#341 Our Data Trends & Predictions of 2026 with DataCamp's CEO & COO, Jonathan Cornelissen & Martijn Theuwissen

Generative AI is transforming industries at an unprecedented pace. But as AI changes how you or your team work, one thing is clear. Your skills also need to evolve. At datacamp, we offer everything you or your team need to adapt and thrive with AI. Whether it's business users looking to get the Most out of ChatGPT and Copilot, or developers and data scientists looking to fine tune models. You can learn the entire AI skills spectrum on Datacamp. Power your AI transformation today. Start learning@data camp.com. Welcome to Data Framed. This is Richie. It's a new year, which means time for another round of trends and predictions about what's going to come in the world of data and AI. In this second session of two, we're going to make some new predictions. I'm joined once again by Data Camp's co founders. Jonathan Cornelisson is Data Camp CEO and Martin Thoron is Data Camp coo. The format has changed slightly from last year. This time it each of us will suggest predictions about what we think is going to happen, argue the case, let other people provide counterarguments, and then we're all going to vote on whether we believe the prediction or not. Historically, I'm a terrible futurist since I once dismissed LLMs as being too niche. So let's hope I make better predictions today. Welcome back, Jonathan. Welcome back, Martin. It's been so long.

Thanks for having us again, Richie.

Thanks for having us.

Wonderful.

Yeah. So last time we did a review of last year's predictions and now let's dive straight into things for 2026. So, Martin, I think you wanted to go first. You've got a careers angle.

Yeah, yeah, indeed. So my first prediction is actually that I think there's a lot of talk about this junior hiring crisis, like, okay, people who just graduated can't find a job and that you're better off with or that companies will only need senior people. So my prediction is actually going to be that I think the junior hiring crisis is going to get a lot better and that it's going to be flipped and it's going to be the people who are in the middle of their career who are going to have a bit of a problem unless they substantially like reskill themselves. The reason is that it's going to be more and more important that you can think in an AI native way. So rather than companies valuing years of experience, I think they're going to value the ability of someone to work with AI. With AI, agents have the ability to set up their own automated workflows using AI. And I think that if you're a junior person and like, okay, basically from the get go at school, at university, you're taught how to work with AI, that all is going to become a lot more native and that's eventually going to make you a lot more valuable in the workforce. The mental model I have is like, okay, if, if I compare, like I grew up during the rise of the Internet and if I see how I work with the Internet versus like how my parents interact with the Internet or the worldwide web, like, it's clear that it comes still, still today a lot more native to me than it comes to my parents. And so I think we're going to see the same thing on the AI side. So which brings me to like, okay, if you're in the middle of your career, you're too young to retire, so you need to get the hang of this new world. You're going to need to do substantial reskilling to remain valuable. So, well, in short, like whatever your career stage upskilling, reskilling with AI will be or should be one of your highest priorities. So that's my first prediction.

I'm torn on this one because I'm very happy for the young people, but also I'm very much smack back in my career, like, oh really? I'm going to learn new skills again. Okay, bye. Jonathan, do you have anything to add to this?

Yeah, I think this is, this is one of the big open questions and I'm actually very curious. In a few years we'll probably be able to analyze this. The rise of AI has coincided with a lot of tech companies getting fit and interest rates going up. So I think there's a natural dynamic in the last year or so where organizations are just hiring less and it's easier to just not hire a bunch of junior people versus letting people go. But it's also easier to say, hey, this is because of AI. We're doing so amazing in terms of AI adoption that we're not hiring junior people is a much easier message from company leaders. So I'm a, I'm a bit skeptical that like AI is the biggest driver for the fact that there's less junior hires in tech. But I think there's some truth to it in some areas for sure as well. So I'm very curious how that's going to play out. And I agree with Martin in the sense that I think a lot of organizations will wake up to the fact that things in every kind of department, in every team can be done much more Efficiently, much more effectively with AI. And so they're going to start looking for not just ICs but also managers who really understand how to drive that AI transformation. And the reality is I think that a lot of the more junior people are kind of faster in adopting AI in a lot of organizations. So I think directionally I very much agree with the prediction here.

Okay. So yeah, I think just being AI native is the important thing and maybe junior people are going to be better than doing that because, you know, it's easy to learn stuff when you don't have the baggage. All right, so before we vote on this, Martin, can we have an official prediction then that's going to be something testable for next year.

The entry level positions that are open that like we see in. A rise in those. Okay.

A rise in the number of entry level positions is going to be around AI or in, in general, but with.

Dimension of AI, like I would say.

All right, so Martin, I presume you're voting for your own, your own idea, but yeah, Jonathan, what's your prediction for or against this?

Oh, I agree, I agree. I don't know how many points I can, I can put to work here, but I agree with this prediction for sure.

Okay. I mean, I like the idea, but since you sent it to me hard on mid career people, I'm going to vote against it.

You actually put something on the line that's cool.

So the next prediction, I'm gonna go with this one. So my prediction is that I'm gonna have a personal shopper by the end of the year. So I feel like we've already had bots for buying things like concert tickets. And there's a whole thing with sneaker heads who use like bots to buy like the brand new sportswear shoes whenever they come out. So this is something that Tom Tungas was talking about when he came on data a couple of months ago. We've had bots for buying things for a while, but it seems like this is going mainstream in sort of more general agents to buy things for you. So basically it's a personal shopper for everyone. So at the moment the three big players in this, it's the usual suspect. So it's OpenAI and Google are kind of going build sort of building protocols for enabling shopping within the chatbots. Amazon's got a big push. It's got a feature called Buy for Me where it's basically bots buying things for you on Amazon. So there's a huge push in this direction. Like a lot of big Players involved. And there was a report from McKinsey saying, well, actually first of all, like, global e commerce is predicted to be worth $6.4 trillion this year. A report from McKinsey saying they're written that AI orchestrated US retail. It's going to be like $3 trillion worth of retail is going to be AI enhanced. So they're looking like half of global commerce is going to be AI orchestrated by 2030. That feels like a lot. I don't know that I believe that, but I can certainly see this is like a very big market for this on the AI, like the AI company side. It's like, okay, it sounds cool that you're like these bots buying stuff. It's going to have a big impact for retailers because then retailers have to make sure that bots can buy things on your site. That means having better APIs for like finding products and actually going through like the payment systems means that like all your data has to be intact because if bots. Otherwise bots will buy the wrong things. And basically you just got to make sure that all your websites are machine readable. So a lot of work to be done here. There's gonna be a lot of stuff around like, how do you stop bots doing fraud? Can you govern bots? So many different, like topics to explore. But it sounds like it's gonna be a big thing. So I guess. Do I have your own opinions on whether we're going to see AI personal shoppers?

Yes, I think the. I would partially agree and actually I had a prediction that's somewhat adjacent but slightly different. So it depends on what exactly you mean with a kind of an AI personal shopper. So I completely agree that it's unavoidable that there's going to be a tight integration between E commerce, I would add travel to that as well, and AI. I think those worlds are starting to collide and it's not really happening yet, but it will be huge very soon. I think the consumer will want to be in control most of the time though. So I actually wanted to predict that kind of the OpenAI app store or I don't think Google has announced they're going to do this, but it seems like the obvious step for Google to integrate apps into Gemini. I could see that kind of blow up and be huge. Where for example, I live in the us I was traveling to Belgium. I want to book a hotel. I have certain criteria. I was asking ChatGPT to help me find something. I don't want to leave the app and go somewhere else. And then I want to book a restaurant. Say I don't want to. I just want to do that within kind of the gemini or the OpenAI interface. So I can totally see the integration. But I think most consumers, before they spend the money, they want to make sure they approve what it's buying. So like just a bot that goes off and kind of goes shopping for you. I think there's people who will like that, but that's a very small market. So it depends on how you. What, what you mean. Richie.

Yeah, okay, so that's a good point that there are very different scenarios. So if it's just, do I ask my favorite chat bot, like for advice on what to buy that's already very easily doable, but having that complete thing of, okay, I give it a shopping list with some requirements so it goes off and then finds the products and I give it my credit card details to buy stuff. So it's that end to end experience with like no human involved. That's a, that's, that's the kind of thing that's supposedly coming next, but that's a very different play.

I agree that's going to be possible and some people are going to use that. I just think the number of use cases where you're not ultimately approving feels like it's going to be small. I could be wrong, but like maybe if you say, hey, I want to cook this or that and then it goes off and buys the ingredients. But like in principle it could book your whole holiday. But do you really want that? I think you want to approve. If you're going to spend thousands of dollars, yes.

What you buy matters. So certainly a surprise vacation bought by an agent that's like, it's a novel thing to do once, but I wouldn't necessarily want to do that every year. So I think that the thing that I think is going to be coming first is going to be B2B buying where you have regular orders and it's like, okay, check. Is the stock level low at this point? We always, once the stock level drops to a certain amount, we always buy the same thing every time. So when you have repeat orders for similar things, that is very programmatic. It's a small stretch from existing rule based stuff to I've got an agent that's a little bit more flexible about like where I'm buying from and things like that. So I think B2B commerce is going to be the initial thing on the B2C side. It's going to be small purchases. It Was like, I'm fine with it. Like buying toiletries, I guess. But yeah, not quite the vacation yet. All right, so to make a prediction then, to make this official. So in 2026, I'm say the AI shopping agents will directly execute $64 billion in global E commerce. This is gonna be 1% of all, all global commerce values. We're going, we're going big here. And that's beyond, like the full execution. I'm giving the bot my credit card.

1% I think is too high. Like I'm going to say no. Like, I don't think one out of a hundred dollars spent is going to go through an AI agent fully automated, to be clear. I don't think that's going to happen. You look very disappointed. For those who are listening, Richie has a very disappointing face. Tala.

Shot down. Do you think I went too high with the 64 billion? I was like, 64 billion sounds like a cool number. So let's go with that. Jonathan, what do you reckon?

It depends on the fine print of how you define this. I'll raise you that. I think the amount purchased through kind of AI apps or integrations I think will be higher than whatever the direct bot purchases is because I think it's a much bigger market. So, like, if you're just to be specific about what I mean, there's apps in OpenAI now, I'm guessing. I'm hoping this will come to Gemini and others as well, where you say, hey, I want to buy, I want to book a hotel, I want to book a restaurant. You kind of. I want to book a plane ticket. You kind of book it through the app interface or for the integration into OpenAI or Gemini. I think that will. Will be bigger than bots going off and kind of based on high level instructions doing the purchasing for you.

Okay, so this is human has final approval, but basically.

So it's integration within AI. That would be my bet that the integration within AI is the big market rather than the AI agents going off to purchase. I think eventually that might actually be the bigger market, but for the next 12 months. I don't know a lot of humans who are okay with bots kind of going off and purchasing stuff for.

Okay.

People will have to get adjusted to that idea.

We're keeping the human in the loop for this year then.

I think so.

Okay. All right.

I guess that sounds like it's a no on the original premise review. I'm not going to cheat and update my prediction for the sake of things, so. All right. I'LL let you both say no on that one. But I do think you have a good point there, that maybe human in the loop is going to be necessary in the near term.

Yeah. What do you guys think on that prediction? Should we immediately do the vote like.

Oh, you want an extra? Okay, all right, yeah, let's vote on this.

So on that one, I think if the, if the tools get good enough. Yes. So, because I was actually also thinking about like for example, you have Perplexity's Comet browser and I tested it and it didn't really work yet. But I, I went to the website of like the grocery chain that I buy all my groceries from. I took a screenshot from like the grocery list that lies around at home. I put it in there and I said like, hey, go and order it. Sorry, put it all in the basket and I'll give final approval if you get excited now. It didn't work. To be clear, it got 80% there. Now if it gets a hundred percent there, then yes, like that type of use case where I finally say like, oh yes, everything is in the basket, let's pay. Like, yes, I think that's going to be use case and it's going to be huge. Yeah.

All right, so if we're going to vote on this, I think you need to put a number on it though. Uh, how big is it going to be?

I, I heard six. I'm, I'm literally just going off of what you said. I heard 64 billion sounds like a nice number and honestly that, that seems very high to me. But I, but I think it will be significant and I think once it starts to work, it can go really fast. Like I would immediately use it for a lot of things I'm, I'm purchasing.

Personally, there are a lot of really bad shopping websites which I would rather avoid having to.

Exactly.

So yeah, I think that's a big message for retailers. Like if you improve your website, that's how you keep people there. Like, you know what, the bar has just been raised. Martin, how are you voting?

I'll, I'll vote on the 65 billion, but with the caveat that like it works and it's like not, not necessarily a direct purchase by the bot. It's human in the loop.

Okay, we're keeping the human in there. All right. Okay, next one. Joe, he had one on one education.

Admittedly the next one is a bit self serving. The prediction is that everyone's going to get their personal tutor and I think everyone knows that like one on one tutoring has always been the gold standards in education and the issue was the cost and the fact that there's just not enough exceptional tutors in the world to teach everyone. But technology is now at a tipping point where AI driven teaching is ultimately going to get better than the best teachers in the world. We're not quite there yet. That's something datacamp has been working on. For those who follow the news, we acquired a company called Optima in late 2025. We've launched our first AI native courses where the course is essentially delivered by an AI tutor and everything adapts to the learner. So it's truly personalized, it adapts the pace, it adapts to the context of the learner and the early results are really positive. Even though this is a completely new experience, it's already better in terms of engagement, in terms of retention. And I would argue if you kind of analyze the sessions with learners, it's also just better from a tactical point of view. And so I think there's tremendous potential here. Just like in other areas, there's protections around having your personal shopper, for example. I think the area where it's going to have a very, very positive impact on the world is everyone ultimately having their personal tutor that knows them really well, that adapts to them, that is not just a teacher, but also a coach. Helps them be accountable, helps them clarify their goals. I'm super bullish on this and that's exactly what we're pursuing with datacamp. So it's a bit self serving, but yeah, that's the prediction. Everybody gets their own tutor. Maybe to make it something tangible that I can be held accountable for in a year. I think one way to phrase it is I would say half or close to half of engagement on datacamp shifts towards the AI native experience by end of year. That's not an insignificant amount. We have about in the last year, I think we'll have around 10, 11 million hours of learning on the platform. So that's quite a lot of learning that would happen through AI.

Okay, yeah, I think I'm okay with self serving this case because it's such a huge societal shift. Having a personal tutor for everyone that actually works. And yeah, this is how we make education better for the world. So yeah, I'm okay with that. I also love that you're like publicly stating your okr for as a bold move. I hope I don't have to do that in future.

Okay.

Martin, did you want to add anything on AI education?

No, I'm fully supportive of this Prediction anecdotally, like, I'm a heavy taker of datacamp's new course format. Like, thanks to the Optima acquisition, Like, big fan of them. Like, they're really, really a lot of fun to do. I feel myself a lot more engaged. Like what George tells about the personalization Taylorization, like, it's definitely true. So, yeah, very bullish on this one.

Absolutely. Yeah. No disagreement for me here. So I'm fully in favor of this. It does sound like a big target you've given yourself there.

As you know, Richie, we like ambitious goals. This is not going to be a walk in the park. And to be honest, I think the maybe to take a step back here already today. I think more than half of students prefers the AI native experience. The real kind of question is datacamp has 500 courses. Can we bring the new experience to enough content to really shift the engagement to the new formats? But we're very optimistic by the end of this year we'll be in a great place there.

Next prediction. This one's from you, Martin.

Yeah, so I predict that we're going to have a new GPT3 moment. So to frame that a little bit. So I think there was a big leap when we went from like GPT2 to GPT3. Like GPT3 was when like AI really got known to the more general public. It was probably like the biggest step forward in the last like four or five years. Like LLMs went from a research project to something that was actually useful, mass consumed and so on. And so I think that in 2026 we're going to have a very similar moment and let me explain why. So this year Gemini 3 came out. There was a lot to do about Gemini 3. Michael gave very big model update upgrade. Google stock went through the roof thanks to that. Basically what Gemini 3 show was that the scaling laws still matter. So basically pushing in more data will lead to better models. Now, if you take that into account and then there's a second big thing coming, or there's a big thing coming next year and that it's Nvidia will start shipping its new chips. They're called the Blackwell GPUs and so it's going to start shipping them to their clients. And I was listening to some other podcast and the way that somebody described it as like, okay, so today there's the Hopper GPU and that's like an F14, like that, that airplane and Blackwell will be like an F35. So like the latest, the best, like whatever is. So there's a massive difference in here. And so what's going to happen in 2026 is that we're going to get new models coming out and they're going to be trained on this much better infrastructure. So they're going to be trained using Blackwells rather than Hoppers. So they're going to be trained on the F35, not the F14. And so if scaling law still holds, that means that these models will be much, much better. And we're going to all observe as consumers like that stepwise increase just like we've observed it when we went from like GPT2 to GPT3. I even want to go a little bit more specific. I think it's going to be grok from Xai who's going to come up or come out with the very first big lead there. The reason is like, okay, Blackwell will come out like rumor is that Xai Elon Musk Musk will get the first ones. And okay, he has proven that he can set up gigantic data centers very quickly. So I think his models will hit the market the fastest. And so yeah, therefore I think it's actually we're going to get this new GPT3 moment driven by grog. So maybe to put that in an official prediction like okay, I think gogtree or so new GOK model GROQ model coming out of after the blackwall deployment will take the top of the charts there.

The infrastructure is now going to be the important thing like that sort of produced latency and high performance is going to be the differentiator in these models. It's slightly different to what we were talking about in the review of the last year's predictions about the prediction that there are going to be AI products that take more than an hour to respond. And that's kind of a different aspect to like, okay, we're letting something letting AI run for hours versus we really care about the infrastructure and we're going to get fast responses. We've got high throughput.

That's about inference. And I think this is really like my statement is really about the training of the models. So my prediction is really on like a new model will be trained using this new infrastructure thanks to Nvidia's new gpu. And that will lead to like a stepwise improvement in the model like we've seen with GPT3.

Okay, so it's about continuing to scale the training side of things. Do we get these bigger models?

All right.

Okay. Jonathan, do you have a take on this?

I would tend to agree and I would feel nervous about betting against Elon Musk. There's a lot of people who've lost a lot of money doing that, so it seems like a bad idea. Talking from the educational perspective, in the last year we've kind of seen a transition from the models getting so good. You could essentially have an AI tutor deliver content in a very effective way. And while being able to trust that the quality is very high. Our bet for, for this year is that the models will get so good that you can really accelerate content creation for that format to the point where it allows datacamp to not only teach data and AI skills, but really go beyond that. Because to some extent we're downstream from this and our bet is the models will continue to get better and as a result, not only will we have an effective AI tutor and AI coach, but we're going to be able to scale our content while maintaining a very high level of quality.

Yeah, so this is interesting. Like, I think for a lot of sort of simple use cases, the large language ones have reached a point where they're just good enough. But for things like the tutoring use case, where people are asking technical questions about data science, about AI, and then you need to. The AI needs to come up with a coherent and correct response that is understandable for user. It's really hard problem. And so we do need models to keep getting better in order to get those sort of high levels tutors. So, yeah, we need scaling to happen in some sense. Martin, so what are you going to predict to make sure that this is measurable and testable for next year?

Yeah, that's harder because it's hard to capture a moment. So I think the closest I can get to is like, okay, Grok will be at the top of the leaderboard.

Okay, we're back on the Chatbot arena leaderboards then.

Yeah, if we need to put it in a factual metric. Yes.

Okay. All right. So, yeah, this is interesting, but the.

Prediction is like, it's going to be this new stepwise improvement.

A stepwise improvement? Yeah. I mean, I feel like there's enough money being thrown into AI research that we should be guaranteed, or at least close to guaranteed, like some decent improvements over the next year. But yeah, having Grokker at the top of the leaderboard is always in number two position at the moment, the latest GROK model. So yeah, only, only a short step to number one. I'm not super convinced though that it is going to be Grok at the top. I feel like Google's going, guns are blazing at the moment. And they, they're certainly going to be able to get access to the latest Nvidia chips. So I suspect it's going to be Google number one next year.

Just to avoid a discussion we'll have a year from now. Are we saying at some point Grok in the next year Grok will top the leaderboards and then maybe it falls off? Or are we saying exactly December 31st, it's at the top of the leaderboard? What are we saying?

So I'm saying like Grok will hit the top of the leaderboard and it will be due and that's how it's going to be communicated due to the fact that they're using this new GPU by Nvidia.

Okay, yeah, then I agree.

So Grok getting number one at some point during the year. I think, I think that's, that's likely to happen. They just seem like the leader kind of swaps regularly. Whoever releases the latest model gets the number one spot. It's going to be harder to track, you know, finding a. Did they hit number one at some point during the year? Yeah, I think that's, that's, that's an easy yes then. But I think by the end of the year we're going to see someone other than Grok being the overall leader. Next prediction is from me and that's the AI app builders are going to be the next hot tool that everyone needs to learn about. So I feel like over the last year, well, I'm almost every software development tool has now got an agent involved, but the biggest sort of winners of the last year, well, it was Cursor and the other sort of AI native ide. So the idea is that you basically you have agents built into your coding interface, but that sort of the sheet of code is like still like the standard thing that you're interacting with now. I suspect that in 2026 it's going to flip a bit the AI app builders. So this is tools like Lovable, like Replit, like Bolt. So these, do you say this is the thing I want to create and the agents over in the background and create what you want, but you care less about the sort of the, the code. So it's another layer of abstraction. We also have tools like Sigma is doing a similar thing in the, in the BI space. It's basically creating, it's AI agents that will create your apps for you. It's a subtle shift in how you create things, how you do software development. But I think, yeah, 2026 is going to be the, the year of the AI app build this. I don't know whether either of you have a take on this.

I believe in the category. I believe in, in the fact that there's a whole new group of people that's going to build apps. Where I'm unclear is like, we've heard all the numbers on Replit and Lovable, but I'm a data person, so I always look at, okay, what's their. What's the evolution of their traffic over the last six months, nine months. And both Replit and Lovable are essentially flat. So there's no growth. There was some initial growth and obviously there's been a tremendous amount of revenue growth. And so what that signals to me is that what's really happening is you have a group of people that gets a tremendous amount of value from these products and they're spending probably really large numbers of money unlovable on Replit. And so on the one hand that means it works. On the other hand, the fact that their traffic remains kind of flattish over the last six to nine months, it does single, like, hey. Or it raises a question on, like, why is that the case?

Right.

It's very rare to have a company where like the number of users and the user inflow seems flat, but revenue is going 10x. Like that's a new thing. And I don't really. It gives me a level of skepticism around the growth model. That doesn't mean I'm, I'm, I'm not bullish on the whole category and the idea. But like, for, for example, like, is cloud code getting so good that people just skip it and just go straight to cloud code? I don't know the space well enough. I'm just, I'm a bit skeptical that if this is the next big thing in 2026, you would see both revenue as well as users grow. And unless similar web is like vastly off, like, there's no traffic growth. So I'm a bit confused about that.

Yes, A very odd data signal because you would sort of expect that. I guess these companies have been targeting the software engineering group and then they're kind of gradually moving towards the sort of more, I guess, well, less technical people. So the replica in particular has been pushing hard on the idea of vibe coding. So it's like, yeah, you don't need to understand too much about the underlying code. You just want to be able to describe what you want to build and things just magically happen. So yeah, you would expect that at some point it's going to switch and it's going to go mainstream. You're going to see a large amount of traffic growth. Martin, I don't know whether this is something you've looked into.

A bit of a similar take like Joe. So if the prediction is like, hey, their revenue will grow and will continue to grow rapidly, probably like, yes, maybe going to the cloud code point. Like, I actually think there's also a world where the cursors and the cloud codes and so on actually move into the space of Replit, Lovable and Bolt and you actually see like more and more users switching to those tools, which would actually make the total addressable market of like a cursor or cloud code, like a lot bigger. So maybe Richie, like going back to like, well, asking you like, okay, what, what's the official prediction here?

I was, I was going on the valuation. So currently Cursor, well, last month it was valued at 29 billion. And then the two biggest players in the AI app building space, it's lovable, that's valued at 6.6 billion. Replic is valued at 3 billion. So I'm going to say the either lovable or replic will overtake Cursor as the highest valued AI assisted software startup.

Oh, that's a hard one.

Yeah.

Going on valuation. So, yeah, basically cursor's got like a 5x lead over lovable mode, 10x lead over breathlet, but I think it's going to switch in this year.

Oh, that's very tricky, to be honest, because I would be bullish on all of them. Although I do think the valuations are. If it's true that Traffic is actually now growing, the valuations are very high because they assume a revenue growth rate that, that it's hard to sustain with traffic being flat.

I'm going to say you're wrong, Richie. And I think what's going to, the way it's going to play out is like, okay, lovable or replit or both are going to get at some point a higher valuation than Cursor. But then before the end of the year, Cursor actually raises again over the 30 billion valuation that there are now. And that will like turn your prediction to be false. Well, we're guessing a little bit VC behavior here, which is really hard. But like that, it's like that's outside of like a gateway thing. These tools are really cool. And will it be used a lot more next year? Probably, yes.

Yeah. So certainly, like there's a lot of excitement, like whatever format you're using. So yeah, it's like three big companies and yeah, all exciting stuff happening. But you're saying no. Okay, was that Joe, was that also a no from you or you had a skeptical face?

I'm on the fence so I'm just going to strategically vote no.

My predictions, that's been both no's from both of you on both of these occasions. All right, go on. Next one. Joe, you had another software engineering related prediction.

Yeah, so the next prediction I have, I would summarize is quality is the new differentiator. I think in a lot of spaces we're just talking about kind of AI assisted coding or coding agents. There's not several players that are doing this reasonably well. And so I think in that space the ultimate quality of the product is really going to start to matter. If you think about the education space that we're in, there's going to be more and more companies that claim to have an AI learning engine and AI tutorial. But already today there's vast differences in what that actually means and what the quality is. And I think there's going to be an evolution in how important the kind of output and the quality of the output of the AI models in kind of the application layer is. And so there's going to be much more rigor around the product management, around kind of creating specs, having really good evals and having that iterative loop I think will become super, super important. Whereas I think up to this point in a lot of cases there was a bit of a wild west kind of if the demo works, it's good enough, let's launch it attitude. People wanted to be first to market. I think 2026 will be the year where quality is the differentiator, where people are now really looking at ROI and really going to evaluate whether something actually works and is actually better. Maybe this is wishful thinking to some extent, but I'm still going to make the prediction.

Yeah, maybe wishful thinking we're going to avoid drowning in a tide of AI slop. But I'd say over the last few months I've been running a lot of data framed episodes with trust in the title because I really want to be a sense that you can use AI to create high quality things and yeah, anyone can generate anything cheaply now. So the big differentiator is going to be can you create something great? So yeah, doing that well is good. So yeah, I'm in favor of this.

Yeah, I'm as well actually. So agree with the prediction. Maybe small plug like we have on Data Camp and AI assisted coding. For developers. We also are soon coming up with an AI Engineer engineering curriculum fully in our new AI native learning experience. It has a large section on quality, both from a prompting perspective as from an evals perspective. So one, yes, I think this is going to play out as Joe predicted. Two, if you want to be prepared for this like go to datacamp, check out some of these new courses that will become. Because like we're going to heavily dive into this subject as well.

I mean both these last predictions have been about how software development has just changed so much and building things change so much that yeah, it's going to be a big focus for the Data Camp curriculum team over the next few months. So I guess you start off, you take the Vibe coding with replic course and then you figure out how to like go from Vibe coded something to do it to doing something well tested and high quality at the end of it. Actually how, how are we judging this one?

I'm not sure. Any suggestions?

The point, the point about is going to be a workflow change for software developers. So this is the idea that you've got spec driven development and you can have some kind of evaluation thing. So it's like the idea of this sort of behavior driven development workflow is going to be less like the dominant sort of thing. I don't know how we quantify that, but maybe we'll do Vibes based evaluation next year. All right, this next prediction's one of mine. So I'm a believer that we're going to have a very successful open weights model in 2026. So in 2025 we saw a few success stories. We've talked about them a little bit already. So for example, at the start of the year Deepseek had R1 that was kind of blowing everyone's mind with like the great performance for how cheap it was both to create and to run. And then in the middle year There was the Quinn 3 model from Alibaba was kind of like the darling of all the AI engineers. Towards the end of the year there was the Kimmy K2 thinking model was again like filling up my feed. There's been lots of very successful models in 2025 but as you've seen like the top performers on the particularly on the Chatbot arena leaderboard, it's been OpenAI. It's been all the Gemini models from Google and then Grok more recently has been up there as well from X AI. I feel like if you're a startup creating AI products or creating AI features in products, then probably at least for, for part of the product, you're going to be running inference on like modal or base 10 or one of these like inferences or service platforms. And then if you're using one of those, then it's going to be an open weights model that you use. So I think the sort of the, the latest generation of all these like AI native startups, they are using open weight models by default where they can just get a cheaper. It gives them a bit more control over what they're doing with the models. Also if you're a company where you need to do fine tuning, you want to customize your model, then generally you have a bit more control with these open weights models. So I think we've talked about this already, but the pushes in general towards open weights models being the standard and that means, I think because the gap is closing in performance at some point during 2026 we're going to see an open waste model top the Chatwood arena leaderboards board. So any thoughts on this, either of you?

I agree, actually, I think for the same reason. Ultimately in software open source kind of became dominant. I think there's a lot of similarities where eventually open source will become a big player in this ecosystem as well. The question is when? And just if you look at kind of the trends, I think it's a reasonable prediction that 2026 is the year where they kind of make it to the absolute top.

I agree the timing is the tricky part here. I feel like I'm going to be right eventually, but is it going to be in 2026? I know this slightly contradicts your prediction, Martin, because you were talking about how it's going to be early access to the best chips that enable you to create the biggest and best models that are going to help you win. Uh, I don't know whether open source or open weights models are going to have access to these things. Uh, so what do you think's going to happen?

Yeah, so it's will it happen eventually? Like I actually agree with both of you. I also think it's going to happen eventually. Just for the fun of it, I'll take the opposite side. So I think it will not happen in 2026. Two reasons. One, as I explained with, with the access to the latest advanced chips, I think that's going to be a big difference. Now that's not going to be from the start of the year. So like the, the critical period for me will be like what happens in the first part of 2026. The incumbents, the Closed source models that they still have a bit of Runway and that they'll continue to dominate. I actually think the risk to my prediction as to what will Meta do, will, will they bring out a new model and will that be open source? I think that could flip it, but we'll see.

Yes, certainly. Of all the companies pushing the open weights models, Meta is by far the biggest, has like the biggest financial firepower and 2025 has basically been a complete turnover of the team. Like Zuckerberg spent. Well, not personally spent but like Meta as the company spent a lot of money acquiring both people and companies. So. So yeah, I'm expecting big things from Meta in 2026. Like that new team has to deliver.

All.

Meta's waste a lot of money. So yeah, a lot of talented people. Hopefully they're going to pull something out that I think that's going to be my ace card in picking this prediction come true. All right, we've got one more prediction to go then. So Martin, I think this one's yours.

Yeah, yeah. So a final prediction from my side. So I predict that OpenAI will launch a sign in with ChatGPT or maybe named a little bit different but so they'll launch like a social login, just like you can sign in with your Gmail, like you can sign up your Facebook account to websites. And I think that the reason they do that or what will happen is that the login will not just like authenticate you but it's going to pass through like your AI credentials, like your preferred model, like maybe usage limits, you have like credits, budget, maybe a little bit of like memory settings like we talked about in the tutor context of like okay, how the AI tutor will know who you are well and what you like. I could see like the foundational models that you interact with on a daily basis will have similar type of information. And by the way, I think that Google Meta, like they'll like whoever has a sign in with X evolve in the same way. Like it's going to be like you sign in with X and you're going to share your AI configuration. It will allow you to start using AI software apps, AI features without using like at some point, even without using the vendor's own model of choice. But like your model that you have credits for that you pay for the underlying assumptions. Obviously like the foundational models become quite interchangeably. So like OpenAI models will be as good as Google's model and so on. One of the reasons I think by the way it will take off is because of the cost control. So it looks like a degree of cost control for your users and it definitely it like you centralize your spend, you enforce policies, you avoid paying like an AI tax on every tool that you have. Yeah. So I'm not sure if it's going to take off massively, but I do think that they'll do an attempt to it and like try to make it a thing.

It feels like OpenAI is trying to ape the meta business model. I guess like Facebook is like, it starts off as a basic social network and then add a load of extra features. You can do a lot without leaving Facebook if you want to. And I think a lot of other companies have tried this where they just want to be like a central hub. I mean, I guess that's also. At one point, Elon Musk was talking about X being the everything platform where you can do like, you know, you can buy stuff and, and do everything from. From inside X. And it's not really quite happened yet, but do seem to be like a lot of companies wanting to just be like the center of where you, where you access the Internet from.

Yes. Okay.

All right, so it sounds plausible. Jonathan, do you have any thoughts on this idea?

I like this idea. It's a different question whether it's actually going to happen. I think the implicit prediction here is like, you're going to have vendors that leverage AI, that kind of will integrate with. With anthropic OpenAI and organizations will want their credits to be used that way. I could totally see that this is essentially the consumer version or the scalable version of that if this plays out. Actually, my thought was Google is in a great position because they already have so much connectivity. They have kind of the Google login and that connectivity. So if this becomes a big deal, OpenAI will have to do it to compete with Google. I'm kind of on the fence to vote yes or no. So I think for the fun of it, I'll go with no. But I could totally see this happening and I think it logically makes sense.

Yeah, I mean, certainly the idea of being able to use different models from the same place, it seems pretty important in the business side of things where you do care a lot about the cost of stuff for individual consumers, it's like, do they care about which model is being used enough to worry about that sort of thing? So it's tricky. Yeah. I don't think I have enough evidence to be convinced at the moment, but Martin, you look thoughtful.

Two quick things. Maybe first on Rich and something what Joe Said Richie. I think if you already pay privately for Gemini, take an example, and you go to a service and that service uses Gemini under the hood and they're going to charge you as well for that. Like you're not fully paying double, but like you're, you're paying for something that you're already kind of paying for. So that's like why I was thinking about it. Like again, like a lot needs to be true for that to happen. Like one thing is like models need to be easily be able to switch in and out. Like service providers need to provide it. Joe, you gave a different data I actually found interesting. Like, okay, if you're indeed an enterprise and let's say like you negotiate a great deal with anthropic and you can buy tokens at a discount and the vendor you use buys tokens not at a discount and also uses anthropic. Like, yeah, indeed, you, you want them to use discounted price, you're going to say like, no, like let's use my tokens rather than your tokens. Actually a fair point.

Maybe this is already playing out. Actually I haven't heard any, but. But that seems like. I don't know if this is the eventual model, but at least for some time it feels like there's going to be pressure for that to happen.

I do like the idea of not having to pay twice for AI or like if I get a good discount on AI, I want to be able to use my discount everywhere. That feels like a good shopping experience. I think the cross company politics and the engineering challenges are going to make it not worth it to happen. So I think I'm going to vote no on this one.

My guess would be, actually most engineers working on AI applications are already building it in a way so that they can easily switch models in and out.

Yeah, if you're building stuff yourself, you want to be able to switch it easily. Like it should just be a case of change the name of the model you're using. The idea of being able to bring your own AI to an application, it feels like not in 2026 for me. Joe, which way did you vote in it? You were a bit on the fence.

But before I think I'm voting no.

Okay, another no vote. All right, so with that, I feel we've all been like very harsh about each other's ideas this year. So it's going to be interesting to see who has the, the most correct answers when we come to review this next time. All right, super. So with that we got nine different predictions for you for next year. I hope everyone in the audience is enjoying them. I guess let us know on, on social media whether or not you agree with any of these predictions and with that. Yeah. Thank you, Martin. Thank you, Jonathan.

Thanks a lot, Richie. Perfect.

Thank you, Richie, for hosting.