← Everyday AI Podcast – An AI and ChatGPT Podcast
Everyday AI Podcast – An AI and ChatGPT Podcast
AI in 2026: 7 reasons why the pace of AI this year will far exceed 2025.
00:35:34
Loading summary
Transcript3 lines
- [00:01]
A
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.
- [00:17]
B
The gap between how most people use AI today and what it's actually capable of is mountainous. That's because many business leaders are driving by only looking in the rear view mirror. Many enterprises are primarily still using AI to polish their blog posts or as a Google search replacement. And that's one of the main reasons, I think 2025 wasn't quite the breakout year for AI that many thought it might be. But that's also because the technology in itself in the first half of the year needed a lot of elbow grease and duct tape. It's not the case in 2026. Today's models are so good you can sneeze and accidentally create a million dollar app. I mean, with little experience and a few clicks, you can send a swarm of capable agents to go accomplish real work for you. So why the change in outlook for 2026? Well, I've got five reasons. And on today's first edition of Hot take Tuesday for 2026, I'm not holding anything back. And we're going to be laying out the five reasons why I think the pace of AI adoption will far exceed what happened in 2025. All right, you ready to get into it? Hope you are. Let's go. If you're new here, welcome. My name is Jordan Wilson and welcome to Everyday AI. This thing is your daily livestream podcast and free daily newsletter helping everyday business leaders like you and me. Not just keep up with all the AI advancements because they happen every single day, but how we can make sense of it. No bs, no spin, just grab the most important aspects that can grow our company and our career. So starts here with the unedited, unscripted live stream podcast. But to be the smartest person in AI or your company, make sure to go to our website@your everydayai.com Sign up for the free daily newsletter and little tease here. Talked about this on the show yesterday, but we are launching our AI Inner circle community this week. We have our new prime prompt polish course done. It is self paced. It is ready. We've technically been working on this for like almost two years. So make sure to tune in to tomorrow's show and keep an eye out on the newsletter. All right, Brian, but let's get straight into it. I'm not going to make you wait any longer for the five reasons. So here is what they are. What I'm calling the five pillars of AI acceleration for 2026 that maybe weren't there in 2025. So number one is reasoning by default. So large language models, large are now a natural thought partner and they were just kind of a fun tool maybe last year for a lot of people. Number two is agentic scaffolding. So we've gone from passive knowledge to active execution with tools. Number three is frictionless, frictionless data integration. So rag pipelines. If I'm being honest, I'm bearish, right. I'm very bullish on what most front end large language models offer now is one click versions to bring your company's dynamic context to the front. Number four is exponential tax task endurance. So going from what large language models were capable of maybe a year, year and a half ago, short sprints to now long term projects and then number five, and this is last but definitely not least is economically meaningful work. Yeah, it's large language models are no longer about writing better blog posts or replacing Google. They are about doing the actual work that you're doing now, which I think in 2026, more than ever to get the most out of AI, it really requires a, a mindset shift. Right? And I think this became pretty evident to maybe a lot of people. I've been saying this for a very long time, but a twee has picked up a lot of steam in the last day. Came from the OpenAI CEO of Applications, Fiji Simo. So Fiji wrote a tweet and then a blog post. We'll make sure to link it in today's newsletter if you want to check it out. So she said Frontier AI is far more capable than how most people actually use it. And then in the blog post just kind of talked about how 2026, at least for OpenAI, is about closing that gap through better products, not just better mod. You know, she Talked about how ChatGPT is positioned to become a true personal super assistant. Proactive and personalized, connected to real services and focused on getting concrete tasks done. And then she said that for businesses, OpenAI aims to be the operating system for enterprise automation with agents that reliably handle real work at scale. All right, not trying to say I told you so, but I've been calling ChatGPT and other large language models the AI operating systems for multiple years now. Insane that even if you are a co pilot organization, you need to move all of your day to day meaningful work inside of a front end large language model and start using them as a team. All right, but hopefully today when I talk about the five big reasons why I think you'll start to understand. Yeah, we should probably start doing this if we haven't already. So let's look at reason number one, right? Why is the pace of AI going to be so much faster in 2026 than it was in 2025? And the first thing is, well, the models are better by default. So what's interesting to note is a stat that's not talked about a lot. Last year, Sam Altman said that only 7% of queries involved reasoning models. And that to me is absolutely nutty. All right, so if you are kind of newish or not super technical, let me explain very briefly. Kind of this big shift that happened in 2025 and it was a slow shift, right? Because maybe a lot of the, a lot of how people view large language models is the original ChatGPT, right? A friendly chatbot that's, know, really fun can spit out, you know, large blocks of text and sometimes hallucinates. And that's not really what large language models are today. I like to say that there's almost a line in the sand. And to me, that line, it's not between, you know, normal AI and agents. It's actually between, you know, kind of quote unquote, old school transformer models versus new reasoning models. And I think that this has turned large language models from something that humans have to put a lot of work into, into being a true agentic partner. Because the ability to reason is an absolute game changer. All right, so like I said, models from yesteryear, they were great. You know, nothing wrong with a GPT4O or a, you know, Gemini 2 or a Claude Sonnet 35, whatever model you want to throw out there, right? But they weren't that good, if I'm being honest. Right? What it required, you had to be a real dork like me, right? You had to really put in the work on prompt engineering. You had to really almost be obsessive to get human level output out of large language models in early 2025 or in 2024, right? For a lot of reasons. But one of them is just because these models didn't think they couldn't plan. They weren't, they didn't exhibit human traits, right? They just spit out as quickly as possible. Next token predictions, right? Reasoning models are much different. They think they plan, they can agentically start going down a certain path and then decide, oh, this isn't the right path. And they can then go backwards and start down a different path, much like a human would, right? Looking at, if you have never done this before, go into ChatGPT, use one of the thinking models, ask it a very tough question and read the summarized chain of thought and you'll see exactly what I'm talking about and why. This leap between non reasoning models and reasoning models is actually exponential, right? And it's crazy because it didn't really happen until kind of the middle of 2025, which is one of the reasons I still think people had this old original November 20viewpoint of Chat GPT when it came to AI implementation in the enterprise, which is a dangerous viewpoint to have. So as an example, right, when we talk about when did you know agentic or reasoning or hybrid models come to the forefront? Well, it wasn't until the middle of the year, right? So Google brought Gemini 2.5, their first hybrid reasoner in March of 2025. Anthropic brought Claude 3.7 Sonnet in February. And then it wasn't really until, you know, most chat, Most paid chat GBT users started to get the O3 model right around the same time. So it honestly wasn't even until, you know, that first quarter or middle of the year that people even got a taste of what reasoning models were. And I still don't think right, even just by looking at that 7% of users were actually using them. Right. I think as they become the default, right. I think that changes what companies think AI can do for them or should be doing for them. And reasoning models I think help close the gap that let non technical people get much better result from front end large language models because you don't have to, you know, do a bunch of, you know, advanced prompting to get the most out of them. All right, so let's now go to reason number two or pillar number two which is, well, they now have hands and feet to act. This is the agentic scaffolding that models have. So that's the technical term, right? Agentic scaffolding. But this is what I mean by that. Models now know when to enable tool calling. They can plan out their responses, when to write code or use computer vision without being told by the human, right? So many times I will send a prompt, right? Whether we're talking about Gemini 3 Pro, Claude 45 Opus, you know, Gemini 5 Or, or, or sorry, GPT 52 Thinking Pro, one of my favorite models, right. And I think that it's going to be straightforward and it's like, okay, it's going to reason, it's going to go do a little research on the web. It's going to look at my, my data and that's it. And then all of a sudden I'm like, oh, no, it's using computer vision and it's using Python and it's doing all these other things that maybe I didn't expect it to. And then when I look at it, I'm like, oh, wow, I'm getting much better results than I thought I would because the model took advantage of all this advanced scaffolding. So what does this mean? Well, the agenta capabilities are there by default in these reasoning models. So that is more or less the AI's ability to act as an agent by independently planning, you know, executing a series of steps to reach whatever goal it is. For example, you know, if you ask, you know, one of these models to plan a trip, it doesn't just look up flights, it automatically might check your calendar, it might look up live weather, it searches multiple booking sites at once and might pretend, you know, present a final itinerary to you in one go and create different charts and graphs that you didn't even know you needed. And you're like, wait, I actually really needed this. You know, and then on the scaffolding side, this is more of the kind of invisible support system that large language models have that's built around the models and it gives it the tools to be agentic. Right. So in the ChatGPT interface, for example, the scaffolding includes the model's ability to access a web browser code execution, know the memory of your previous chats, previous files. Right. So that just allows it to kind of step outside of its own brain to verify facts or run calculations, or to take multiple passes, you know, at a certain problem that you may be throwing at a large language model. And I do think the combination of these two things, it's taking large language models, I think, in my mind, from input output devices to problem solution. Right. I find myself, especially over the past 18 months, spending much, much, much time working with large language models, telling them more about problems and giving them data about problems and then investigating different solutions together. So I'm not using large language models very much anymore or as much anymore for simple input output. It is really thought partnership. All right, so let's go into reason number three. AI adoption in 2026 is going to far outpace 2025, and that is rag is one click now, right? Yeah. All right, I'm sure I'm going to get something from an engineer here and be like, no, Jordan, you're an idiot. You're wrong. All right, I'm simplifying here, I'm simplifying here as I always do for our more non technical audience. So let me talk a little bit about rag why it's important to understand, right? And you don't have to be an expert on, you know, vector databases or anything like that, right? The simple way to think of this is traditional rag, right? Very popular, you know, even before chat GBT came out. But you know, especially once Chat GPT came out, I think in 2023, early 2024, Rag was all the rage, right? Retrieval, augmented generation, to say simply, it was a way to insert your company's most important dynamic data in front of the model. So before it even went to its training data, essentially it would look at your databases first and then just kind of pass off the relevant invite insights to the large language model. I'm oversimplifying there, but hopefully that on a podcast that can make sense for you without showing you, you know, fancy graphs that I have on screen here, right? You needed that in 2021, 2022, 2023, not so much anymore, right? Everything is moving to the front end, right? Not everything, but for non technical, everyday knowledge workers, right? You're in hr, you're in finance, you're in marketing, you're in PR comms, you're an executive leadership whatever, right? You probably, if I'm being honest, you probably don't need to go pay someone, you know, six, seven, eight figures to go build a RAG pipeline. Because now with a couple of clicks and this is one of the biggest changes that I think was overlooked in 2025, you know, OpenAI I do think led the way. I think anthropic caught up in a huge way and Google through one of their Gemini people don't, I feel a lot of people don't even know that Gemini has two different versions. They have the version that probably most people use and then they have a business and enterprise version as well. That's completely different, right? So I have two completely different versions of Gemini that kind of do different things, right? But one I think has a little bit better way to ground your data, right? So when we talk about grounding, you know, that is a simplified way to talk about the benefits of having a RAG pipeline, right? So now whether you want to talk about things like, you know, connectors, you know, ChatGPT connectors, or now they're kind of leaning more into apps, right? Google calls them different things. Same thing with Anthropic, they have connectors or integrations, but you can bring your entire Google Drive, right? If you use box, if you, you know, your Outlook calendar, your, you know, your Gmail, where wherever it is that your company's data lives within literally two to three clicks, it can be indexed, waiting there, dynamic, and in certain cases, with a little bit of know how, not a lot, you know, you can get that grounded truth from large language models, right? The, the holy grail of what we thought we would never get right. These things are just spitting out, you know, more lies than a politician two weeks before reelection, right? No, not anymore. You know, having this ability to connect AI to business reality was a major bottleneck, but I don't think it is anymore. All right, let's keep going. Number four here is AI can work for a very long time, all right? And this is important for a lot of reasons. I think the easiest way to illustrate this point is to talk about coding and to specifically even talk about a kind of exact kind of stat here from meter, which is the model evaluation and threat research. So meter, so that's M E T r. They're essentially a nonprofit that tests if AI models are becoming powerful enough to be dangerous. Right? To simplify it. And they've been doing this for a while, and usually it's kind of through the lens of coding because that's a great way that you can kind of get this time horizon. So let me, you know, so instead of just giving, you know, different AI models a score, they measure a model's stamina for finishing long projects measured in human hours. So what they look at, it's called a 50% time horizon or the effective horizon. All right, so that's kind of what has been popularized by meter. So in short, it measures the length of a task that an AI can successfully complete at least half of the time. That is that 50% horizon, okay? And this is the human hours scale. So the task length is measured by how many hours it would take an expert human to finish the same job. So this isn't how long a model can work autonomously on its own. This is essentially, can a model, you know, by default, you know, do a task that would take a human an hour? Can they do a task that would take a human four hours, six hours? And can they do it, you know, 50% of the time? So what's interesting, and I'm going to go ahead and share something else on my stream, on my screen here for the live stream audience. So hopefully you guys can, can see this. What's interesting is I think the new model from Anthropic, at least on this meter metric completely changed, you know, what people think large language models are possible of now, right? So it's hard to describe to our podcast audience, but I'll say this hockey stick, right? Everything else was pretty linear, right? Looking at growth from, you know, GPT 3.5, GPT 4, right? And at this point, you know, even middle of last year or no, let's, let's go back about two years, right? So when you look at some of these, some of these models, some of the GBT models, they couldn't even do 30 minutes reliably, right? They couldn't do a task that would take a human 30 minutes. They couldn't do it reliably. You know, more than two years ago, 2025, we started to see some huge growth with reasoning models, with models that had that agentic scaffolding that I talked about, right? So you saw some pretty big jumps from 04 mini, from GPT5, right? And at that point you're getting into the one hour range, the two hour range. Then you saw GPT5 1, Codex, you know, two and a half hour, almost three hour range. And then Claude Opus 4.5 thought it was Gemini 3 Pro in the fact that it went nano bananas and shot straight up the charts and is now nearly at the five hour mark. So why is that important? Well, you have to look at the rate of growth and what's happening now is legit the definition of exponential, right? Because the time horizon previously was doubling every seven months on average going back to 2019. So now it's accelerated a little bit more in late 2024. So it went from seven months doubling to now it was doubling every four months. But now it is nuts. Right now it is nuts because as of late 2025, the top AI models are, like I said, well past the one hour mark. But if the current trend continues, if the current trend continues that we saw set with Claude Opus 4.5, that means that AI models as early as 2027, maybe 2028, we'll be able to do a month of human work and get it correct. Let me repeat that. Maybe as early as next year or 2028. Are you still running in circles trying to figure out how to actually grow your business with AI? Maybe your company has been tinkering with large language models for a year or more, but can't really get traction to find ROI on gen AI. Hey, this is Jordan Wilson, host of this very podcast. Companies like Adobe, Microsoft and Nvidia have partnered with us because they trust our expertise in educating the masses around generative AI to get ahead of. And some of the most innovative companies in the country hire us to help with their AI strategy and to train hundreds of their employees on how to use Gen AI. So whether you're looking for ChatGPT training for thousands or just need help building your front end AI strategy, you can partner with us too. Just like some of the biggest companies in the world do. Go to your everydayai.com partner to get in contact with our team or you can just click on the partner section of our website will help you stop running in those AI circles and help get your team ahead and build a straight path to ROI on Gen AI. AI models are going to be able to complete what would take a human a month. Okay. Yeah. And that hockey stick curve just hit in 2025. And it's, it's a lagging effect, right? So many of these things are a lagging effect. I am absolutely mind boggled at when I talk to smart people, right. All the time. I'm not saying on the show that you all get to hear. I. What's, what's crazy is, you know, I talk to other people outside of the show, right, About AI companies, maybe hire us, you know, maybe I'll jump on a call with people and sometimes I'm shocked, right? Smart people, growing companies don't even know. Oh, there's a difference with a reasoning model. Or they're using the free version of ChatGPT, not even knowing anymore. And they're like, oh yeah, well, you know, if I ask it a hard question, it'll route me to a thinking model. It's like, nope, OpenAI got rid of that. That router doesn't exist anymore. It sends you to a bad model. Right? And the majority of people, I think 95% of people are using the free version of ChatGPT. Literally. Enterprise companies that are making billions of dollars are using in some instances the free version of ChatGPT, which is I would not recommend. Right? So again, this is just another thing that signals just this huge gap, you know, going back to that Fiji quote from earlier, that tweet that she put out, just what models are capable of today versus what people are using them for. Right? They're still using the 2022 version of ChatGPT, not knowing you can go out and legit do hours of human quality work. All right, and that brings me to the last number. Five, the fifth. And I think maybe the most important reason why the pace of AI adoption in 2026 will far out exceed exceeding 2025. And the easiest way to say it is, well, AI is finally doing economically meaningful work in a big way and out of the box. Let me ask you a yes or no question, all right? And answer it silently. Live stream audience, you know, go ahead and put it in there. I'll. I'll do a little 10 second delay. All right? Can Chat GPT without any, you know, additional features or anything like that? Can it create a PowerPoint? Not the agent mode, the normal version of chat GPT? Yes or no, can it create a PowerPoint? All right, give everyone a second here. You at home, too, listening in the car, you know, walking your dog, whatever you're doing on the treadmill. On the treadmill with your dog. Question number two. Can ChatGPT without any additional bells and whistles, right? Can it create spreadsheets? I mean, these have historically been two things, at least when you look at Chat GPT. Well, and Google, Gemini and not as recently, Anthropic Squad. These have kind of been one of the Achilles heels, right? It's like, all right, well, I can get all this great content out of these large language models, but it puts in a chart and I got to copy and paste it and I got to go create a spreadsheet, right? Or this just gives me stuff to. Then I got to go build, you know, an outline or, you know, gives me an outline, but I still got to go create a PowerPoint. All right, if you were playing at home, yeah, Chat GPT by default can do those things now, right? Previously, it could only do an agent mode, and it was super slow and it really wasn't good. But the new model, the GPT5.2 thinking model, can create Excel spreadsheets by default. It can create PowerPoints that actually look pretty good with visuals. By default. Yeah, I was doing a little bit of testing over the weekend, just some kind of PowerPoint building between Claude4.5 opus and GPT5.2 with thinking mode. And they're both very good. Sorry, consultants. It's gonna be a tough 2026, because that's what consultants do. They go do a bunch of research, they make some spreadsheets and do PowerPoints, right? Yeah, but number five, yes. It's not just they can do this, it's not just, oh, my gosh, you know, cool Claude and, and ChatGPT can make decent PowerPoints and spreadsheets. It's not just that they can actually, on their own, complete, economically meaningful work at or above an expert level. All right, so I've talked about this once or twice on the show before. I should probably do an entire dedicated show on, on GDP val. Maybe I'll get the, the creator of the benchmark on the show at some point this year. All right, so this is a, in, in a benchmark that OpenAI created. And I've, I've said this once or twice before, I love when a frontier company makes a benchmark, they publish the results and they aren't the leader and it's one of their competitors. Right? Because that's exactly what happened when OpenAI announced GDP Val. Right. They were not the best model, Right? It was Claude Opus. All right, so here's what GDP VAL is. It stands for Gross Domestic Product Valued Evaluation. So essentially it measures real world economic deliverables instead of just, you know, random puzzles or, you know, multiple choice questions or something off, offline data sets, which is what a lot of these benchmarks do, right? You know, essentially benchmarks. To oversimplify things, it's either, you know, there's these kind of like acts for large language models, these offline tests, there's these complex, right. Kind of like ARC AGI. There's these complex like reasoning puzzles, you know, kind of like spotting patterns. You know, those are all cool benchmarks and all. But what do we use them for? What do we use large language models for? We use them to do economically valuable work. Right? Which is why I think this GDP VAL is such an important metric to look at. Right? But this is where models, it evaluates models ability to create actual work artifacts. Right? Legal briefs, engineering CAD blueprints, nursing care plans, financial spreadsheets, whatever. Right. Things that the economy requires. And it's a broad professional scope. So the benchmark consists of more than 1300 specialized tasks across 44 distinct occupations. Right? The in the nine, in the nine sectors that contribute most to the US's GDP. And then there's blind expert grading. So outputs, outputs are graded via blind pair pairwise comparison by human industry professionals who have an average of 14 years of experience. And then they judge whether the AI's work is superior, equal or inferior to a human expert's attempt. So it's kind of just like a blind, you know, a blind taste test on real work that all of us do, right? Literally all of us. And well, here's what they found. They found that the model GPT5.2, and I believe they use the thinking version, achieved a 74% win, tie rate. Right. Whereas the model from just a couple months prior, GPT5, that was kind of not well received, right? It got a 38% win tie rate. So it almost doubled its win tie rate of creating economically valuable work. So not only that, obviously, well, it's preferred to the expert human work with a 74% win tie rate, but it is obviously extremely efficient and it completed those economically valuable tasks 11 times faster at less than 1% of the cost. So if you need, number one, if you need a reason to invest more heavily in large language models in moving your day to day knowledge, work tasks inside of a front end large language model, there's, there's your answer. It does better than human experts or at least a win tie rate of 3/4. It is 11 times faster and it costs less than 1%. So if you're still thinking of large language models as that cute cheeky chat GPT, right, the fun little chatbot. Oh my gosh, look at this. It wrote a haiku. Oh, cute. Oh, look at this. It helped me seem less mad at an email. No, large language models, if you know what you're doing, they instantly connect to your business data. They can think like experts, they can use tools that experts sometimes might not even know how to use. They can work for hours on end and they're doing work better than human experts that is more economically valuable. So there you go. Buckle up. 2026 is going to be wild. So that is a quick recap on the five reasons why. So I hope this was helpful. If it was, please tell someone about it. All right? It's great that you listen. I love it, appreciate your support. But don't keep this thing your own secret in 2026. Right? I can only keep doing this thing for another 700 episodes if you tell someone about it. So please Repost this on LinkedIn if you find it helpful. If you're listening on the podcast, please leave us a rating. You know, follow the show. I'd really appreciate that. And keep an eye later this week, maybe tomorrow, maybe Thursday. We'll see. We're going to be launching the AI Inner Circle. It is a free community. We have a lot of special things planned aside from our, you know, updated prompt engineering course, which I just finished recording like 72 hours ago. It is hot, fresh off the presses, so make sure to keep your eyes and ears peeled. And you got to do that at our newsletter@your everydayai.com. so thank you for tuning in. Hope to see you back tomorrow and every day for more Everyday AI. Thanks y'. All.
- [35:14]
A
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.