Everyday AI Podcast – An AI and ChatGPT Podcast - Ep 713: Company AI Brains, No More Code, Slop Debt Kills internet and Agent Societies. 2026 AI Predictions and Roadmap Series: Vol 2 of 2

Episode Date: February 13, 2026

AI Slop is gonna kill the internet. Companies are gonna have agent brains. And if you travel down the same path in 2026 as you did in 2025, you're toast shorty. I spend thousands of hours each year working in and around AI. And once a year, we do our predictions and roadmap series. It's a literal cheat code to skip through the 95% of B.S and get the plan for the 5% that moves the needle. Join us for Part 2 of our AI Predictions and Roadmap series. Company AI Brains, No More Code, Slop Debt Kills internet and Agent Societies. 2026 AI Predictions and Roadmap Series Part 2 of 2 -- An Everyday AI Chat with Jordan WilsonNewsletter: Sign up for our free daily newsletterMore on this Episode: Episode PageJoin the discussion on LinkedIn: Thoughts on this? Join the convo on LinkedIn and connect with other AI leaders.Upcoming Episodes: Check out the upcoming Everyday AI Livestream lineupWebsite: YourEverydayAI.comEmail The Show: [e-mail] with Jordan on LinkedInTopics Covered in This Episode:OpenAI Consumer Hardware Predictions 2026Disposable Software Adoption in EnterprisesNotebookLM as Fifth Core AI PlatformAI Native Ads Maintain Premium PricingMulti-Agent Societies: Enterprise Default ArchitectureMicrosoft Copilot Usability Reset ForecastBig Four Consulting AI-Driven RestructuringVibe Coding Rebranded as Agentic OrchestrationProfessional Services Launch AI Flanker BrandsSlop Debt Crisis and LLM Data IntegrityFrontier Labs: Humans Rarely Write CodePortable Context Engines Replace Prompt LibrariesGDP-Val Benchmark Scores Surpass 80%2026 AI Roadmap: Unlearning and RebuildingTimestamps:00:00 "AI Insights and Actionable Steps"07:40 OpenAI's Lead and Growing Race11:50 "Disposable Software and AI Duct Tape"17:26 "NotebookLM: Game-Changing AI Tool"22:47 Advertising Evolution and Faster ROI29:11 Domino's Strategy & Copilot Reset35:22 "Consulting vs. AI Efficiency"42:54 AI Transparency Demands Rising48:29 "AI Slop Debt in Training Data"51:34 "AI Coding Revolution Predicted"54:57 "Portable Context Engine in AI"01:00:51 "Unlearn and Rebuild with AI"Keywords: 2026 AI predictions, AI shortcut, AI trends, AI roadmap, OpenAI, consumer hardwSend Everyday AI and Jordan a text message. (We can't reply back unless you leave contact info) Start Here ▶️Not sure where to start when it comes to AI? Start with our Start Here Series. You can listen to the first drop -- Episode 691 -- or get free access to our Inner Cricle community and all episodes: StartHereSeries.com Also, here's a link to the entire series on a Spotify playlist. 

This is the Everyday AI Show, the everyday podcast where we simplify AI and bring its power to your fingertips. Listen daily for practical advice to boost your career, business, and everyday life. Meet Firefly AI Assistant, now live in Adobe Firefly, the All In One Creative AI Studio. Just describe what you want to create and the assistant handles the rest, orchestrating multi-step workflows across Photoshop, Premiere Express, and more in one conversational interface. You direct the outcome. The assistant accelerates execution. You don't have thousands of hours a year to devote to staying on top of AI developments.

But you need that much time. That's because you've got to follow everything, but 95% of the AI movement is sheer distraction. But that 5% is a force multiplier for your work. But to know the difference, you've got to be an AI news hawk and you can't do anything else except stayed glue to the AI developments. Or you could listen to 700 plus everyday AI podcasts and go read every single newsletter. Or crazy thought, you'd just listen to our 2026 AI predictions and roadmap series.

That's because between today's show and yesterdays, we're not only condensing down thousands of hours of AI insights, but we're projecting the trends on what's coming next based on all of those hours. And we're also building you the roadmap as part of that AI predictions and roadmap series. So today we're kicking off part two of the biggest AI shortcut there is. So make sure you check out part one from yesterday. All right. Enough. Let's get into it.

Welcome to Everyday AI. My name's Jordan Wilson. If you are brand new here, well, we've been doing this thing every single day. for three plus years now. And it's your daily guide to stay on top of what matters in AI with our unedited, unscripted daily, live stream podcast and free daily newsletter. So it starts here with the podcast. But if you want the actionable next steps, make sure you go sign up for the free daily newsletter at your everyday AI.com. We're going to be recapping today's show and giving you everything else you need to know to stay in the loop. Speaking of everything you need to know, well, if you listen to

our show all the time, you know, I'm lucky enough to get to talk to hundreds of the smartest people in AI. And what I've realized after a lot of conversations, right, 700 plus podcast episodes is, well, I'm able to kind of grab these nuggets of wisdom from a lot of very smart people that are building AI. And I obviously have hundreds of conversations every single year that aren't on the show. And I've realized I'm able to kind of connect these dots, right, by just stealing these insights from the smartest people in the world. And then I have time to reflect. And I'm like, oh, wait. So throughout the year, I always have a working notes file for this very show. So I literally plan for this thing every year. I'm already planning for next years.

So just keep that in mind. That's where this comes from. And I always audit myself, too. Because sometimes I have these crazy predictions and sometimes they mostly end up coming true. So if you want to go check on my audit work from last year, make sure you go check out episodes 674 and 676. That's when I did a kind of an audit or a rewind on my 2025 AI predictions. And it's actually, aside from, wow, I was very much on target, even though they were kind of crazy predictions. Aside from that, it's a great catch-up to get, you know, an entire year of AI developments

in a couple of minutes. All right. And then like I said, make sure you go check out yesterday's episode, Volume 1. There we covered a lot. So make sure you go check that out. And also, you're probably going to want to repost this on LinkedIn, just saying. So if you are listening on the podcast, make sure to check your show notes. We always have a link to the LinkedIn live stream.

So go repost this because I started with, I forgot the number. It was like 150 or 200 different AI predictions. And I boiled it down to like my top 53 or something like that. So it's 2026. So I'm sharing 26 of them, but if you want them all, just go repost this show on LinkedIn. I have an incredible guide with even more insights on the ones that I'm sharing. I have more insights on the 26 that I'm sharing that I don't even have time to get to. So that is in our bonus guide, the 2026 AI predictions and roadmap series bonus guide.

Go repost that and I'll send it to you. All right. Let's get straight to the predictions. These are in no particular order, FYI. So let's, like I said, yesterday we did one through 13. Today we're doing 14 through 26. So number 14, open AI ships no consumer hardware this year. Number 15, disposable software becomes routine practice. 16, notebook L.M becomes the fifth core AI platform. 17 AI native ads maintain premium intent pricing. 18, multi-agent societies become enterprise default architecture. 19. Microsoft launches a co-pilot. Works Now, Reset Campaign. 20, Big 4 Consulting announces an AI-driven restructuring. 21, vibe coding rebrands as agentic software orchestration or something similar.

22 professional services launch AI Flanker brands. And at 23, the slop debt crisis makes some large language models unusable. Spicy. Oh, there's more. It's not 2023. I'm tired. 24 Frontier Lab. A Frontier Lab publicly declares humans rarely write code.

25 portable context engines, replace prompt libraries. And 26, we're going to see GDP VAL scores that cross the 80% threshold. All right. Fun stuff. Let's get into it. Number 14. Open AI is shipping nothing when it comes to hardware. All right.

So there's been a lot of buzz and a lot of rumors on this third device, right, that OpenAI is reportedly working on. And we've seen reports over the past year. Seemingly, they've nabbed more than a dozen Apple hardware leaders. You know, they have Johnny Ives, you know, I.O. company, a design company. But I don't think we're going to see any of it. And I think one of the reasons, well, it's the code red that. happened in the fourth quarter of 2025, right? Reportedly, Open AI put out this code red, right, that they said, wait, Google is starting

to eat everyone's lunch. We got to get our stuff together. And granted, on the software side, on the AI side, Open AI has been shipping a lot. And I actually think one of my predictions I didn't get to is on Atlas. I think Atlas is actually going to be a standout product for OpenAI. It's one of the few products that they are updating. routinely, both Atlas and Codex, right? Like poor agent mode doesn't get any love.

GPTs are ignored, you know, not actually ignored, but I'm saying, you know, they have a lot going on on the software side. But it was all of these, I think, hardware, you know, endeavors with this third device, right? It's a laptop, it's a phone, and then whatever Open AI, right, says that we need, right, whether it's a hockey puck, we stick in our pockets or a pen or whatever. But I don't think we're going to see any hardware. I think that their competitive leverage is going to shift back to just competing on model capabilities and distribution. And on the distribution side, that's where they've continued to keep, I won't say an untouchable lead, but a sizable lead over Anthropic, Google, and Microsoft.

And on the model capability side, not quite the same story there, right? because if you went back 18 months, no one was touching Open AI. Now, you know, it's a race, right? I think at any point, like I said on yesterday's show, Google is going to, what's actually funny, that prediction didn't take long, FYI. It was literally about five hours after that show

when I said at any point Google could, you know, because at the time, I say at the time like it was years ago, but at the time of the show, they did not have the top model on most text benchmarks. And then five, literally five hours after the show, they released a new version at Gemini 3 Pro Deep Think, right? And it swept, you know, all of the important stats. So when I say Google at any time, they want to, they can come in on the benchmark side,

on the model capability side and be the leader. They will. But I do still think it's always going to be 1A and 1B between at least when it comes to general purpose between Google and Open AI. But I don't think, I think that some of these other endeavors, right, all these financing deals that, you know, kind of distracted, all of the, you know, kind of the NBA transactions of, you know, different people going different places.

I think it all became maybe too much for Open AI. And it seems like they've re-centered themselves. So I know a lot of people are looking forward to something from the hardware side, but hardware has brutally low margins. right and a lot of the you know hardware AI plays that came out in 2024 and 2025 were a disaster and maybe that's a signal that the consumer industry isn't quite ready for that at scale right and you know maybe they want to be the electricity company and not the toaster company right next one 15 disposable software becomes a routine practice so here's the prediction in 2026

enterprises are going to regularly build and then discard short-lived applications created for specific one-time tasks, right? I think teams are just going to create full, fully working software tools for certain campaigns, for certain migrations. But I think the life cycle for some software is going to go from years to weeks, right? I think that governance is going to have. have to adapt to that kind of sprawl as well. But I think these disposable patterns are going to become standardized, right?

I think, let me tell you this. I don't think people are going to believe me. But I've always been a software nut. I've used, and I'm not exaggerating when I say this, I've used thousands with an S, thousands of pieces of software. At all times now, for the most part, when I'm doing this show, I have software being written for me. Right. Like right now, I literally, let me see if I, if I have both Codex and Claude Code. Yeah, I have Codex and Claude Code right now writing software for me. At this point

for me, right now, I don't think it's disposable for me. I think it will end up being semi-disposable. But think of how a lot of software has worked traditionally. Sometimes you spend so much time scoping it. You sign up for all these trials. And then you find like, ha, you know, you end up making concessions. And you're like, okay, well, we actually just needed these three features, but you got to pay, you know, $500 a month. And you get all these other things that we don't necessarily want. So then you spend time saying like, okay, well, what's going to be worth it? We have to buy it. Even though we only need three out of the 10 features that everyone's offering. So we got to buy the other seven. So which of the other seven can we put in a place?

We got to test it out, right? Versus now. I mean, my gosh, I've been talking about this now, for two minutes, you could have already built a simple version of at least one of those three features that you need. So I do think that we are going to be coming into this disposable software series. And whatever anyone says, right, I've seen some things on the internet pick up steam lately, right? Everyone's like, oh, you know, vibe coding's great, but, you know, you end up, you know, going back and forth, you know, with whatever models for, you know, hours or days, I mean, if you're trying to build, you know, a billion dollar company, like, of course, if you need something that solves an annoying task that you do over and over, or if you're just trying to, you know, to use some terminology that I go to sometimes, right, duct tape.

I say sometimes humans are the duct tape between AI, right? But you can build duct tape disposable apps with AI fairly quickly. It's funny, I typed into Codex earlier today, an idea for an app. And then I went upstairs and got a drink. I came down, it was done, right? Was it perfect? No. Was it working?

Yes. It was something I was paying, you know, $10 a month for. So, you know, I do think people, when we talk about vibe coded software, the de facto response to everyone in the enterprises, you know, everyone's like, oh, well, well, well, you think your company is going to, you know, vibe code Salesforce? It's like, no, absolutely not, right? But think of maybe how many, you know,

Salesforce adjacent products people might have, right? Or, you know, plugins that people pay for, right? I remember paying a lot of money for different plugins that work alongside different types of software. I think it's things like that or, you know, pieces of software that you used to use that are no longer supported, like I said. or, hey, we have these five, again, five pieces of software.

This is, I think, is going to have the real enterprise utility in the long run. When you have expensive software that you're actually using multiple pieces, and you're going to create, you know, a version of just what you need, right? If you're only using 20%, but you're overpaying the 80% for five pieces at that point, it is worth, you know, putting a couple months of vibe coded development. if it ends up safe, right? Especially if you're paying per seat in it, right? A lot of people, their software debt,

so not their software utility, what they're actually getting out of it, but their software debt is in the millions of dollars. So yes, disposable software is going to become a thing. I think enterprise vibe-coded software will still happen. It's already happening, but I don't think that's going to be a norm. But I think if you're listening to this show,

come talk to me next year. you're going to have at least a couple pieces, right? If you're already mildly technical and you listen to this show every day, you're going to have plenty of disposable software. You're like, oh, yeah, I use this for a couple days, right? I have some sometimes that I just use once and then I'm done. Because, you know, it's like, oh, I put in a certain, you know, model or, you know, it's a process that I know I have to do once for a big project.

It's just like, okay, I'll, you know, use like an opal or, you know, something easy like that, right? And it's like, okay, it's done already. Right. And I think especially now with some of the new Gemini models and, you know, the Gemini CLI with quad code on the desktop with codex, it's just too easy not to. All right. 16.

Notebook L.M becomes the fifth core AI platform. This might be one of the more random and kind of out there. But I think, okay, so obviously when I say big core platforms, I'm not talking about back-end APIs. I'm talking about front end large language models. So for the most part, right, that's Microsoft co-pilot, chat GPT, Anthropic Claude, and Google Gemini. So, yeah, I'm not an idiot. I understand that Nobogelam is under Google and it is powered by Gemini,

but it is a completely different product. And I think, right, there's been times, right? if you've been a Google user for a long time, you know Google comes out with a lot. There's a lot out there on the Google AI side that you've probably never even heard of. Go look at like go check out Google's AI labs. It's insane.

So good. Right. But a lot of stuff just well, you know, they'll release it and then that's done. Right. That's not how No Booking LA is. Their team is cooking. Right.

I'm talking to a couple members of their teams. specifically of the notebook lm team just you know on dms in different places and just you know getting some more insight and intel in terms of what they're working on and just how hard they're cooking on these products only this isn't a a side project for a couple of people this i think is has been one of google's main drivers uh when it comes to gaining some of that market share um from open AI. It's been notebook L.M. And I do think that from an outsider's perspective, I think notebook LM is going to come into the conversation where it becomes a verb, right? Oh, you better notebook that. I think that's where we're going to be. And people, there's going to be people,

I think, in companies specifically with some things that are coming out. Okay. There's going to be companies that are using notebook LM that they're not even going to know it's Google. And then they're going to see this, oh, direct integration, you know, this, you know, notebook and Google Gemini kind of kind of backward compatibility. And they're going to be like, oh, that's real cool, right? I think it's got to become that big and that well known and that useful and still, which is crazy to say, that's separate from Google Gemini, right? Some of the best features I think of any AI, any consumer AI we've ever seen have come from

Notebook L.M, the AI audio overviews and the deep dives. the notebook L.M with their nanobanana slides, the video overviews, right? The fact that you can dump in hundreds of sources, thousands of pages, and it will create a custom video with a voice with amazing graphics and illustrations, and it's free, and it takes a couple of minutes, right? I think especially with people's attention spans, right? Like, you got to call it what it is. I was actually one of my podcast episodes this year talking to a chief evangelist,

Richard from Google, you know, he had this little line that stuck with me. He said demos over memos, right? I think memos in general is they're going to die, right? And I think, unfortunately, as a former journalist, this hurts me to say the power of just the written word, I think is losing traction. And what do most large language models excel at, right? Aside from code, right, it's the written word. Ultimately, I think what people care about, what is going to engage them is what we're seeing at a notebook LM, which I think is going to be one of the key factors that leads to its rise.

That is the multi-modality, the personalization, customization at scale, right, to consume information in a way that feels very personal and relatable, but also extremely high quality. All right. Next, 17, AI Native ads are going to maintain their premium intent pricing. So here is the prediction on that. So I'm going to say, well, through 2026 but also 2027, ads embedded in AI assistance are going to command structurally higher CPMs due to explicit user intent. This one, y'all, yeah, going to be big. So the key player here is obviously Chat ChbT. They just started testing publicly testing ads this week.

So it's very fresh. But again, going into my background, right, I've been in different MARTEC comm positions over the last 20 years. But I remember working on Google Ads campaigns, what, like 15, 16, years ago? What's going to be possible from a brand perspective? It is. 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.com. Almost unfathomable, right? I mean, we'll see. I think Open AI has a lot of work to do to give the type of advertising tools that you have from like Google ads, right? But the level of intent that brands and advertisers will be able to tap into, right? Because, I mean, you spend, big companies spend millions of dollars a month burning it just to learn more about who is clicking their ads. who is buying their products, right?

That is going to flip. That is going to flip. And so what happens? You're not going to have to, you know, show an ad to a thousand people that you think are interested just to get the 50 people that are interested, right? Because it's going to be completely different because now those 50 people, chat GPT is going to know instantly because people tell chat ChabitT, everything right everyone treats chat gpti like their personal life coach their therapist their

business advisor all those things you don't treat google search like that right you don't treat uh facebook like that or uh matter or whatever the kids are on this day ticto snap chat snapchat still around um you share psychological intent with a chat gpt you share key words with a Google search. And I think what that is going to be is a huge premium that companies are going to end up paying to chat GBT. I think in the in the, or sorry, to open AI, I think in the early time, it's the CPMs, right, the cost per milly or right, essentially the cost that you pay per impression. That's how open AI is going to be doing it first by impression, not through, you know, different attribution going all the way through and checking.

out and buying something, although I think that they will offer that in the long run. They have to, I think, get some data on that first. But I think people are going to end up paying early, maybe double of what you might be paying from other platforms just from an impressions standpoint. But I think in the long run, when they're able to offer more than impressions, people are going to pay a lot more and they're going to be happy with it. They're going to be extremely happy with it because it is going to allow brands to go market so much faster, right? Startups, new companies, companies expanding into new markets with new

service offerings, you're going to be able to do it like that. You're not like, you know, there's always been this, you know, advertising funnel. You have to have your, you know, your cold, you know, you're cold, you're prospecting, your medium, you know, your hot buyers, right? It's not going to be like that anymore. You're just going to have buyers instantly and you're not going to have to build all these funnels because, well, chat GPT is going to be the new funnel and it is going to command premium pricing. Number 18. multi-agent societies become the enterprise default architecture. So the prediction is by the fourth quarter of 2026, all right,

given myself a little bit of time. Most serious AI deployments at the enterprise level are going to involve coordinated teams of specialized agents rather than just single, you know, single assistance. So here's what it's going to look like. One agent's going to plan. another will execute. A third will verify. And this essentially mirrors how human teams are already functioning, right?

And I think that's one of the reasons why, you know, agents didn't really take off in 2025 because people were looking at a single agent. But I think agents are going to be societies. Agents are going to be sub-agents. Like we talked about on yesterday's show, there's going to be agent-to-agent commerce. There's going to be agents that are negotiating with other agents. without real human oversight, right? But I do think that eventually,

at least for forward-leaning enterprises, not saying every enterprise, but AI-native forward-leaning enterprises are going to be doing this. Right? So a user-facing example might be something like, you know, build me a quarterly sales report. And then behind the scenes,

the planner agent breaks it down. You know, they grab your data from Salesforce, a chart from tableau, you know, two-paragraph summary. They assign subtas, then the executor does them, then a verifier checks numbers before anything. You know, then a team of design agents take it over from there. Then they say, okay, we need to design a deck. But in the deck, we need a couple charts.

Let's give that to the data visualization agents, right? That's how it's going to be. It is going to be a society of agents that work together. And like I said on yesterday's show, it's going to. going to be the, I think future of work is going to be the burger. Humans are the buns where the front end, the back end, but the juicy stuff, the real meat of the work is going to be the agents or society of agents, right? We are the plain kind of boring bun. We give the tasks, we check it on the back end, we hold it together, right? But the actual meat of what is going to

get done, someone start using that. I just made that up, I think yesterday and then I took it a little bit further today. Yeah. Can that be the next vibe coding, right? It's, I don't know, the, the Asian burger. But that's what it's going to be. And I think that you've already started to see this, right? If you've used in the last week, because a lot of this is new, right?

If you've used the new Claude Opus 4.6 in Claude code, which I have doing now, that's what's happening already. Sometimes the quote unquote agent you're talking to will do something on its own. Sometimes if you give it a much more robust task, it's going to assign a team of agents. One thing I'm playing around with right now is how much leeway I have controlling those teams of subagents. I'm still trying to understand from a prompt engineering perspective. It's weird because I think prompt engineering is starting to make a comeback strangely enough, not, right?

Not like what I talked about earlier, about how context engineering is replacing prompt engineering. I think that's just for, you know, work, you know, tax inside large language models. I don't think that's changing. But I think prompt engineering, at least when it comes to multi-agent societies, it's going to have a short-lived run here, right? Because what I'm finding is, you know, some different prompting techniques. But some definite prompting techniques that we're sticky in 2020 through 2023,

you know, we might get some added juice out of using those. All right. So multi-agent societies, enterprise default. No longer one agent. 19. Ooh, this one, a little spicy. Got to take a drink. I think Microsoft is going to launch some sort of co-pilot Works Now reset campaign.

So here is the. prediction. So in 2026, Microsoft will publicly reposition co-pilot around reliability and governance and real workflows rather than just, you know, broad AI, you know, AI everywhere. So copilot's early positioning, I think centered on it being everywhere, like everywhere. And I think the next phase is going to focus on performance in specific use cases. I think Microsoft's going highlight, you know, policy controls, audit logs, orchestration clarity, but I think they're actually going to do like an about face. All right. I don't have any insider intel on this, but I go back and think, you know, a fun example I used to talk about a little bit more, you know,

when I did did a little bit more, you know, brand work back in the day. Domino's, right? One of the most successful rebrands of all time. And what Domino's did, they ran a very famous commercial where employees and executives, you know, they essentially, you know, looked in on these focus groups and, you know, someone's like, oh, the pizza tastes like cardboard, right? And Domino's paid millions of dollars to air these types of commercials everywhere that said their pizza tasted like cardboard, right? Then they essentially, it was a very public about face. It was a public reset campaign.

I don't think we're going to see that level of it from co-pilot, obviously, because that would cause Microsoft to lose, I don't know, like $500 billion in market cap overnight. But I do think we're going to see a softer but somewhat similar version of that of like co-pilot just works now. Microsoft knows one. one of their biggest climbs. And I think why a lot of enterprises, even if they're still paying,

you know, Microsoft 365 co-pilot seats, they're paying less. They're using it less. Utilization for a lot of companies I talk to is going down because it's getting easier and easier for enterprises to use ChatGBTGBT Enterprise, to use, right? A lot of people don't even realize this.

Google Gemini has a completely separate business and enterprise product. And it's really good, right? It's different than the Gemini, you know, many of us, they just came out with it. I don't know, like six months ago. So it's not like been there forever. It's new. You know, Anthropic, they have a great enterprise product as well. I think Microsoft's been losing a lot of enterprise customers.

They were obviously first. There was no other options. You know, it was the enterprise option. Or if you wanted, you know, you either. had to, you know, rag it, you know, build your, build your pipeline, spend, you know, six, seven, eight figures. Or you had to say, all right, well, we're just going to go on this on, you know, this consumer large language model because there was no choice, really, until 2025, you know, when Open AI got serious about the enterprise, so did Claude. And then, like I said,

late in 2025, or sorry, 2024. And then in 2025, Googled it as well. But we've seen a lot of reports out of Microsoft, right, where it's CEO, Sadia Nadella, reportedly said like, hey, if we don't change, we're going to get gobbled up, right? One of the biggest companies in the world said, if we don't change, we're going to get gobbled up. And again, according to reports, the CEO, Sadia Nadella, is in a somewhat product management role, not actually, right? But he's getting his hands on the co-pilot product, which is not normal.

is not normal for a CEO from one of the world's largest companies to get involved in product. And he is, which is one of the reasons why I think Microsoft might come out with some sort of marketing messaging about how co-pilot just works now. It's easier, right? I've talked to literally countless, countless, countless, countless people. Even when I'm at Microsoft conferences, I'm always, you know, just fly in the wall talking to people. And so many people are just like, hey, like when co-pilot works, it's great. But the majority of our people can't get it to work.

They don't know where it is. They don't have access, right? One thing about chat GBT, you know where it is. It's easy to get access, right? You go to chatgbt.com. If you have the enterprise version, you have instant access to everything. Copilot is a little hard, right? Because you can do the same thing in like nine different places.

and you could complete the same task in nine different places, but your team might only be in three of those places and another three of those places. Maybe you don't have access to all the folders that you do in other places, right? It can be a sticky spider web of access. So I do think Microsoft maybe simplifies copilot a little bit in terms of usability and reliability and just simplifies its messaging on it just works now. We'll see.

Actually, right, this is one of those like swing for the fences because no one's, no one's saying this. It is kind of random. But once I saw that story of Sadie and Adela working on co-pilot, I'm like, okay, how are they going to respond? Something's going to change. We also know that Microsoft is, you know, working on their own models now, now that, you know, their agreement with open AI has changed a little bit.

So they're going to be pushing out their frontier models as well, which I think is only going to, well, it's going to help them in the long run. Obviously, Microsoft has so many other things going on than just Microsoft copilot, even on the AI side. But I do think that we are going to see some sort of about face effort from Microsoft, essentially admitting without admitting that it's been hard to use copilot. 20. The big four consultants, consulting announces AI-driven restructuring.

So here is the prediction in 2026, at least one. big four firm publicly announces major restructuring, citing AI efficiency gains. So here's what I mean by major AI restructuring. We've already seen from big four consulting companies, and I called this last year layoffs in mass, right? There's been multiple cases of 10,000 plus employee layoffs at the Big Four over the past year or so. This is different. I think one of the Big Four consulting companies is going to,

essentially similarly to similarly to how I talked about Microsoft doing in About Face, I think they're going to publicly say knowledge work has changed. So we're changing too. And I actually have a fun follow up to this one. Oh man. I, okay, I'm actually going to skip ahead one and go back to 21. So I'm going to go 20, 22, 21. I should have had AI put this together because I put in one slide out of order. But I think that Essentially, it's a pyramid model, right? Not a pyramid scheme, but consulting is a pyramid model, right? You hire junior associates.

You charge high rates for their grunt work. And then, well, agents are doing that grunt work and studies and benchmarks show. They do it better than junior associates. So eventually, companies that are spending millions of dollars and have spent millions of dollars on consulting, for the past, you know, decade, two decades, three decades. Well, they're running these agents in-house and they're going to be like, wait, why would we continue to pay our consulting company, right? I got an email.

I got an email the other day. I forgot the exact amount, you know, but someone said, you know, oh, I was, you know, going to pay a consulting company, you know, 80 grand for this project and, you know, accidentally did it myself, not accidentally. But I did it myself in an hour in the output was better than what they, you know, projected, right? You're seeing stories like that all the time, right? But when it's small little things, it doesn't matter. When it's big enterprise companies, it's going to matter.

And I think consulting companies, I'm getting this on the record. You're going to want to be a first mover, right? The rest of the world is going to, you know, come to in terms of smart. If you know what you're doing with agents now, you can one person can outperform an entire consulting team that doesn't have AI. Okay, so let me say this. Let me reframe this because you have to obviously know what you're talking about. So I'm going to say a single consultant with AI can outperform a team of 20 consultants without AI.

AI agents are that much of a multiplier. especially over the past two months, we need to wake up to this. This is why we've been seeing, you know, especially if you follow, you know, AI news and AI chatter the last week or two, things have gotten a little weird and dark and gloomy because I think people outside of that, you know, inner circle, right? Even the, you know, some AI researchers today are like, like just today, today or this week, you know, we're getting stories of it. It's like, yeah, I don't think people, society understands what's happening.

what impact this may have for jobs, right? But you have to start, where this is going to start is high-cost professional services, consulting, legal, finance, accounting, etc. Right. But these AI tools are going to compress delivery timelines from weeks to hours. And the end client is going to know. So I don't think the billable hour is going to die, but for consultancies that still lead with the billable hour, they're going to lose, right? It's going to so much of what consultants do, right?

And this isn't a knock on consultants, right? but you read and ingest information, you synthesize it, you personalize it for, you know, your client, your market, your competitive viewpoint, and then you create some output, some deliverable, right, a PowerPoint and spreadsheets. The models over the last three months do that. All of those steps in less than one percent of the time. Quality-wise, can you one-shot it, right? I don't know.

if you have a former consultant at the keyboard, I think yes, right? The PowerPoints are the prettiest. But if you're using a design skill, they're way better than what a junior would do. They're way better than a template. So yes, this is, yeah, the consulting industry is going to get absolutely rocked. And one of the four is going to admit to it. And they're going to restructure maybe not an entire company because we're talking, in some cases hundreds of thousands of employees,

but they are going to publicly restructure how they work. And they're going to have to. And now I'm going to skip from 20 to 22 because this is related. But we're going to see professional services launch what I'm calling AI flanker brands. All right. Here's what this is. So you have Verizon Wireless, right? Not cheap.

And then you have, what is it? I think straight talk, right? I have to, you know, AI mode this as I'm going live here. I believe it's that. Yes. So, Straight Talk is owned by Verizon, and it is just Verizon's services, right? You are paying for the same thing, right? But you're paying less on Straight Talk.

It's essentially, you know, what you could call a flanker brand. And I think we're going to see that from major professional services. So the prediction here is by late 2026, major professional services firms are going to introduce lower cost AI-driven service lines to preempt the disruption that's coming. Think about it.

Firms right now, they segment by complexity and risk. So lower tier services become semi-automated. And premium advisory can still remain human heavy. And that's where I think the brand separation is going to preserve the pricing power. But cannibalizing or the cannibalization is going to become a defensive strategy, right? Companies are going to create essentially these little flanker brands. They're not going to put their name on it. Just like Verizon doesn't necessarily slap their name on straight talk.

They're like, here's our low cost offering. right, we're not going to give it the same marketing and maybe not the same customer service. Companies are going to do this. I wouldn't be surprised if it's already happening. I haven't seen any stories of it. And I'm not saying it's going to be, you know, Deloitte or PWC or, you know, whatever the world's biggest law firm is. But it's going to be some of the bigger ones. They're going to do this because they're going to have to.

Because these AI native brands for professional services, high price professional services, I mean, we've already seen it. Look at like Harvey, right? What they're doing in the legal side. There's a couple of them, you know, doing it on the health side. They're going to be eating up the middle, right? The big, big, they're untouchable, right? These losses for the next year or two are going to be a line item. The small, small, small ones, you know, they'll be fine because they're small enough to adjust. But the middle, they're going to get eaten up by some of these flanker brands. So, Some of these companies are just going to have to create them. So it's kind of like a budget airline, you know, lower costs, semi-automated, but it's like consulting light, but powered by AI. So I think clients are going to increasingly start demanding transparency. Because like I said, maybe in 2023, 2024 clients were none the wiser. And then in 2025, maybe they started questioning like, hey, how are you using AI, right?

How is this affecting billable hours? When I hired my attorney for everyday AI, you know, the first questions I asked the attorneys, how are you using AI and how are you billing me for it? And then I gave them advice and I'm like, okay, well, that's good. That's not. Right. But clients are going to know. So you can no longer, you know, charge, you know, $250,000 in research, right?

Oh, yeah, yeah, yeah. We got to scope this project out and, you know, become acclimated, you know, with your industry and, you know, do all these interviews and reviews and, uh, no, you know, an agent now does that. And it's going to cost, you know, $14. And it's going to run for three days overnight and then it's going to deliver it all to you, all the answers, right? Not having to waste that 95%. No, here's the 5% that matters brought to you by an agent and that costs you $14. Why are clients going to continue to pay $250,000 for that? No, someone's going to say, all right, well, we're going to sell this for, you know, $500. And we're going to give it to everyone, right? And then these companies are going to be like, okay, we need one of these companies. We need an AI native that takes our kind of skill sets that takes our IP, our expertise, repackage it and get it out the door.

All right. Now I'm going to go back one. 21 was vibe coding gets a rebrand. Yeah, a lot of rebranding here, shifting around. It's going to happen. I think it's going to be cold. All right. I might not get this word, right, this phrase, but maybe something like agentic software

orchestration. But the prediction here is by, or sorry, in 2026, the enterprise language replaces the term vibe coding with something more structured. you know, might throw in the word, engineering in there instead of orchestration, right? It might be agentic software engineering or something like that. Here's the reality. The AI labs, their vibe coding to build the technology that we're all using.

Even AI assisted coding, which I think is, you know, most enterprises would admit to that, right? You know, they've probably been using, you know, GitHub code. pilot from Microsoft, something like that, it's not going to be like that anymore, right? It's going to be strictly vibe coding, but vibe coding doesn't have a good ring to it, right? It implies this, you know, this informal, you know, facade that, you know, enterprises aren't going to want to tolerate that, right? You're not going to approve something that says, oh, we're going to vibe code this.

But that's essentially what's happening, right? That's the reality. what's going to happen is vibe coding is going to grow up, right? Kind of like how, you know, we used to think of startups, you know, hoodies and in laptops and dorm rooms, but no, now startups are billion-dollar companies with boardrooms and private equity and venture capital, right? I think vibe coding is going to do the same thing. And this is going to get called something else. But I think that, you know, the job of vibe coding is shifting from laying bricks, to supervising the bricklaying machines, right?

Like I said, even right now, I'm technically vibe. I'm much more obviously in the true vibe coding, right? But if someone has software engineering experience, what they do is not vibe coding anymore, right? Vibe coding is you're playing around, you're tinkering and, oh, look at that. I have a working app. Cool.

Agentic software orchestration is saying, okay, here's the different platforms, the different tooling, the different scaffolding, the harnesses that we need, that our agents need. And it's, you know, got to get done. Right. Right now, you know, vibe coding, it is still the human duct tape, right? Still pulling it together.

Oh, let me look at this. Why is in this working? Let me feed it to document. You know, let me tell it to go look up the solution, right? I don't think agentic software overstration is going to be anything like that. But eventually what they're going to be doing, it's vibe coding in a suit and tie. All right. 23, the slop debt crisis is going to make some large language model data unusable. So the prediction is by late 2026, at least one major lab will publicly acknowledge

that part of its historical training data is too contaminated to trust. All right. So this is a little bit different than model collapse, which a lot of people talk about. That's not what I'm talking about, this cyclical regurgitation. I think eventually it's going to come to fruition that so much of what's on the internet right now is slop. Right. So what do I mean when I talk about slop debt? I think that there's going to be so much extra human time needed to weed out AI slop from training data. Right.

So how training data works. the very oversimplified, maybe 95% accurate version is, you know, essentially every single big AI lab scrapes the internet. They do have paid partnerships now. There's offline and online data sets, but everyone uses the same data sets. And they're extremely large. And there's good stuff in there. There's bad stuff in there. There's, you know, novel ideas and inventions waiting to be connected. And then there's, you know, racist, homophobic, you know, terrible things in there as well. So humans have to go through and take the bad stuff out. They have to train the model on, hey, when someone asks about XYZ, here's what it means. But now so much of that training data is

slop. Right back in, you know, when Open AI, you know, in, you know, the early transformer days, 2015, 2016, 2017, when a lot of these models, the foundation of them was being built, there wasn't AI slop out there. Now there is and it's bad and it's inaccurate. That's the thing. When I'm saying AI slop, I'm not even saying, oh, look at all these M dashes and so many delves. Oh my gosh, we're diving in delves. No, I'm talking about information that is just not accurate, but it looks accurate, right? It is misinformation and disinformation dressed up on a blog post on an enterprise website that no one has an idea. And I think it is going to cause silent chaos.

And I think that a big AI lab is going to have to say, yeah, we have a problem here with this internet thing. It is contaminated beyond repair. You know, it's like, you know, making a copy of a copy of a copy of something, right? And then eventually you can't see what it is. And someone just decided, oh, I'm going to fill it in with marker. And no one checked it.

And then they just keep copying. But knowing now it's not accurate anymore. Because someone just, ah, filled it. It's not even words anymore. It's alphabet soup gibberish. All right. Three more. 24.

A frontier lab is going to publicly declare humans rarely write code. All right. Caviat on this one. But I do think that at least one major AI lab executive states that AI now writes the majority of internal production code. So this is kind of already started to happen. But here's the thing. You know, Claude Code is a great example. The creator of Claude Code from Anthropic said, I believe in December on Twitter, that, you know, I think, you know, in that month that he wrote zero code.

Okay. That's a difference between that's, you know, one individual saying, hey, I don't write code anymore. That's one individual, you know, going into one product. You know, how many people at Anthropic are writing code? Thousands, right? And I would still assume that there's a good overwhelming majority that are still writing some code from hand, right? Maybe it's only 1%, 2%, you know, 20%, I don't know. But I'm guessing a majority of people, even at the big AI labs, are doing still a little bit by hand.

I think by the end of the year, it's not going to happen anymore. The jump that we saw from quarter 3, 2025 to quarter 4, 2025, in terms of what was capable, you know, not even just talking about sui bench, terminal bench, all these, you know, coding benchmarks, but just what was possible with code now, it's weird. You know, you have software writing itself. You have models recursively learning, not just improving themselves, but writing themselves. That's what's happening now, all right? That's why things are starting to get a little weird, you know.

We're starting to get some, some, you know, glimpses, you know, some of our first official glimpses of, hey, this is groundwork for artificial superintelligence, right? When models are improving themselves and humans are no longer hands on keyboard, right? Humans are just like, cool, good job, model. That's where we're getting. And I think that the statement when it comes is going to be controversial. Engineers are going to push back on the nuance. They're going to say, oh, yes, but, you know, but CFOs in the enterprise are going to

amplify that messages. And competitors are going to echo it as well. In enterprises, it's going to take a little bit wild to follow suit. But at least one major AI lab is going to say, we don't write code by hand anymore, which is going to be wild, but I do think it's going to happen. All right. 25. Portable context engines are going to replace prompt libraries. Here is what I mean. In 2026, organizations are going to standardize portable version. context engines that travel across AI tools. So think back to prompt libraries, right? You might have a Chrome extension or something like that.

Great one out there. You know, AIPRM, you know, that you can just plug into different large language models. Oh, here's all, you know, my saved prompts with placeholders. You know, you know me. I'm not a fan of saved prompts. However, prompt libraries, prompt libraries are brittle. what I mean by that is they can break easily. One small update that, and I'm not saying like,

oh, going from GPT4 to GPD5, no, I'm saying, do you know that GPD5 instant was updated two days ago? And it changed a lot. And that's the model that hundreds of millions of people use. Did you know that, dear listener? No, probably not, unless you read our newsletter, then you did read it, right? But if you had an entire prompt library, right,

It could have no longer work. Context engines, different. Context engines are modular, version controlled, and audible. This is how you, your team, your company, your industry works. Think of it as like, you know, everything you, your company knows in markdown files, right? But teams are going to ship context updates like software releases. I'm already starting to do this, right? I have my different markdown files for different things and I'm updating them at all times.

and whether I'm working in quad code or I'm working, you know, something, you know, toying around with anti-gravity from, from Google or Codex, I'm constantly keeping these markdown files up to date. So I'm putting this into practice, even though I'm, you know, part of a very small team, right? I'm doing this myself. But I think that the, kind of the MCP formalizes how these tools and contacts are going to connect to the models. And that just does enable this concept of a portable context engine, right? Because, yes, I do advise companies to find your AI operating system of choice and move as many of your day-to-day processes in there. However, there's always going to be instances where you're going to be wanting to use multiple models for whatever reason.

All right. In our very last, here we go, prediction 26. This is not technically about AGI, right, artificial general intelligence. And that's how I ended last year's show. And I don't care what you say. If you go back and look at definitions from 2010, 2011, 2012, 2013, 2014, 2015, we achieved AGI. By far, right? The thing is, over the past 10 years, the definition of AGI has changed and it's been punted, right?

We're not moving the goalposts on the definition of AGI. We're moving the entire football field, right? It's changed so many times. There's very few dorks out there as dorky as me. I went back through for a show last year. I read hundreds of definitions of AGI from as early as 2020, right, using archive.org. We've already achieved the old definition, but that's beside the point here. Let me talk about GDP value.

So GDP Val is, well, short for GDP valued. So it's a benchmark created by OpenAI, and it's developed to measure AI on real world economically valuable tasks rather than, you know, academic or synthetic or blind taste test, right? And this is quite literally giving models, economic, valuable work, giving the same thing to humans. All right. So, and then you have expert judges, and this is across dozens of different sectors in the

U.S. and producing economically valuable work, right? The consultant example, right? Go do research, you know, pull all this data, put it in a spreadsheet, put it in a report, right? An entire project or task front to back. A model does it. an expert human does it, and then a set of humans judge it. And they don't know who is whose, right?

So right now, that is a GDP valve score. So the best models in the world are at 70%. And that means that either 70% of the time, the model either wins or ties. Right? How crazy is that? So I do think that we're going to see 80%. Okay. And I think the win rate is actually going to increase.

increase a lot more. The tie rate is a little more right now. I think the win rate is going to go up even more than just the 10 percentage points, right? That's the win tie rate. So this is so important. And I think a good way to end the show and to end our 2026 AI in prediction series. At this point last year, AI could not do this front to back, right? You had to have a lot of duct tape, a lot of huge.

human duct tape. Now people don't realize this. They can do this scheduled, right? Not having to, you know, go in and get this fancy agent and, you know, give it this memory and this scaffolding and connect. No, right. One model can go out, schedule on its own, right? So semi-autonomously and do entire projects and create deliverables. right, based on your context, based on your company's data. That's where we are. Right. So I think when you are making your roadmap for 2026, I want you to keep GDP vow in mind, right?

We're already at 70%. So that means, right, AI model versus expert, judged by experts. Model wins or ties against the human more often than not, which is why I sometimes simplify it and shorten it and say, well, AI is smarter than us, better than us. There you go. A single AI model can do all of these things in one step, in one shot,

with expert level outputs. So, what does this mean for 2026, right? I gave you, yes, some of them, you know, are kind of fun off-the-wall predictions, but I hope this is laying out a roadmap. My takeaway is this. There's no more waiting. There's no more waiting. Things, right, I've always for three years, every single day, try to tell you the truth.

In terms of, A.I can do this. in terms of AI can't do this. Sometimes I'm wrong. Sometimes I'm early. Sometimes I'm late. But that's why I do this. I put it all out there for you. And my takeaway here is this.

There's no more waiting, right? For you or your company. Companies that are still trying to reskill or upskill are going to die. Right? I'm not saying you're going to go out of business, but you're going to start a slow death. That's why since literally 2023, I've been telling you all to unlearn and rebuild. Okay. So if nothing else, right, even though some of these predictions are off the wall and in fun,

I'm not trying to end on a scary note. I am trying to say, 2026 needs to be the year that you unlearn, tear it down, and rebuild. The models are better than us. So we can't keep doing the same thing. That is a recipe for a disaster. And that is not the roadmap I am leaving you with. The roadmap, the plan ahead, don't travel down the same roads. Those roads lead to dead ends.

We have to rebuild new roads. We all have to learn together how to travel from point A to point B. No one knows the best way yet. But we're going to do it. it here together in 2026. And we're going to do it every day on everyday AI. All right. I hope this is helpful.

Like I said, there was a lot more we didn't get to. Make sure you go listen to yesterday's episode. If this episode was helpful, if everyday AI is helpful to you. If you're still listening, my gosh, you literally owe it to yourself at this point to listen to me for an hour and three minutes. Go find the LinkedIn post. Go repost this. And I'm going to send you the commencement.

Complete guide. 50 plus trends. I didn't get to everything, right? Last year, I did five and a half hours worth of shows, you know, this year, a little shorter. So make sure you go repost this. It's a guide. It's ready to go. I already sent it to everyone yesterday who reposted yesterday's show. So make sure you go to go to do that. And then while you're at it, make sure if you haven't already, please go to your everyday AI.com. Sign up for the free daily newsletter. Thanks for tuning in. Hope to see you back. Well, next week. And, every day for more everyday AI. Thanks y'all. Meet Firefly AI Assistant.

Now live in Adobe Firefly, the Allman One Creative AI Studio. Just describe what you want to create in your own words and the assistant handles the rest, orchestrating multi-step workflows across Adobe Creative Cloud apps, including Photoshop, Premiere Express, and more in one conversational interface. You direct the outcome while the assistant accelerates execution. Stand control with the ability to step in and refine at any time. See it today at firefly.adobie.com. And that's a wrap for today's edition of Everyday AI.

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