Limitless: An AI Podcast - AI in 2026: Predicting the Next Trends That Will Change The World

Episode Date: January 2, 2026

🌌 LIMITLESS HQ: LISTEN & FOLLOW HERE ⬇️https://limitless.bankless.com/https://x.com/LimitlessFT------In this episode, we predict the future of AI by 2026, debating AGI breakthroughs and the role of GPU clusters in space.Get involved. What do you think about the future of AI?------TIMESTAMPS0:00 Predictions for 20262:27 GPUs in Space4:22 Biggest IPO of 20265:34 Long and Short Trades7:36 Industry Predictions8:42 The Layoff Lotto10:29 China's Open Source Dominance14:19 The AI Crash Out15:37 AI Bubble Implosions17:11 Breakout Hardware Devices18:50 Comeback Player of the Year20:45 Most Valuable Robot21:53 AI Entertainment Apps23:36 The Largest Model25:24 Context Window Records27:32 Winner Takes Most29:08 The Biggest Loser30:44 AI Blue Chip of 202631:57 Best Model by Intelligence34:08 The Hottest Trend for 202637:27 Predictions Wrap-Up------RESOURCESJosh: https://x.com/JoshKaleEjaaz: https://x.com/cryptopunk7213------Not financial or tax advice. See our investment disclosures here:https://www.bankless.com/disclosures⁠

Welcome to our all-knowing predictions episode. As to hosts of an AI and Frontier Technology podcast, it seems fitting to end the year with an episode all about where we think the world is going. We just did an episode that was kind of recapping the biggest winners and losers of 2025. If you haven't seen that, I would highly recommend. But this episode is looking forward into the future. This is where the puck is headed to. This is where we're going to try to predict the most impossibly fast trend that is going up into the right in the world of AI. So everyone without further do welcome to the Delimitless 2026 prediction show. This is where we're going to talk about all of the things, starting with a topic that I think everyone's kind of most excited about. Wait, hang on a second,

Josh. Hang on. What do you got? Fit check. Josh and I went above and beyond. I've got my fleece fully zipped up and I've got my tinfall hat for all my conspiracy theories that you're about to hear. I apologize in advance. And Josh, Josh has gone with his favorite Apple CEO, the Tim Tim Cook, Steve Jobs look. He's got the turtleneck on. He's got the glasses. Might I say, Josh, you're looking mighty predictingable. I feel very sophisticated today. Although I will say these glasses are like a minor prescription and they're hurting my eyes.

So we're going to try to get this episode done quick so I could take them off. Okay. Question one. So what we're going to do for those who are watching, we have gold stars. We're going to hand out. This is the Limeless Gold Star. If you are a receiver of the gold star, congratulations. We are bullish on you in 2026.

The first question, the first prediction we're going to reveal EGS is what I think everyone is kind of most curious about is the AGI question. AGI, here's our alarm bell. Are we ringing the bell or are we not ringing the bell by the end of 2026? Will there be AGI or will there not be? Okay. My answer is yes, but it comes with a clause, which is specifically for science. I think we're going to make a major breakthrough.

By major breakthrough, I mean we discover a cure for like a major disease by the end of the year. And I'm saying this for two reasons. Interesting. One, science progress in AI has been friggin amazing. And two, according to Sam Altman's timeline for OpenAIA AGI, he's predicting we get science AGI by the end of 2026. So that's my prediction. Okay.

And how would you define AI in that case? Something that shocks me by the end of the year, Josh, that I'm like, oh my God, we could never have thought of dread this up by the end of last year. I think if you create a drug that literally saves millions of lives, that's AGI. Quote me. Okay. Okay. I think my prediction for this is going to be no.

I don't think we get to AGI. And for me, the loose definition of AGI is like an intelligence that's more capable than us, pretty vastly more capable than us in anything that we can do. And a big part of that is the physical embodiment of AI. Like it shouldn't be limited to just bits. It should be also extended out to atoms in the world of robotics. And I don't think robotic understanding is going to be good enough by the end of the year to feel like it is a truly super artificial general intelligence. So that's why I'm going to say no on EGI for next year.

We're not ringing the bell or ringing the bell in some categories. Who knows? I guess moving on to the next question here. Okay. Houston, we have a GPU. Over under, Josh, do we have a 25 GPU cluster in space by the end of 2026? AI data send us in space. I was the biggest hater in the world about two months ago.

I am taking the over. I think we're going to have more than 25 GPUs in space by the end of next year. Why? Because Starship is going to work, and we are going to hopefully go to a launch and watch that thing work. If Starship works, there is no world in which they don't just ship up, like SpaceX themselves ships up a cluster just to prove a proof of concept. We right now, if I'm not mistaken, have a single GPU cluster in space, which is training. So that would imply a 25-fold increase.

I think it's happening. Ejess, where do you stand on this? I'm also going to take the over on this, but it's not going to be a super bullish take because regardless of us having over 25 GPUs in space, I don't think it's going to be better than the clusters that we're going to have on Earth in 2026, but it's going to be a good initial proof of concept.

You're right, StarCloud has already this year launched one GPU up there. It seems pretty feasible with all the SpaceX launches that we're going to get more than that in 2026. Yeah, it feels totally contingent on the, Starship launch. So we're going to be watching Starship and monitoring that. I'm going to be going to be buying SpaceX Chaz when it IPOs, just to be clear. And next year for the IPO too. Oh, we should do biggest IPO of 2020. All right. Wait, we didn't add this, but I just, I'm going to

right now. Okay, okay. Let's go. Biggest IPO, what's it going to be? Because, wait, before you answer, we have some rumors of IPO. So we have SpaceX, 1.5 trillion. We have open AI, which is roughly $800 million. We have possibly anthropic going last year. So like, there's a lot in the pipeline. Who's the biggest winner? Okay. It's not going to be. what you expect. I'm going for Open AI. I think Open AI is going to have the biggest market cap by the end. Purely because more

people understand the AI play. I'm not saying this makes sense. I'm saying more people believe and understand the AI play and so that like retail people will buy it as well as long-term investors versus space people are still going to be like and there's a lot of Elon haters out there. I'm not one of them but like

I feel like that might play against them. Okay. I'm going biggest IPO SpaceX for sure. We just published an episode last week about why $1.5 trillion is undervalued. So I expect SpaceX to hopefully eclipse the two trillion dollar market cap, put over 25 GPUs into space and be the biggest IPO of the year. But we have a lot of good IPOs to cover. Next year is going to be ridiculous. Anyways, now it's probably time to get to the spread trade rate of next year. If you're looking to trade markets, if you're looking to make money, we're going to talk about what company or sector is most likely, it's what you're most likely

the long next year versus most likely short. So EJ's where are you starting on the long front? What category, what industry, what company are you most excited for that you're bullish on that you think will make you the most amount of money? Okay, I'm going to cheat, Josh. I'm going to give you two answers. I'm going to give you a company. That's fine. I'm going to give you a sector. Good thing is this is our show. So we could kind of, you know, go for it. Exactly. Okay. I think the company to long next year is going to be Amazon. I think Amazon has been incredibly slept on. I think their new tranium chips are going to get used way more than people expect. And I think their AWS spread out for AI specifically, because they're creating AI factories is going to be insanely good.

And the sector that I'm longing next year, this is going to be unsexy, but I think it's going to be true, is still GPUs. GPS are going to kill. Okay. So that's Nvidia. Are we excited about AMD also? Any other GPU?

Nvidia. AMD, if I could put my Asian stock market hound, sorry, I've run out of headspace. I've only got the tinfall hat. It would be Samsung. I think Samsung's going to absolutely kill next year. They've got new chip fabs. they're working with Elon, it's going to be sick.

What about you? Just today, Samsung announced their two nanomino microchip ahead of Apple for the first time ever, which was huge. For me, the company I'm most excited about is Tesla, which should probably come to the surprise of absolutely no one. I am so unfathomily bullish on Tesla in terms of their ability to, one, change the transportation complex of the entire world through autonomy, and two, this humanoid robotic line. It's massive.

Tesla also has this gigantic energy sector with batteries and solar panels, and all of those things are required for where we're headed to, I think Tesla has a monopoly on pretty much all of these pillars, and they are going to crush it in the year of 2026. In terms of maybe industry that I'm most excited about, I think the picks and shovels is still the move. The like Nvidia, TSM, ASML, Avago, like all of those, the compute, the foundry, the lithography, the networking. We are going to be building data centers so fast. And the people who are capable of putting those together are probably going to be the winners. But wait, let's go to, can I, can I, wait, can I, before we go, Can I tweak my answer on the industry?

I want to update my answer, Josh. Okay, what do you got? I'm going with the energy sector. I just realized none of these data centers or GPUs are going to work if we don't have energy. And I think we're going to make a lot of investment in energy grids next year. So I'm longing the energy sector. Big time. All right, what are you shorting?

Okay, I'm going to go with METO. Sorry, obvious, but I don't see them turning around a new frontier model. I think they made the mistake of spending way too much money for the wrong types of people to build the models. So I'm bearish 2025 meta. I'm also bearish 2026 meta. I think it's going to play out similarly to the metaverse, unfortunately. And this conflicts with me because I was super bullish with them earlier on. The sector that I'm likely to short this year is going to be neoclouds.

I think I'm bullish. I can be bullish GPUs, but bearish too many cloud, neo cloud providers that are supplying the GPUs. I think it's kind of bubbly. I think it's kind of pyramidy. I think they're going to collapse slightly. Okay, I think my choice this year, I'm shorting SaaS companies in general across the board. I am not a fan of any seat-based software with a weak moat. That's who I am shorting this year.

And I'm taking a basket of all of them because the switching costs are not very high to go from something like, and not to single out Slack because I think it's a great product, but to go from something that you pay millions of dollars a month for in something that, like, AI can probably automate with an engine and a couple of prompts. So I think SaaS companies as a whole who have been sitting pretty, making tons of money per month, selling millions of seats to companies, are going to have a very tough time when you're able to build applications so easy with these new AI tools that we're going to have for the next year. So I think that's probably my biggest loser. I'm long, big tech, short SaaS companies. And let's move on to the next one here. Okay, Cathedral of Compute. What do you think, Josh, will be the, who do you think will have the largest data center, or? cluster by the end of 2026. I'm taking the cluster part of this because I want to say XAI. Those guys are

absolutely crushing it. They have 200,000 coherent GPUs already. They are planning to push to 500,000, then 1 million. If anybody in the world can engineer a solution to do that, it is going to be Elon and the hardcore XAI team that stays up all night, every day, making this a reality. I'm team XAI. Elon haters won't like my answer. I'm also going with XAI purely because the proof is in the pudding. And you heard it from the man himself. Jens said, Wang. He has never met someone who is built and scaled data centers as quickly as Elon. I want to remind you guys of a very famous 2025 statistic that was pulled out this year that it typically takes you about three years to spin up something like a two gigawatt cluster and at least start

laddering that up. Elon was able to do it in 30 days. That is just insane for the premise of like a year's foundation. Crazy. Yeah. It's pretty incredible. Okay. So next one, a little, a little more more bleak maybe. This one isn't a fun one. This one doesn't get a gold star. This one gets a little, maybe like a red crossing sign. The layoff lotto,

you guys, which industry sees the most layoffs from AI? Okay, so this one kind of, my prediction is, it's going to hit close to home. I think knowledge workers are going to get screwed.

Interesting. So here's my reasoning behind this. A lot of focus of the new AI models that get released, they're focused on this one benchmark called a GDP Val. And that specific benchmark is focused on knowledge work.

And it's gotten really good. GBT 5.2 is chosen in 70% of cases right now by human experts versus human experts that can actually perform the same job. So I think this scales. I think by next year, knowledge works. So things like document writing, product strategy,

all that kind of stuff. It's going to be done by air. That's a good take. I think my take stems from a post that I saw from Andre Carpathy, actually. And he was kind of, describing how anything that can get verifiably measured can be automated by AI. These are things like call centers, like administrative work, like data input and output, anything that you can create a

verifiable answer and then train backwards against, you can replace by AI. So those are the first industries that come to my mind. But basically, if you have a job where there is a very clear outcome that you're guiding to and there is a reprogramable set of steps to get there, chances are an AI is going to be able to replace that very, very quickly. So I think that the goal, if you're a human being who's listening to this and not an AI, the goal is to be kind of like a polymath across a lot of industries. The dynamic range of knowledge is going to be very important as we move forward because part of the intuition that AI cannot replace

is just understanding lots of different categories very deeply and kind of connecting them together and the connection of those dots dynamically. It's intuition. You set the line. It's intuition. That's it. So if you have a intuition-based job, you're good. I want to push you. I want to push you because I want to just select a specific industry. That was a great answer. I love it. But which industry?

Well, yeah, the first one is like customer support. It would be call centers, but also customer support is a very subjective industry depending on the level of it. Like if you're working in hospitality, you're not getting replaced. People want human-on-human interaction. But if you're just like, you know, filing support tickets, things like that, things like administering report, data input, anything like verifiable is what I would be a little bit worried about.

Okay. I like it. All right. Moving on. Now, China in 2025 has dominated the open source model. So I'm going to question. They have the best open source model. So the question now is, will Trina retain the number one spot in 2026?

You over on the... I am slamming the over on this. China is open source world dominance. And the United States has no incentive to make an open source model. Meta tried. They got crushed. There was no benefit from doing it. And now they've pivoted to going closed source.

So there's no indication that anybody from the United States will release an open source model. We saw Open AI do it earlier this year. People used it maybe for half a day and then never looked back. China is innovating via open source. They will continue to maintain their dominance through 2026. What do you think? I'm going on the for one simple reason and it's political. I do not think the U.S. government and the U.S. in general wants U.S. founders building

on Chinese open source models. It's no secret that this is already happening in Silicon Valley. In fact, Brian Chesky of Airbnb integrates AI features into Airbnb. Guess what? He's used a quen derivative, a quen model from Ali-Bah. I think this trend is going to get increasingly worse and people are not going to want that for fear of China implementing a backdoor in one of their open source codes. So with all of that said, I think there's going to be an additional focus on open sourcing, just not all, but certain US models. And they're going to take the lead.

Okay. Well, next up we got the crash out catastrophe. This was actually your suggestion here. The A founder most likely to crash out. For example, publicly insults appear or even go to jail. Who's it going to be? Yes, yes. I had a specific individual in mind, and it's going to surprise you listeners, because you always call me an Elon fanboy. It's Elon Musk. I think Elon's going to crash out. I didn't say it was going to be a good or bad crash out, but I think he's going to crash the hell out because he's going to be super annoyed by something Sam Altman or Trump says. And by definition of a crash out, I mean he's going to yell insult or crash out on public news or on a public forum, probably X. So I've got him. I'm doubling down on Elon. Okay, I'm taking Sam Altman on this one.

He had a little issue. Yeah, yeah, yeah. If you remember the episode with Brad Gersner a few weeks ago, it was fairly recently where he just kind of had like, he was just like, hey, listen, dude, if you hate this, like, I could sell your shares. And he got very, like, snippy with him. And he's got a lot of pressure on his shoulders. They have a lot of debts and a lot of payments that need to be made. And things are going to get stressful. And I'm not sure how well he's going to be able to deal with stress in the public eye.

So I'm going to take Sam Olman on that one. All right. All right. Next up, we have bubble buster. This category is the AI bubble, most likely to implode. And note that I didn't say the entire AI bubble, because I don't think that's how it's going to work. But I think they're going to be implosions with smaller bubbles. Josh, what's your bet here. I think my bet is going to be open source models is what I'm going to pick here. And I think the reason behind that is because China is starting to develop an advantage when it comes to chip. for the first time. And part of the reason they've maintained their open source nature is because they have wanted to kind of, you know, accelerate things. And when things are open source, you could build on top of each other like building blocks. Now that China has started to get like Blackwell chips, they have very clearly defined like very good models, it makes sense for them

to close the door and to shut things down. And if China shuts down the open source models in the United States don't have any incentive to do so, I think the open source model industry probably gets crushed. There's going to be no leading models that are really excellent to use. it's all going to be closed source, closed gated in the race to AI. So that's my bubble buster. I like that. I like that. I'm going to go with Open AI partnerships.

I think that bubble is going to boast. Oh, the circular economy? The circular economy, but specifically for Open Air, who have signed to the tune of $1.4 trillion in payments. I think there is no way they're going to meet their targets next year or the year after that unless there's something written in the contract, which in some cases they are. But for example, the $300 billion deal with Oracle, no way they're paying $100 billion next year when they're losing $12 billion in generating that. So we're going to see some collapse there.

Whether they have IPOed in that time remains to be seen. We'll see. Okay. Well, next we have what topic that's very close to my heart, which is the breakout hardware device. Do we get a breakout AI device that people love, not including a phone? It cannot be an iPhone because, well, Apple can't even ship, but even if it was, do we get a breakout AI device? Um, yes And I know this is going to get under your skin, Josh

But I fully believe it Google Glass 2.0 Project ORA As they've turned it I think it's gonna I think it's gonna slap Because it can't be worse than Google Glass V1 They've watched Meta do the Rayband display

And absolutely flop They're not gonna put out something that people hate That's like saying Going to one prison is worse than the other Like they're just they're both suck Like being better than nothing is not a good place to be. I think those devices still do not get to a place where they're actually useful. Who you got? My breakout hardware device, which may be off by a few months in terms of timing,

but it's the Open AI Johnny Ive device. I am so excited. I think this is going to be the biggest event of the year next year in terms of how it's going to change the way that we interface with AI. This is the first time of companies rethinking the way that we engage with computers as a whole. It is the person who designed the most popular handheld device in the world, the iPhone. And now he's doing it again for Open AI, the biggest company with the most amount of users. And I think whatever device, whatever product they ship next year is going to actually impact a tremendous amount of people and really alter the way that we use hardware, where you will not need to be reliant on a cell phone in your pocket to navigate the worlds of the internet and AI. So I'm

excited. Is my tinfall hat on? I don't think they ship next year, Josh. I don't think they ship. They might not, but I'm going to be bullish and optimistic. All right. Moving on. Comeback Player of the Year. is someone who has been beaten down on in 2025, but they're going to make a big 180. Similar to how Google did this year. Josh, who you go? I got Apple, man. I'm stoked for Apple. I'm so excited for them. I think they're going to absolutely, no, I shouldn't say that. I don't think they're going to crush it next year. I think they're going to do much better than they did this year. Why? Well, because they're outsourcing their intelligence to Google. And that's what they, that's what they've always needed to do. They did this with their search results where Apple didn't create a

search browser, they just put Google default and Safari on the iPhone. They're doing the same with Gemini. I think Edge Compute is a huge thing to be wary of because if you can run AI on your phone, the inference charges are free to people who are building there. So there's a strong incentive for developers to build on iPhones. There's a strong incentive for Apple to offload their intelligence to Google and just focus on the experience but they're exceptionally good at. So I think if those things happen, Apple will have a really home run year next year as it comes to as relates to AI. Yeah, I think that's a good take. My take, and again, I'm wearing the tinfoil hat for this one,

cursor. Curser has been beaten down on this year because things like Anthropic Claude Code and a bunch of other competitors are just better vibe coding apps. But cursor has a crazy customer mode, and so many people still use it. They still might use Claude via it, but they love the cursor UI.

So my conspiracy theory, Josh, is cursor runs on VS code that is a fork of Microsoft. Microsoft is currently bleeding AI users. I think Microsoft acquires cursor and remolds cursor for their enterprise audience. And it slams. You haven't mentioned Microsoft

a whole lot in these episodes. We have not. Okay, so this one's totally not biased at all. Please ignore the humanoid robot to the left of this prompt. That is the Tesla Optimus. Most valuable robot, EJez. Which one is it going to be?

I know what answer you're going to give, so I'm intentionally going to give a different one here. And also because I believe they might actually pull it off. Figure. Brett and Cock, I think, is an awesome CEO. I have seen so many demos of this figure robot. And in my opinion, it's the only one that can go toe to toe with Tesla Optimus. I think because of the fact that they've been focused 100% on this for years now,

they've got a good chance of scaling this next year. I'm excited. Okay. My answer for most valuable robot actually is not Optimus. It is the CyberCab. I think Optimus will not reach the production scale required to actually create a lot of value. And they would likely just have early prototypes. By the end of next year, we will have cybercabs rolled out, hopefully across the country,

that are genuinely self-driving cybercaxies. And that is going to be such a mind-bending reality for a lot of people who've never sat in one before. So in terms of value generation, in terms of shock and awe, in terms of acclimatizing the average person to the world of AI that we're headed towards, I think the cyber cab is going to be the most valuable robot of next year. Love it. Okay, moving on, most popular AI entertainment app. Now, the idea of this is not the model, not necessarily chat GPT's app, but what kind of new breakthrough AI feature or product do you think people will be consuming? What's the TikTok of 2026? Josh, who you got? I think the most popular AI entertainment app next year is just going to be YouTube.

I think it's a continuation of what works. I think, YouTube is actually leaning into AI very heavily in addition to supporting short-form content built around AI in a way that I don't see a lot of other companies doing. There's probably more hours spent on YouTube than just about any other internet website on the planet. And them signaling that they're planning to lead hard into AI means that we probably get a natural extension. They're building where the people are. I think YouTube next year will be the most popular AI entertainment app. Josh, V-O-3 YouTube videos next to it? Dude, we'll sample some all the time. It's all. all it's all there. All the parts are there. Maybe some AI avatars of us. People already in the comments

think that we're AI avatars. How are you going to verify that we're not? Yeah, that's true. That's true. I might lean into it. All right. My big bet is, it's going to be a big year for the gooners. I think adult AI content is going to absolutely slap. I'm not saying it's good. I'm just saying Sam's already indicated that they're going to have an adult version of chat GPT releasing very soon. He said end of year. I don't thing that's going to happen. I think it's going to happen in 2026. And I think that video models are getting just good enough for the, you know, the AI adult industry to really kill it next year. I'm not happy about it. I just think that that's exactly what's going to happen. GROC companions are going to get the next level up. It's all going to get crazy. Oh, God, that's a scary future I'm not

looking forward to. Agreed. Okay, so here we have the largest model. Like, Victoria's sick. That's why you turn on. You know, it was a... That's funny. I hope you guys get that joke. That is so funny. Yeah, that's a deep cut for anyone who is around in 2021 on the internet. Anyways, the largest model, not the one walking down the runway, the one that is powering your AI tokens to your computer.

Which do we have for largest model? Josh, I have a general hot take on this. Are you ready? Okay. And also, how are you going to define the largest model, too? Okay, so in my opinion, largest model is like how many parameters the model has. So we started off with models with like 10 billion parameters, and then we skisks. all the way to a trillion parameters. If you're asking yourself what the hell a parameter is,

think of it as like the genetic makeup of a model. You're giving it the characteristics, what its eye color is, how intelligent it is, what its tone is, stuff like that, you know, stats, characteristics. I think parameters will not matter in next year. I think the thing that matters next year is post-training. 2025 has been all about pre-training. Oh my God, the most amount of compute that we give a model, the smarter the model is going to be. pre-training, pre-training, I don't think that matters at all. I think we're going to have a breakthrough in 2026 for post-training. I think people are going to spend more time with reinforcement learning and reasoning,

and that's what's going to give the level up for the models. So the resulting trend will be, models will still get slightly larger, but it won't matter. The resulting intelligence that comes from a slight model increase will be exponential because of post-training stuff. Okay, that's a pretty good take. I think in terms of, I'm just going to throw an arbitrary number for, parameter count. I feel like 15 trillion is a good number. 15? So X-AI with GROC-5 is planning to release a 7 trillion parameter model and that is coming

in Q1 of this year. It's going to be massive gargantuan model. So for them to double that in a year seems possible. Seems feasible. Because when you think about how quick, I mean, I mean, even from this year, we went from 03 to 5.2. And it was just like these unbelievable model jumps. And I suspect we'll get the same thing. Whether or not that matters is probably the more interesting question, like you were mentioning is like, do we even need a higher parameter count than that? Or is it just more on the like more niche post-training stuff? That remains to be determined. We will see. But anyways, next. Okay. Context window world records sticking along the theme of like size of model, what matters a lot is how many tokens or how many words you can prompt the model with. The more

tokens you can prompt the model with, the more context it has the smarter. The answer it gives. Now, 2025 was a record-breaking year. They hit, I think, million tokens. In some cases, unofficial, 1.5 million tokens. Josh, who do you think? What do you think the record's going to be? I think the limit does not exist. I think the context window expands to infinity.

And this is a very bullish take, very optimistic take. But there are a lot of researchers like safe superintelligence with Ilius company and thinking machines who are attempting, and even Google's trying to do this to remove the constraint of the context window altogether. And I don't know what type of magical moon math they're doing, but I suspect that next year will be the year where possibly we can get that. And what does that imply? Well, it can get this gigantic understanding of the world around you and refer to it precisely without a lot of fuzzy loss.

So there's going to be a lot of breakthroughs, but my hope is that the context window world record will be infinity by the end of next year. Okay. Wow, I mean, that is a crazy prediction. I love it. I don't have a strong take on a particular number, but what I will say in accordance to my previous answer, which is I think post-training is going to be really important. Context windows are going to matter a lot here.

So I'm bullish, large, large, large context windows next year. Okay, big context window year. Next up, we have winner take most, EJS. Does one model become the default for greater than 50% of the usage? Yes. Okay, so I'm measuring this, Josh, on amount of tokens processed. So, purely mathematical,

how, like, how used is the model? And I'm going with whatever the latest version of Anthropics Claude is going to be. Maybe it's Opus 4.5, but they're probably going to release Opus 5 or Opus 6. By the end of next year, coding is the most used contact. If you look at any of the coding measurements,

it's like 50% of the tokens that are being processed right now. I think they gain a market dominance on that. I'm going to call it. I'm going to take OpenA. and chat GPT as the model with greater than 50% of the usage because they have this dominant monopoly

where it's like something like 80% of the users. And I guess I'm kind of thinking more on the user front where it's slowly decreasing, but the question is how quick is that rate of decrease or how quick are other companies able to eat their market share. And it's been accelerating. We've seen Gemini take over some of it. We've seen Anthropic takeover some of it. But I think Chatchipitian OpenAI can sustain 50% for at least the next year. And then it makes sense that they eventually find a resting spot around

like 25, 30%, maybe slightly less depending on how many big winners there are. But I'm going to take Open AI as the model that becomes the default for greater than 50% of the usage, as it relates to users, at least. Okay. Next up, we have the biggest loser. Can meta keep the lead? Josh, who have you got? Can meta keep the lead? I think they can, but I don't want to bet against them. Because one thing that I've learned as I've invested in these companies are considered

investing in a lot of these companies is you want to let the winners ride. And there's only so many companies that can win when it comes to the scale that's required the amount of compute, the amount of money. And I think meta has a chance. They have a ton of resources. They have a ton of people that support them. I think the biggest loser of next year is going to be the bears. As a broad category, anyone who is bearish on the industry who thinks this bubble is going to pop for the next 12 months,

they are going to find themselves sadly mistaken and very badly hurt if they attempt to short this huge wave that we have coming up. Okay, I'm going to give a specific company and that company is Microsoft. They have already given revised numbers before the quarterly report, which shows that they have less customers using their AI model. in general on the AI features. This isn't a dig at AI in general. It's a dig at Microsoft's capability to ship good AI products.

I think Mustafa Salaman, the head of AI, kind of does a bad job at what he's doing, which might be a crazy take, but it's just like I don't hear about his updates. I hear about Demasas. And I think that the best thing that Microsoft has and will ever do in AI for now, at least,

is their investment in open AI. Nothing new with their company in general. Big time. Yeah, Microsoft has sadly been absent from a lot of these conversations we've been having. There's just not much going on over there. Anyways, the AI blue chip, EJAS. Who is the most valuable AI company by the end of 2026?

Anthropic. This is Anthropic. Wow, that's a big one. I think anthropic IPOs. I still think... That's big. I think anthropic IPOs. They're already rumored to do it.

They're engaging investment banks to set it up. You realize how big some of these companies are, right? Yeah, yeah. And Vividy is a $4 trillion. Oh, okay, I guess the way that I'm looking at it is like most ROI that you could get from like a public perspective. That was the way I was thinking about it. Not the largest market cap. Okay, well, that counts too.

Is that allowed? Is that okay? Yeah, I'm going to go. These are our rules, man. Okay, okay, okay. So I'm still going to go with Anthropic. I think they're going to IPO at a crazy valuation, but that valuation is only going to get higher. As more and more people realize that coding AI is the ultimate intelligence that you kind of want to own.

and Anthropic dominates. Okay. I'm going to say the most valuable AI company in the world by the end of 2026 will be Google. They will,

they're already close. They're right there. They will continue their dominance. They will continue to not only build CPUs for their own in-house compute, but we'll start selling them.

They have such a head start across so many pillars. They have the best models. They have unbelievable rate of execution. And if they just continue along this pathway, they're already trillions of dollars ahead of companies like Open AI.

And there's no reason in my mind, that they won't be able to outpace a company like Nvidia when it comes to just growing pure market cap. Like what a incredible business. For me, AI Blue Ship of the year is Google. And that brings us to the crown. Hmm. Who has the best model measured by intelligence?

So this is kind of the way we've been ranking the top number one AI models throughout the entirety of 2025. I'm split on this. Josh, I'm curious to hear your answer. Okay, I could give my answer first. I think, and this is probably a hot take, but I'm feeling optimistic, the answer to this question by the end of 2026, and I'm giving them two gold stars, is going to be XAI. I think they will have the best model measured by intelligence by the end of next year.

The reasoning is because they have been in pursuit of truth, and they have been in pursuit of synthetic data, where they are really trying to hone in the quality of intelligence that they used to train these models, and that is paired with the fact that they're able to build these clusters, larger and faster than everybody else. So it seems like their lead is going to start to make itself clear towards the end of next year because they're, honestly, they're going to be the ones with enough compute to actually run something that powerful. It's not that other companies won't get there.

They'll just be the only ones that can actually run it. Okay, okay, I see it. I'm going to go with Google Gemini, and I think I have a very strong case for this. It's funny, you mentioned synthetic data. I'd say Google are the leaders in world models, And I think world models are going to be the most important thing kind of next year.

And to help simulate kind of intelligence, they're winning that with Genie 3. I think they have nailed multi-modality. They've got the best video model. They've got the best audio model. If you combine all of those things into the number one, LLM, which is Gemini, and it all feeds back. You end up with the smartest model.

That's how they were able to catch up so quickly. I think Google takes the ground. And the final one, trendsetter. what's the hottest trend that isn't so hot right now, but we'll be hot next year in 2026? My answer is going to be the death of the text box. I think the trend that we're going to see next year, and this is an optimistic thing that I really want to see,

is LLMs or AI companies in general, moving away from the LLM, from the text box, from the way that we actually engage with these AI models. Basically, the entire interface is that of Google's in the early 1990s. It is a single text box. you write something, you expect answers. I think one of the hottest trends is going to be removing that friction from the experience when people engage with AI and creating more of a fully embodied ambient intelligence system where it predicts more of what you want. It kind of understands the

context better and it's not just limited to text in a box. And I'm expecting that to really evolve pretty rapidly as companies kind of fight to onboard more users. They're going to be inclined to want to give them more value. And in order to give them more value, you've got to kind of tell them how to extract it from these models. And that's the trend that I think is going to be hot. And that I'm very excited and looking forward to in the new year. Ejiz, what do you got? I like that take. I am going to double down on something I mentioned previously, which is world models. I think it is going to be the single most important type of model that goes viral next year, purely because it creates a bunch of synthetic data. Why is that important? We're running out of data. All these models are trained

on the same data. They're not going to get any more intelligent using the same data. We need new data. What is the best way to do that? Well, what if you could create a simulated environment of the earth and stick in your AI model there? It's how agents are going to get smarter. It's how AI models in general are going to get smarter. And Josh, it's how robots are going to get smarter. Tesla's already using it for their full self-driving AI model, right? They're using it to simulate car accidents so that, you know, they can figure out how to keep people. safe. So I think World Models is going to be crazy. I kind of cheated because Demis was really bullish on this on a podcast episode that I listened to recently. But hey, he's the godfather rare.

Could we maybe do the inverse of this too is what was the hottest trend in 2025 that will absolutely not exist in 2026? Do you have any off the top of your head? Okay, you go first. I got to think about this. Okay, because my answer, the reason why I'm asking is because I feel like death to the agentic browser is certainly a trend that we will consider to see. And that's one that I like a great amounts of joy from because the agenic browser just seems like an unnecessary complexity when I could have an AI go off and do the hard thing for me. And we talked about this with the CEO for Plexey Arvin, who was on the show earlier this year. And he was talking about how he thought it was important. But the reality is that there's two buckets. There's leisure and

there's productivity. And I don't need a browser for productivity. And I think that's going to be the trend that kind of goes away and never returns. Okay, Josh, I'm going to be hyper-specific on my answer because it really friggin't annoys me. Reservation. agents. I don't need you to book me a restaurant reservation. I don't need you to find flights for me. I'll do that myself. But if you can give me an amazing health stack or supplements based on my blood types and everything, go ahead. If you can automate 50% of my day job, go ahead. I don't care for the reservation agents that got released by Open AI. Google this year, kill them. I hope they never make it. I love it. And I think that probably concludes our

our 2026 prediction show. No, no, no. There's one more, Josh. What more? And this is the most important one. A surprise one? You and I have strived to keep episodes on average to 20 to 25 minutes this year. What's your over under on this episode being 25 minutes now? We are absolutely taking the under.

Every episode will be under 25 minutes no matter what. That is a hard cutoff. And we will absolutely not sway any higher than that. Yeah, there's no way that's not going to happen, right? Yeah. Certainly not. No, this is very serious. This is a hard limitless deadline.

We want to deliver the best 60 minutes of content every week. And that comes in the form of three episodes. And they will not be any longer, including this episode, which is probably already at minute number 40 or something like that. We've been going on for a long time. So maybe we should end it here. I think this has been the predictions episode. The task for anybody who's listening, this is very important, is you need to put your

predictions in the comments section. That way we can timestamp them and reference them next year. If you want to be top dog, if you want to be the Oracle who sees the future in a way that we don't or no one else does, then drop your comments underneath this video or this podcast episode. You could even post it on X, tell us what your predictions are, make a nice little thread, we'll share it with everybody. What do you think is going to be, you can even answer these questions. What are your predictions for next year? I want to hear it.

Let us know what you think. Definitely. And I'm going to let my ego enter the video for a second, Josh. we are less than a thousand subscribers off of 30,000. And listen, I hold no allegiance to the number 30,000, but it's a really nice round number that I would love to reach by the end of the year. So if you enjoy our episodes, if you enjoy our content, if you've enjoyed our predictions,

if you think we're going to be at least 50% right, I would call you crazy, but I appreciate it. Please like, please subscribe, please turn on notifications, give us a rating if you're listening to on Spotify or Apple Music, and we will see you. Josh in the new year. We'll see. We're going to see

when all this comes out. There was one last point I wanted to make, which is for the love of God, if you do anything, please watch this video because EJazz is wearing the most ridiculous hat. And if you're listening to audio only,

you are missing out. And I'm boiling. Tinhead over here. It's, yeah, so it's worth, it's worth watching. Spotify's my preference. YouTube is great.

But yeah, we'll see what the calendar plays out if this is the last year, the last of the year. Well, we're going to end for a pretty mighty 2026. But if it's not, we'll be back again, at least one more episode before

of yours. So as always, yeah, like you guys was saying, thank you so much for watching for being here with us and we'll see you guys in the next one. Peace guys.