Latest / In the Loop AI / 2025 AI Awards: Ranking the Best and Worst AI Companies This Year
Transcript
- Ejaaz: [0:03] Ladies and gentlemen, welcome to the first annual Limitless Awards, Ejaaz: [0:07] where we'll be crowning the biggest winners and losers in AI. As you can see, Josh and I dressed up for the occasion. I've got my tux, I've got my bowtie, and Josh has got his Hollywood gear. Josh: [0:21] I don't own a suit in this apartment. I don't have one with me. I don't have many suits, so I figured we'd go Hollywood. The problem is these sunglasses that I'm wearing are polarized and so are my displays. So I can't actually see anything. So I'm going to be going through the episode looking like this or something, but this was the best we got for our workshop today. Josh: [0:37] You have one formal, one a little bit informal, and we have a very exciting episode here. Ejaaz: [0:41] Amazing. Amazing. So 2025 has been an insane year for AI tech from someone marrying their chat GPT to Mark Zuckerberg spending $15 billion to hire one single person. There's been no shortage of drama and some insanely impressive progress in the space. Yeah. Josh: [0:59] So without further ado, let's get into the awards. I'm going to have to lower the sunglasses here so I can see what I'm doing. Welcome. This is our little award gala. Here's the visual. We put a lot of effort into preparing for this. This is kind of how it's going to work. So within this, we have a dunce cap and we have a crown. Those are the two things. You're either a winner or a loser. If you win, you get the crown. If you're a loser, dunce cap. Here are the nominees for all of our awards this year. It's just a series of all of the top AI companies in the world. And what we're going to do is start Josh: [1:25] with our first question, which is rookie of the year. So who is our rookie of the year? Maybe I'll start off first, Ejaz, and just kind of get into the flow of this. This is the breakout newcomer that went from who to oh, them. They are the guy that's now on the radar that no one knew about before. So I'm going to get over to my answer for this, which is DeepSeek. Deep Seek. And in particular, there are one model, which was remarkable this year. Prior to this year, I had never heard of Deep Seek before. I just... Like, I didn't know. And then this year, DeepSeek R1 came on the scene, and it absolutely shattered everybody's expectations as to what is possible. And it shocked the market that day, too, because it was down a considerable amount because people weren't sure if these scaling laws were actually going to stick, if you could spend more money on GPUs and if it actually resulted in more revenue. So DeepSeek really came from nowhere, threw a wrench in everybody's plans, and continues to do so as they innovate with restricted hardware. So Ijaz, I guess I'll pass it to you and ask, what is your rookie of the year? Who is your winner? here. Ejaaz: [2:26] Okay, drumroll, please. Josh: [2:28] Oh, and also before you get the answer, I just want everyone to know we haven't reviewed this with each other beforehand. Yes, I have no idea what. Ejaaz: [2:33] Josh is picking. All right, drumroll. For my winner, it is Kimi K2. It is a pro-China consensus today, it seems. Kimi K2 is the other frontier Chinese AI model lab, which open source their insanely impressive. I think it was 750 billion parameter model. Josh, I think DeepSeek and Kimi K2 are like the kind of leading companies in China that have released their AI models. What impressed me the most about Kimi K2? Well, there were two things specifically and why they gave them this award. Number one, they barely trained their models on kind of redundant, I think, A100s from NVIDIA, which was insanely impressive to reach frontier intelligence. When they released their model at the time, it was the best or competed with the best at the time, which I think was GPT-5, insanely impressive. The other really cool thing was how they did it. They used a mixture of experts model, which basically didn't use up the entire model parameter at one time. And so I have to give them the award for creativity. Like that team is just insane. I think it's literally only about 15 people in that team versus the hundreds of people that work at Google and OpenAI. Insanely cool. Josh: [3:43] So the Chinese companies, Rookies of the Year, both of them across the board. And then we have some honorable mentions. I know StarCloud and N of One are both two top contenders for this. Ejaaz: [3:53] Yeah, yeah, yeah. Okay, cool. So StarCloud is a bit of a rogue shout, but I needed to mention them. They were the company that started the AI data centers in space, Josh. If you remember about a month ago, we broke this insane story around this team from Y Combinator that is trying to launch a data center in space to train AI models. they ended up doing that just last month. We put out a whole episode on it. Definitely go check that out. And then I had to shout out N of 1, Josh. I'm so tired of AI benchmarks, man. And this is the only benchmark that got me excited. It is the benchmark which measures can an AI model make money? Ejaaz: [4:28] And maybe there's a little more on this later in this episode. Josh: [4:31] Well, next up on the list, we have the first dunce cap that we're going to be allocating to someone. And that is for the biggest flop or the most overhyped thing. It was the loudest launch with the quietest retention this year. And Ijaj, you're at first. Ejaaz: [4:43] I feel like there's a lot of contenders for this. Josh: [4:44] There's a lot of flops this year. Ejaaz: [4:47] I feel like there's so many. Okay, okay. Okay, drum roll, please. Here we go. Mine is meta. And can you guess, Josh, specifically what from Meta flopped so hard? Josh: [4:59] Oh, man. The loudest launch with the quietest retention. Ejaaz: [5:03] Come on, come on. Okay, I'll give you a clue. You're wearing it on your face. You're wearing a volume. Josh: [5:07] Oh, the glasses, the Meta Ray-Ban displays, the abomination. Ejaaz: [5:12] The thing that I hated the most. Josh: [5:13] Oh, my God. Yeah. And I was so bullish. Ejaaz: [5:16] I was so bullish on it. Of course, we're talking about the Meta Ray-Ban displays, which was meant to be this bleeding edge AI form factor where it would portray all your apps, social media, whatever it might be on the lens of your screen. So you can become this superhuman that doesn't need to pull out a phone or screen from their pocket. Obviously, it ended up flopping so badly that some of their biggest supporters wrote outward criticisms about it, saying it was one of the worst products they've ever tested. So that's my answer. Josh: [5:44] Josh, what have you got? Mine is three companies here because it is tied around one theme, which is the agentic browser was the biggest flop. Everyone thought it was the biggest thing in the world, led by Perplexity, really, who transitioned their entire company towards the agentic browser. We had the CEO of Perplexity on the show to talk about it, and it left me feeling a lot of desires that were never really met. So that's why there's this cluster of three. We have ChatGPT, they made their browser. Anthropic, they made a browser extension. Everyone was so, so excited about this. And from the day it launched until now, I have not seen any demos, any examples, and people are still using Google Chrome. And if they're feeling ambitious, the browser company. So to me, the agentic browser, biggest flop of the year. Ejaaz: [6:29] And who do we have as honorable mentions, Josh? Josh: [6:32] Oh, my goodness. Oh, MetaVibes is a good one. If you remember MetaVibes, you might not because it only lasted about 48 hours, But that was their, that was Meta's response to the viral AI video feed, where it was a video feed similar to TikTok, but it was only for AI. And then we have, oh my God, the friend pin. The friend pin is funny. Most people watching this might not even know what the friend pin is. It's this little pin that sits around your neck. And... It listens to you. It's supposed to be AI. It doesn't do it that well. And then on the topic of hardware, humane. What an abomination humane was. Ejaaz: [7:05] How much did they raise, Josh? They raised like hundreds of millions and they were valued at $10 billion at one time. It was like an ex-Apple Hardware CEO. One of the worst products I've ever seen. Josh: [7:17] They tried to be John DeIva and Steve Jobs in their reveal trailers. And that was the pin that had the lasers that would project stuff on your hand. Really just brutal failure. And this year they were actually acquired, I believe, for probably pennies on the dollar. Sad story. Josh: [7:30] Hardware and AI is difficult. But alas, we carry on. Ejaaz: [7:34] On to the next one. Okay, the next category is best comeback story. This is going to describe a team that was left for dead, that came back swinging with the ultimate glow up in 2025. Josh, who you got? Josh: [7:49] Best comeback story for me is Google and Gemini. I opened up this year thinking oh my god Google is dead they had built the transformer they had all the hardware they had all of the people but they could not actually execute and deliver anything of value I didn't use their products now and we're going to get into the biggest winners but Google is a force of nature and they really turn the ship around and you could track that directly to when Sergey Brin started coming back into the office so founder is back in the office. Google is turning it around. They were the team left for dead that came back swinging with a crazy glow up. Ejaz, yours? Ejaaz: [8:29] Let's reveal it. Josh: [8:31] Boom. We matched. Ejaaz: [8:33] We have consensus on this one. Similar to you, Josh, I couldn't stand the company for the longest period of time. And I remember when they released their first image model. Do you remember this, Josh? And you could ask for it to produce a picture of the American forefathers and it produced a really inaccurate version of what that was. And I was like, Google's cooked. They have all this money. There's no way they're going to make it back. And they did the biggest 180 ever. And I have been eating my words since. Hugely impressive. Honorable mentions. Okay, I have to take the first one. I put out a thesis on this in our newsletter, which if you're listening to this, you should definitely sign up for. It is awesome. We drop alpha every single week. Amazon was a sleeper hit in the AI company category this year. I feel like they deserve a shout out. They're training their own models. They're building their own AI chips. They just signed a deal with OpenAI to the tune of $10 billion, whether they're going to use their chips, just a sleep ahead of a company. Josh: [9:32] Yeah, and then the other one here that we have is XAI. They shipped probably faster than anyone else on the planet this year. They went from nobody to frontier model in the matter of 12 months. And that was remarkable progress, mostly led by the engineering team that was Josh: [9:47] able to get all those coherent GPU clusters together and really just put on a show. So that brings us to the next question, which is the model most likely to make you rich. This is the model that you give your money to and then you say a prayer. Ejaz, what did you pick for the model most likely to make you rich? Ejaaz: [10:05] Josh, I thought long and hard about this one. I was like, which model would I give a thousand bucks to and trust that it could turn it into a million dollars? And I base this on really factual data, Josh. So, you know, I want you to take my opinion with a lot of grace here. Naturally. It is... Josh: [10:23] Croc 4.20. Ejaaz: [10:26] 420. The unofficial but official model that is going to be released in the next couple of weeks, which got tested kind of like proactively during the end of one benchmark test. Josh, it made a total return of about 12.5% in two weeks on $10,000. And in some cases it returned 40% in one of the instances of the competition going on. I just think it's a super cool model. If you remember from the first iteration of that competition, which we made an entire episode about, we made episodes on all these, definitely go check it out. It at one point was up 50% and it made five grand on the 10 grand that it had made. If I was to trust a model, it would be Grok because it measures my risk taking. It matches my risk taking rather. It is a savant in that case. And I like to roll the dice. How about you? Josh: [11:13] Well said, my friend. We're aligned. I don't think I have anything to add to that. I want someone with the most information who is bold and brave and believes in the same things that I do. And XAI, I mean, they're not going to twist its words. It's going to tell you how it is. It's going to bet straight. And it's going to do so better than anyone else based on our few data sets that we have. So we do have a few honorable mentions. There's three of them here. Do you want to start with DeepSeek? Yes. Ejaaz: [11:35] Okay. The company that started it all, because they were a hedge fund out of China that created the model DeepSeek. And so it didn't surprise me at all that at one point in the trading competition, they were the top performer, Josh. Do you remember? It was like a calmer, more calculated version of Grok, less risk-taking. So definitely deserving a shout out. I see we've got Kimi here as well in GPT 5.1. Josh: [11:59] Dude, the Chinese models, man, that's where the quants live. deep seek and gimme if i if grok 4.2 didn't exist they're getting my money those guys built it hardcore mathematicians engineers they were the quants they were the early leaders in the first version of that trading and then gpt 5.1 as well was an early leader until again grok 4.2 came along and just smashed everybody out of the water. Ejaaz: [12:20] Oh yeah okay so if models making you money doesn't enthuse you we'll we'll do Ejaaz: [12:26] it the the old school way we'll look at the biggest ai stock or the best AI stock of the year. This is the company that you buy, set it, forget it, and you make money. You look back in a few years, I'm going to up massive amounts. Josh, I'm curious who your winner is here, man. Let's see. Josh: [12:44] My AI stock of the year again is Google. We're starting to see a trend here. Google world domination. They went from the underdogs to the top dogs. Their stock is up 60 this year which far outpaces most other companies that are publicly traded and it's just been this remarkable story where they've they've hit their stride and they're able to vertically integrate across the entire stack they have hardware with the tpus they have software with gemini they have image generation they have physics understanding they just have everything you could want in an ai company and they are moving full force across the spectrum so ijaz. Ejaaz: [13:20] Let's see if we go two for two. Dust AI sucks. Let's have a look. Josh: [13:23] Wow, it's like we're on the same podcast co-host together or something. Ejaaz: [13:26] Okay, okay, okay. I have my reasoning behind this. Josh: [13:29] Okay, what's your reason? Ejaaz: [13:30] My reasoning is I think if I was to chuck in all my money, if I could only invest in one stock and I would have to lay out the probabilities of this company still surviving and being number one, it has to be Google. They own all the distribution, Josh: [13:46] Josh. Ejaaz: [13:46] They own the TPUs themselves which would compete against NVIDIA. And they own some of the best models. How could you not go with Google? It's pretty obvious. Nothing you haven't covered. Honorable mentions here. Josh, we have Tesla, number one. This was a sneaky one that I wanted to include because whereas they may not explicitly have an AI model, Josh, their full self-driving, by the way, we made a sick episode on this yesterday. Definitely go check that out, is powered by a general neural network. And it is so cool. And I think it'll be one of the coolest, if not most profitable AI models, maybe not directly to the consumer, but indirectly through their sales of Tesla cars in the future. So needed to shout them out. Josh: [14:26] Yeah. And we're going to have a predictions episode coming soon also. Josh: [14:30] And I have a feeling we'll be seeing a lot more of Tesla in that episode. But anyways, our next prompt, our next award that we're giving out is the Moat Builder Award. Looks defensible until you ask, so why can't I just switch? Ejaz, who is your Moat Builder where we're going to? Ejaaz: [14:47] I'm super curious. Let's find out for all. Let's find out. Josh: [14:50] Oh, Gemini again and Google. Oh, my God. Ejaaz: [14:54] Okay. Hear me out. Hear me out. I know I sound like a broken record at this point, but they are the only company that has a dominant number one or number two position every layer of the AI stack. You could choose an app and you're like, oh, wow, browser, you've got the Gemini app, you've got Android, they own all of these things. Then you could go down to the model layer and you're like, they own DeepMind. They have Demis Hasebis. He's so sick. He creates some of the best research. They've got science models, all these different kinds of things. I think Google has the stickiest mode. And I think that's their best quality and also potentially the worst quality for everyone else. Because if Google can kind of like outcompete you and like not turn on ads for any of this stuff, how can they lose? Mote is very important. Google wins. Josh: [15:40] Who have you got, Josh? Well, my Mote builder, finally we have a different answer here, is ChatGPT. Okay. ChatGPT is the stickiest app for me that has existed, period. Because of the memory, it has so much context of our conversations that whenever I need some helpful resource that can reflect that context, ChatGPT is the easy choice for me. They've built such a sticky Mote because I don't want to go anywhere else. When I talk to Grok, It doesn't know everything that ChatGPT does, and it adds so much value having that memory. I find that it has been the biggest moat winner of the year, for me at least. And then honorable mentions, we have two here. The first being Cursor. Which seems pretty self-explanatory. I mean, Cursor is the best coding model or coding, what is it, an IDE in the world? Ejaaz: [16:24] IDE, yeah. They were the best, Josh. They were. They were. They were the best. Josh: [16:28] They were. But they built this incredible product that people were using to code. Of all the products that were built, this one probably contributed to the most actual economic creation. Yes. Because all of the code was written. And then the second one, you just have perplexity. You want to explain why? Ejaaz: [16:41] Yeah. I had a feeling this would be one that you would hate, Josh. But in Perplexity's defense, it is an awesome search engine that I find a lot of people that may not be you or I, but like are the average consumer that kind of just wants to interact with AI seem to love. They have the most trending posts on X from their Perplexity point. Joe Rogan reps it on all of his podcast episodes. He uses it live. Jamie's about to get his assistant is about to get replaced with Perplexity. Ejaaz: [17:09] I just think it's a cool product. Worth a shout out. But moving on, we have, Josh, this is probably the category that had the most amount of nominees. Biggest waste of money. Josh: [17:23] Congratulations, you played yourself. Ejaaz: [17:25] I'm sure you're listening to this and you're thinking, Josh: [17:27] Oh my God, where's the money? Ejaaz: [17:28] What do you mean waste of money? There's no waste of money in there. Unfortunately, there is a lot. And Josh, I'm eager to know who your winner is. Josh: [17:36] It's tough to pick the biggest loser because there's so many. The biggest waste of money. I chose Meta for wasting money because, oh my God, they spent so much money on talent and got absolutely nothing in return. Meta has failed across the spectrum. But when it comes to poaching people, the expectation was that they would deliver something. And so far, EJ, if you'll remember, they spent, what, $30 billion on employees, probably more now, just on hiring individual employees. They were acquiring entire companies just to get an employee or two. And then the first product they released from that was MetaVibes. And it's that like TikTok feed that no one ever used. And I'm not even sure we tested it because I didn't even know how. So Meta, they spent an unfathomable amount of money on talent and have absolutely nothing to show for it at the end of 2025. Ejaaz: [18:22] Okay, are you ready for mine? Josh: [18:24] Please, let's see it. Ejaaz: [18:26] Okay, it's Meta as well. If you look at, how do we quantify this, Josh? If you look at the most amount of money spent per kind of like redemption on that amount spent, like in terms of value, Meta sucked so hard. To spend $15 billion to hire one guy which ended up not releasing a model within like a couple of months of you buying him, like it is just the most insane and stupid amount of spend. You add on the hardware, the Ray-Bans, I can't get over the glasses, sorry. You get over the Meta Vibe app. Their resource allocation has been absolutely horrendous. Meta takes the crown on this one. Josh: [19:03] I'm sorry. Wasting a lot of money this year. We have a few honorable mentions here that are quite funny. Do you want to talk about the first one? Ejaaz: [19:09] Okay, right. Okay, we know you guys love or hate, rather, the circular economy. And that is the instance of everyone investing or forming a partnership, a multi-billion dollar partnership at that with OpenAI in exchange for the compute that OpenAI is going to reinvest that money to spend on that exact same company. And it boosts everyone's valuations. And everyone thinks this is a massive Ponzi scheme. I don't know why you would think that. Definitely an honorable mention to make out here. And then we have social feeds here, Josh: [19:39] Josh. Yeah, social feeds, they were just a big L across the board. Open AI released Sora, which was hot for about two days to a week. Now I haven't generated a video in a very long time. Again, the MetaVibes app, I'm not even sure where to find that if I wanted to try it. Josh: [19:56] Anyone who invested in social feeds this year, they kind of lost. It was kind of a waste. But anyways, moving on to the next one. This is also a dunce cap that we will be allocating to the most unhinged AI drama. This is boardroom chaos, lawsuits, leaks, subtweets, just the peak Silicon Valley telenovela. Ejaz, who do we got for most unhinged AI drama of the year? Ejaaz: [20:18] Dude, I have an interesting one. Tucker Carlson calling out Sam Altman on his podcast, accusing him of murdering a whistleblower that left OpenAI, I think in 2024, 2023, I can't remember which year. This, of course, relates to the story where you had an OpenAI employee that left, leaked some secrets and ended up dead two days later. Really mysterious kind of suicide where he somehow figured out a way to, you know, end his life himself. But Tucker had or was bold enough to bring this up directly with Sam in an interview where he was just talking about things like gpt5 and stuff like that and suddenly this came out of nowhere it was the most awkward 60 seconds of my life which sam oldman replied josh do you know what he said are you accusing me of murder tucker it was it was so it was so awkward dude yeah Josh: [21:12] That's pretty bizarre it makes you feel uncomfortable if you want your like skin to crawl you can go watch that clip i chose this sick of fancy, phenomenon that happened with chat gpt and open ai chat gpt removed 4-0 as a model briefly and the entire internet had a meltdown because that was who people had an actual relationship with eej as you mentioned at the top of the show someone got married to their ai model to 4-0 people were going on dates people were using it as their companion their emotional support like this is incredibly unhinged because it feels real it's the closest thing to this like real dystopian future that i've seen this year so that's why it feels unhinged to me is like people are literally marrying their models and open ai pulled one and the backlash was so extreme that they added 4-0 back into the chat gptf so people could continue their relationship so for me the sycophancy phenomenon that was the most unhinged drama of the year we do have some honorable mentions here windsurf yeah. Ejaaz: [22:10] Oh my do you remember windsurf okay also another ide and integrated development environment that competed very heavily with cursor they were worth i think seven to ten billion dollars at the top and then they got gutted josh over the weekend they had two acquisitions which might sound weird but the first acquisition was effectively four of their top employees being sold to i think it was Google or Microsoft. I can't remember which at the time. And they spent like $10 billion to acquire them, leaving the rest of their employees that were 200 plus employees out for dead. And then they ended up getting acquired by Cognition, which was the AI engineering development team. Just a crazy ton of events. And I had to shout out the meta AI feed as well. Just an horrendous display of product insight here, Josh. They released a social feed, which revealed everyone's AI prompts using the meta AI assistant the one twist they had no idea it was happening so you would have deeply personal prompts being revealed to a gabillion people every single day insane I'm adding Josh: [23:16] An additional one as we speak which is opening up for profit, They finally converted from a non-profit to a for-profit after decades of insisting we are not here to make money. That was pretty funny. I feel like that deserves an honorable mention. Ejaaz: [23:29] That's a good one. I'm glad you included that. All right, moving on. Next category, the Monopoly Man Award. Who owns all the chips? Who owns entire categories or layers of the stack in AI? Josh, who you got? Josh: [23:42] This is a recent development, but SpaceX is the Monopoly Man. They own all of the capabilities to get to low earth orbit and in a world in which we are moving ai data centers to space they own the entire through way they own the highway they are the toll gate preventing anyone from getting into low earth orbit and training these ai data centers in space and if people are as certain as they seem the leaders across these ai companies they fully intend on doing this and spacex is the singular person that will enable that to happen so spacex has all the chips when it comes to this new frontier of getting off of the earth into orbit for these new AI data centers. Ejaaz: [24:20] That's a good one. Okay, here's who I got. I went boring. I went old school. I went with the big dogs. How can you not look at NVIDIA that own currently, I think, 90% of the most valuable commodity on Earth, just GPU. They are booked out 6 to 12 months in advance for future GPUs that they haven't even created or released just yet. Josh: [24:44] That's so crazy. Ejaaz: [24:44] In fact, Elon's latest and greatest data center, Colossus 2, is going to be running their latest Blackwell and potentially Rubin chips as well. They are the best of the best. I don't see a world where they die or lose their monopoly anytime soon for one simple reason. People can't get enough of this AI stuff and they're going to need to buy more and more and more GPUs. Honorable mentions. Who have we got here, Josh? We've got TSMC. Josh: [25:07] TSMC is one that I'm super excited about because, man, they are the only people that can do this. When it comes to making chips and creating these like lithograph, these very complicated machines, TSMC basically has a monopoly on them. So if anyone's making chips, TSMC is doing it. No one's been able to figure this out. There's a few companies in America that are trying, but still they have the monopoly. And then the other one I just put is the United States electrical grid, because my God, that is the single bottleneck. If they turn that off, if they restrict that to AI data centers, everything comes to a halt. So they very much more than anyone, I think the electrical grid and the ability Josh: [25:41] to generate energy is the single largest monopoly to whoever holds that. But moving on to the next one the crown and this is when things start to get big this is this is the part of the show where we're handing out the larger awards so maybe what we could do we'll make this crown just a little bit bigger here because. Ejaaz: [25:56] Thank you this is the model Josh: [25:58] Of the year the model you'd pick if you could only keep one in your toolbox for all of 2025 ijaz who's your model of the year okay. Ejaaz: [26:08] I before i reveal i thought a lot about this one josh because there are many models that i use and there are many different uses of why you would use a particular model. And the one that I kept landing on that I couldn't get over was... Claude Opus 4.5 from Anthropic. This is a new model that they released a few weeks ago, but Josh, it told me a few things. They will always be the number one frontier developer of coding AI. And in my opinion, coding AI is the only AI you will inevitably end up needing. It is the magic wand that you can wave to create whatever you want. In the future, that's going to become increasingly more digital. It has laid waste to a ton of employees at various different companies. It's going to do so much more in the future, but it also gives one single individual the power to create whatever they want. It gives them the power or the army of 100 people, which I think is super cool, really underrated. They were the dark horse of this year, Josh. We didn't really talk about Anthropic, but they came out of nowhere and I'm utterly impressed. Josh: [27:11] Yeah, not until recently. And again, this is more of a recent development. My answer to this is Gemini. And that's kind of the Gemini suite of models. So I am cheating a bit, But Gemini 3.0 Pro. And the Gemini suite of AI has this understanding of the real world that you just cannot get anywhere else. And I use it to generate images. I use it to generate videos. I use it to do research. I use it to ingest large papers or books that I want to understand through Notebook LM. And the Gemini model is the, it feels like the smartest brain that exists. And if I could only use one AI model through all of 2025, a majority of the time I'm using it to learn new things. and Gemini is the thing that I want to use to learn new things with. And that's why it shows Gemini as the model of the year for me. But we've got a few honorable mentions. Ejaaz: [27:59] Just a few, just a few. I mean, like I said, there was a lot of nominees for this. So a few honorable mentions. Grok, Josh, if ChatGPT and Claude are two of the models that I use the most, Grok is the closest number two or three that I could ever say. Why? Because it's in the main social media app that I use every day, Josh. If there's something that I read or don't understand, I tap that explained by Grok button and it explains to me. And then I have a conversation with them. Before you know it, I've had like 10 different conversations throughout the day. Awesome, awesome model. Josh: [28:28] We don't have an award for this, but it might actually be my most used model of the year. That was the one that I probably interacted with the most because Grok to me has become like Google, where whenever I just want to quickly search something or learn something, Grok is the way. Ejaaz: [28:40] And then we have DeepSeek. We shouted it out earlier today, but just an amazing Ejaaz: [28:44] open source model. They led the open source frontier. It's what led to NVIDIA market are collapsing 5% at the start of this year with their 2024 end of year release absolutely insane and i had to shout out alpha fold and genie from google uh which have got nothing to do with llms but everything to do with science and breaking frontier simulation theory more on that later but that was awesome and for our next category josh we have product of the year this is a product or a feature an ai thing that normal people actually used on purpose and kept using. That's the most important part. Josh, who have you got? Josh: [29:21] This is a big one. Again, let me increase the size of this crown here because product of the year is no easy feat. Everyone's trying to build the best product. Ejaaz: [29:28] I think we can go a little larger on this, Josh. Come on. Okay, we're good. That's a big crown. That's a big crown. Josh: [29:33] Who's the product of the year? My choice is the OpenAI application. It is so good and it is used by 800 million people per week which is an outrageously large number of people that it's supporting so not only is it an excellently built app it's just smooth it works it has all the functionality baked in whether you're on your phone whether you're on desktop it is supporting so many users on a weekly basis that it feels impossible not to nominate it it's where i go for voice chats it's where i go to interact with anything that has my memory it's just a hot key on my keyboard it's a shortcut on my phone. It's everywhere. For me, it's the product of the year is ChatGPT. Ejaaz: [30:15] Okay. You ready for mine? Josh: [30:16] What do we got? Ejaaz: [30:17] I went with also OpenAI, but specifically OpenAI Pulse, which is the personal AI agent that pings me every morning with the latest and greatest news, financial insights, or AI stuff that I want to get into. And the best part about this, Josh, is that it can read my mind. It pulls all the information that I want to know, it helps me with also a lot of other random stuff. Like it knows what my workout routine is going to be on Friday and gives me some prep and advice in Foresight for that. I just think it's an awesome product and I can't wait to see what it turns into. Josh: [30:52] That's a winner. And then, oh, these honorable mentions are good. First, Claude Code, again, probably increased the actual productive output of the world by a non-zero percent, which really says something because the quality of code this year and the amount of code that people have been able to build is so high. But the one that is near and dear to my heart that I probably would have picked if I couldn't pick JATGBT is NanoBanana Pro. It's one of those models where it doesn't make sense. It feels like it shouldn't be this good this quickly because of how accurately it understands the real world, how accurately it's able to represent things. I mean, we had an episode last month where it covered my wedding. It created me an entire artificial wedding. It was so cool. Like Nano Banana, it's one of those things that it's like you see it and it feels like magic. Like I genuinely do not understand how it works. Ejaaz: [31:40] So for those of you who missed that episode, Josh officially got married to none other than Kendall Jenner. Please send your congrats in the comments below. Thank you. Josh: [31:48] She doesn't know it yet, but I'm very happy about it. Ejaaz: [31:50] We know about it. Yeah, yeah. Okay. Moving on to the next one. Josh: [31:53] Okay, this one is fun. This is the biggest loser of the year. Okay, don't cheat and look at the bottom one, but this is the biggest loser of the year. The biggest gap between resources and results. Ouch. This is the person who just continually lost over and over and over again, and no matter how hard they tried, they could not figure it out. Ejaz, who is your biggest loser of the year. Ejaaz: [32:16] There were quite a few options for this one, but I went with... Josh: [32:19] Apple. Ejaaz: [32:20] Oh, familiar? Heard of that company, Josh: [32:23] Josh? That one kind of hurts me, honestly. I know that's personal. Ejaaz: [32:26] Listen, the company that we spoke the least about this year, and granted, this is an AI-focused podcast, was Apple. They didn't get any mention because they didn't really do anything. They've been delayed over two and a half years for Siri AI. They are about to engage Google, Google's AI model for their own AI product. Absolutely crazy. They don't own their own AI model. Insane. And they've lost probably 80% of their AI talent. It's been absolutely gutted. They're probably going to have a CEO changeover in the new year. Apple's just handled this horrendously. And I don't know how they come back from this, Josh. That's fair. Josh: [33:02] That's fair. I hope they turn it around. That'll be for a predictions episode. But for my biggest loser, I had to choose. This wasn't even, it didn't even feel like a debate for me. This is meta. It's just like so obvious to me. Meta has tried and failed at literally everything since Facebook. Like meta the company the name the reason it's called meta is because they pivoted to the metaverse and you just where is the metaverse there is no such thing it doesn't exist yeah so like the name of your company is literally the failure second to that they hired, billions tens of billions of dollars of incredible talent to build these frontier models they never even shipped the big one they only shipped the two little ones that no one used and they never even got to the third big one that they had announced and third they were trying to make hardware work they could not make hardware work the demos were an abomination one of the most embarrassing things on stage that i forgot to mention earlier is when they they did the meta ray-ban display glasses and they blamed on the wi-fi the fact that none of the demos worked and nothing worked. It was just terrible. You're not wearing them on your face, Ejaz, even though you wanted to. I wanted to try them. They just suck. Meta continues to fail. I hope they can figure it out next year. But that's my biggest loser of the year. Anyways, for Adderall Mentions, who else lost big this year? Ejaaz: [34:17] Right. This is a tongue-in-cheek one, but we had Michael Burry, who famously shorted NVIDIA, and I think it was Intel as well, a few other companies, and then closed down his fund the following week when his position was down super badly. He made a bet, a very logical bet, some might say, that we are in an AI bubble and it is about to burst. But he was proven otherwise because there is just insatiable amount of demand for AI products. And so you need more GPUs. Big, big L. And the other one, Josh, this was your suggestion, Josh: [34:49] The bears. Down with the pessimists. Down with the bears. If you're going to bet against this wave, you better buckle up because you're going to get liquidated all the way up. And that's what happened. I mean, we were looking in preparation for this episode. We were looking at the gains from some of these stocks this year, Josh: [35:04] and they were like 60, 70, 80% gains for... Large cap stocks i mean google was 60 intel i think was like 70 or something outrageous like that we had nvidia was 24 amd core weave both 60 to 70 if you were bearish on any of these companies you got absolutely washed and you missed out on one of the best years in stock market trading history so yep sucks to be a bear in the bullish time ever and that brings us to the final okay The AI company of the year, Ejaz. Ejaaz: [35:36] This is the biggest category. The only category that matters, some might say. AI company of the year. We're going to do this slightly differently. We're going to announce the nominees. And then we're going to have a very elongated drum roll because it's very important. Josh, we need to make that crown a lot bigger, by the way. Josh: [35:53] Yeah, I'm on it. Ejaaz: [35:54] Nominees. We have Anthropic for their crazy coding AI. We have XAI for their insatiable social grok model. We have Google for the all-rounder Gemini 3 Pro. And finally, we have none other but OpenAI who managed to pull a bunny out the hat at the end with GPT 5.2. Josh, who is our ultimate winner? Josh: [36:19] So I got to shrink the crown back because the winner's hidden right here. Our ultimate winner. Everybody, please, drum roll. Best AI company in the world this year in 2025. Congratulations. This is no easy feat. This is a culmination of so much hard work in the most competitive industry in the world and the winner goes to google congratulations it felt right it couldn't go to anyone else it had to be google round of applause bravo you did it you've succeeded you have built a monster in the most competitive place on earth and have done so with grace and elegance and a deep understanding of physics and for that we applaud you congratulations organizations the comeback company of the year the company of the year google has just dominated across the board and they deserve this i feel really good about this choice. Ejaaz: [37:08] When I think about why Google was able to pull this off, Josh, it brought me back to that Demis documentary that released a few weeks ago. I think it's called The Thinking Game. The biggest takeaway from that was he just gamifies every single problem and then slaps an AI model to train on that problem over and over and over again until it gets better. It's what he did with the early AlphaGo model. It's what he did with plenty of other, it's what he did with science. And now he's gonna apply it to fusion reactors. He's gonna apply it to material sciences and he's gonna apply it to anything that he can get his hands on that is computable. Google has had that mindset from so early on. Josh, do you remember a decade ago when they announced their investments in TPUs? You might not. I certainly didn't because I was just a teenager doing stuff that didn't concern any of this stuff, but he had the foresight. just amazingly led, absolute crazy killers that work on this team. Josh, you and I were talking about this off camera, but we realized All the major players, all the major people in AI at some point stemmed from Google. They either worked at Google, they either interned from Google or worked in research jobs with Google. They are the start and potentially the end of all things AI. Insanely impressive company. Josh: [38:20] And that pretty much wraps up our best of 2025, the biggest winners and losers. What my ask for anyone listening is to share your biggest winners and losers from this list. Let us know who you think we got right who you think we got wrong who we just totally forgot and why we may have forgotten them the next episode that is coming up is the predictions episode this is going to be what we think is going to happen in 2026 and that to me feels even more interesting because with the amount of progress we've gotten this year assuming it continues it's going to be a pretty wild year of progress in ai but that concludes our episode we won't hold you any longer thank you for watching please share this with a friend who you think this might either trigger or satisfy because we're here to do both and don't forget to subscribe on youtube and put the notification bell on you can watch us on spotify as well it's where i choose to do so and that has been the biggest winners and losers of 2025 thank you guys for watching and we will see you in the next one see you.