thinking GPT4O and now hundreds of millions of people will have access to a more of a hybrid reasoning model that is much better. this is again can't understate how big of a win Microsoft just got from all of this and rolling it out into Microsoft Copilot Studio Microsoft Teams etc but that's if

IT teams can figure it out because if you're not paying for an actual license I I don't know this information off the top of my head so I don't know if those people who aren't paying for individual Microsoft Copilot licenses right because people still have access to go pilot without the license. There's certain instances of co-pilot that are free to use, but you're using an inferior model. So make sure if your company is paying the monthly copilot Microsoft 365 co-pilot seat, make sure that your organization, your employees, or you, if you are the employee, get access to GPT5 because it is available now and it is going to change how you work. Trend number three, the world's software and applications are going to get

way better. So like I said, both companies moving their API wrapping from GPD 40 to GPD 5 and the people actually building with it, everything you use, right? That's what I'm saying. Like your banking company, your CRM, right? Like literally, you're all software for the most part. If there's any smart capabilities in any software, they're using either OpenAI's API, Anthropics, API, Google's API.

So many of them are using open AIs, API. So now just hundreds, hundreds of pieces of enterprise software that billions of people use got instantly better overnight. As soon as the dev teams switch it over and go through their QA. So essentially, using the internet, not browsing internet websites, but using pieces of software is going to get exponentially better. Talking with customer service bots exponentially. better instantly. Number two, vibe coding is going even more mainstream. Sorry, that's the truth.

Right. So when Open AI talked about making the model router better because they have 700 million weekly active users, it's going to make the model itself better. But like I said, I think even the least technical users are now going to be vibe coding inside canvas, right? It's a little different. So like I said, I've been absolutely loving and using the heck out of Gemini 2.5 Canvas or Gemini 2.5 Pro Canvas. This was available. Canvas mode was available previously. I don't think it was that good for whatever reason. Now, GPD5, Canvas, really good. And you can literally create something, share it with teammates, use it

yourself, right? Let's say you're always, there's something in your, I always just say CRM, because I'm like, what's the piece of software that everyone knows most people use, right? A CRM. Let's just say you're using an old clunky CRM or there's certain features. You're like, my gosh, I have to work an hour every single day to do this one thing inside the CRM because it's not a feature. So I'm wasting thousands of hours, right? Or hundreds of hours a year, just doing this one thing because, you know, if there was a button here, I wouldn't have to do that, right? Now you can just go build that, build it for yourself, right? Assuming you can share the data between the two.

But here's how bullish I am on saying this. Well, you have to follow the leaders. Cursor is the vibe coding king. So Cursor, for the most part, relies on other people's models. Cursor replaced Clawed, right? which the majority of the coding and software development world said, hey, Claude is the best until GPD5 came along.

So they announced yesterday that now GPD5 is the default model inside of cursor. I mean, that's going to be a lot better for people paying, right? Your bills are going to go down. But when the vibe coding king, that's cursor, says, yeah, we're making GPD5 the default mode, that's huge. And that also means vibe coding.

It's here to stay. This is not a trend. And then number one, I think enterprise adoption will go exponentially faster. I think hundreds, hundreds of these, like I call them the, you know, the ink 5,000, right? But, you know, you go Fortune 100, Fortune 500, and then like Inc. 5,000. So let's just say the 5,000 biggest companies in the U.S. I think there was a lot of those companies that were still kind of on the fence with like full AI deployment. They've had AI projects, maybe like one or two teams, but they're like, hey, when it comes to bringing

AI to our entire organizations, a lot of stats have said that, you know, 96% of executives say that AI implementation is one of their biggest strategic priorities, but only 4% of companies have successfully scaled AI from top to bottom. One of the reasons is when you look at these problems, I mean, yeah, if it's agreeing, if models are agreeing with inaccurate information, if they're inactored information, if they're inaccurate in general. The hallucination rates are higher. And like I said, GPT40, hey, it was great a year ago.

It's not great today. I think this is going to push people off the fence. People are going to go from pilot programs to literally full deployment, not overnight, but essentially overnight when it comes to tech innovation, just based on GPT-5's reliability, its accuracy, and its speed at bringing this reasoning and logic to market. All right. I did see a couple of questions.

This has gone, got a little bit longer than I hoped, but this is a pretty sizable update. So let me see here. A couple questions, Sneha said, what's the best way to get started with vibe coding? Literally, just open up canvas mode inside GP5 right now

and say, build me a clone of blank, right? Whatever software you use and then iterate on it, right? I mean, I think the first thing is just say, build me a clone of blank. See how it works. See how you can iterate with something. Right. Then obviously, you know, I don't think that GPD5 is going to be a fully featured vibe

coding platform. Right. If you want to build like an actual business around something, you're probably better off using something like cursor or replet, at least for now, you know, something that handles the back end. But if it's for personal use, if it's for your team use, if it's for your team use, If it's just to learn, just literally go in and start. Let's see.

Question here. Jackie said, so paid users get more prompts, tokens. Yeah, you get more usage. I don't know if they gave specifics. I will have to go and double check that, but we'll make sure to put that in the newsletter. So yes, free users, the caps on GPT5 are going to be a little restrictive, especially early on. On the $20 a month, it's still limited. GP5 is limited.

If you're on the pro, it is unlimited GBT5. Douglas, question. If teams has pro thinking, what's the draw for the pro tier? Limits, yes, unlimited. Unlimited, unlimited GPD5. Yeah, so if you're on the teams or the plus $20 a month, teams, I think is $30 a month per user, your GPD5 use is limited. And then you're going to be rolled back to essentially a GPD5 mini.

So that's one of the reasons. But across, you know, codex, SORA, you know, those unlimited limits scale. as well. All right. I think those for the most part, the biggest questions. I'll try to go through the comments. If you do have some later, please ask them. Also, I'm going to be covering this as I get more time, probably doing a lot of hands-on. So let me know what do you want to see us testing next week or the week after with GPT-5. Like I said, we're not going to do a lot of hands-on live demos because I really like to spend minimum at least like 5, 10, 15 hours before I start doing those types of things. So regardless, I hope that this episode was helpful as we told you what's new,

kind of where everything is headed and seven big trends. If it was helpful, please tell someone about it. If you're listening on the podcast, please make sure to subscribe. And then, if you haven't already, please go to your everyday AI.com. Sign up for the free daily newsletter. We're to be recapping today's conversation and a whole lot more in there. Thank you for tuning in, hope to see you back 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.