Breakeven Brothers

[Theme Music] Cool. Alrighty, we are back, episode 36. Is there anything special about 36, Brad, or is that just a regular number to all of us? It's just regular. Yeah, no crazy update on my end. Cool. How's your week going?

Good. I went to two AI meetups last week. Do tell. Pretty big. I went to what was originally the Claude Code meetup back in October of last year. They renamed it to Agents Anonymous because Peter Steinberger switched to Codex. This was at Sentry's office in SF. So if you're an engineer, you probably use Sentry for bug and error tracking. Not sponsored, but hey, if they're looking for a sponsor, we have a good pod. Anyways, I met with some awesome, talented, in-the-weeds AI people. People that live and breathe AI. Again, I'm very passionate about it, but these people are clearly on the next level. One of the talks that I really enjoyed was from a Cursor engineer describing the long-running agents problem. What he essentially described is that we want an agent to be fully autonomous. We want an agent to do large things. So how can we design a system for a pretty tall task that can be broken up, parallelized, and orchestrated? You can have, you know, 40 million agent hours in a timespan of basically a week.

So, very, very interesting. He had presented about Cursor building a web browser. They had posted an article about this online, so it's public. The engineer, I think it was Michael, talked about the journey of figuring out the right guardrails for this system to be self-contained but also scale. You know, you have one Claude Code agent, one Codex agent, pretty useful. But if you have such a big task, how can you break it down? You say, you know, I have a coordinator, that coordinator is spawning subtasks, and we have sub-agents performing those subtasks. So Michael described the journey through all that, which was very interesting. And I think by the end of it, it was something like 40 million agent hours, which is just an insanely large number. There were maybe 12,000 agents involved.

One theme and one takeaway that I got from that conference this time was that AI agents are obviously here. They're really good at writing code, but they're tackling really large problems too, where I think before, the reliability really broke down over time. Now that the coding models are so good, you can chuck it at a really large problem. But then you get to the issue of how do you review that code? How do you merge that code? Yeah, pretty fun meetup. I got to meet some people that I had met before. It was my second time going there and there were a lot of participants, I think it was like 110 people. But really excellent. If you're in SF, I would highly recommend it.

Cool. Yeah, that sounds like a good time. I think on the Cursor thing, I saw that someone from Cursor published an article on X about solving that long-running agent problem. I didn't read the article, but I saw the guy from Stripe, Patrick Collison, said that it was the best thing he read all year. He gave it really high praise. So I didn't read the article, I just saw him praising it. I was like, oh, that's interesting because I haven't used Cursor in a couple of months. But it seems like they're still doing some cool stuff. They shipped the browser and they've kind of really focused on the long-running agent thing, it feels like.

Yeah, I think they're building for the future. I think that's one of their differentiating factors. Lots of these AI labs will just do model improvements and ask what we can do with them. I think Cursor is building at the right curve of the model enhancements, such that they know something big is coming. And they build these features and UI experiences that make it easier to manage the agents and enable long-running cloud agents. That didn't really make sense a year ago. The models weren't good enough. And now that they've built this and the models have reached that pinnacle, they're in a really good spot. I think their team innovates pretty hard, so shout out to Cursor. But great meetup.

I also went to another one. What's my favorite framework again? Django. No, wrong. Laravel. And Laravel came to SF. They basically come once a year. They had an AI meetup this time. Last time it was just a Laravel meetup, but this time it was a Laravel AI meetup. In this meetup, Taylor Otwell, the creator of Laravel, demoed his new AI package for Laravel. That was fun. I also got to chat with some Laravel people who were out there and met a few new faces. Then Taylor presented for about 45 minutes about the future of using AI within Laravel and kind of making it a first-class package. Whereas before, there were so many packages out there to get things done with AI, but it didn't feel very Laravel-esque. If you use the framework, there are built-in primitives that make things easy. And I think Taylor's goal is to make it as easy as possible, but also allow you to customize it to make it powerful.

So, really fun going there. I met Ian Landsman from Mostly Technical. Talked with Aaron, talked with Ian, talked with Taylor. Cool guys. I think the Laravel community is very chill. If you're in SF and you're around for the Laravel meetups, very chill people. They're extremely hands-on builders. They build useful things, which I think other times, like at the AI meetup, there's a lot of useful stuff, but a lot of stuff is kind of out there. But I think Laravel is the opposite, where maybe they can be a little bit slow to things, but they're slow because they wait until it's proven and then they deliver. So yeah, really exciting from Taylor. I think he's traveling to a Laracon in India and getting to announce the AI package. But it was really fun to talk with folks. And it's nice to get a glimpse into the future of what it'll look like. Since Laravel is kind of my side project gig and I use AI in my apps and websites, it was pretty cool.

Yeah, it's cool. From the stuff that they discussed, did you feel like the Laravel AI package would be immediately helpful to your hobby projects, or was it still too early? It's a good question. I think the current leading open-source package that does AI for PHP, which I've used in my codebases, is pretty solid. But it's not as Laravel-esque, like to fit in the Laravel patterns. To be quite honest, I have worked around that and built it within my app, but I think Taylor's way or Taylor's view on it makes it more of a first-class citizen. And I think oftentimes going the Laravel route, you get other things baked in. So they have a good monitoring product that gives you timing for how long your requests take, how long your database queries take. He had described to me that once this AI package is out, they would integrate it with their monitoring package so I could see the spend per user and the duration of an AI query. You know, I don't think the other package really has that priority, and I imagine this Laravel one would from what he described. So I think in the current mode, I'm happy with what I have, but if I adopted the Laravel package, then I would get some other built-in niceties. So I think it's at a good spot.

Yeah, I saw someone on Laravel Twitter complaining that Laravel wasn't good for AI and didn't have any use cases. This was before your meetup. And I remember I saw Taylor respond and be like, "Well, what do you think is missing? Tell me what's missing." And the guy didn't respond, but it kind of made me think about getting customer feedback and like, you know, I'm not sure if they connected and he gave him feedback or if that guy was just a troll. But I think some of the best founders are ones that get feedback from people and act on it versus ones that have an ego problem and don't listen to customer feedback.

Yeah, I get the sense from dealing with you and hearing what you do in that community that you Laravel people are pretty hardcore, so it must mean that they're doing something right. We gotta fly you out next time, man. Yeah, I'll just be a fly on the wall. I know all these names. I don't know why I know these names, but I guess Brad's talked about them. Yeah, I mean, I think they did a question of how many people use Laravel here or the inverse, how many people don't use Laravel here? And surprisingly, there were a few people. So there are definitely people coming out of their comfort zone saying, "Hey, I want to go to this meetup. I don't know any of these people in the community," which, you know, I've had a history with some of the folks at least and it makes it easier. But being a new person is hard in these meetups. There are definitely a lot of big names, so it can be hard to know who to talk to.

Yeah, good experience to get in that zone if you haven't already. Yeah, for sure. Especially, you know, I think we've said this before, but it's maybe worth repeating now. I feel like in-person will be more and more important just as more and more AI and like, you know, everything's AI slop and bots now on X. So in-person is cool. And it's cool that in SF you have all those experiences.

Yeah, it was literally a Wednesday I had the Agents Anonymous meetup, which is the one at Sentry, and then Thursday was the Laravel one. Either way, it was back-to-back, and so I just felt like I was drained socially after those two meetups. I had normal work and then those meetups, and when I'm at those meetups, I'm like, I got to talk to a few people, say what's up. Pretty fun, but yeah, I think next time, hopefully, if you're listening and you're a creator of an event, just make sure there's a day or two between the events. Even on the in-person side, for the Agents Anonymous one, I think someone was commenting, "Hey, could you do a recording of this?" And the creator said, "Hey, sorry, this is in-person only." So it's not exclusive, but it's different. There's a different vibe. By no means is it a negative comment of you're not welcome here. It's more just, "Hey, if you're around, come hang out. If not, maybe we'll have one in your area." So the return to office collaboration stuff is definitely back.

Alright, well it is 2026, and usually, I guess what we did last year, we did a bingo card for 2025. And I think a couple episodes ago we went back and looked at some of those guesses and some of them panned out and some of them were just hilariously off. So we figured we'd do the same thing for this year, 2026. So we're doing a 4x4 on a bingo card this time around, right? Fewer, but more high-quality guesses or predictions is the idea. So you want to kick us off, Brad?

Yeah, so in the top left corner, I'm leading, so of course it's going to happen. My first guess on the bingo card is a full-length film AI model. So if you've been on any of the AI apps, you've seen that there's a lot more content that feels and looks real. Sora, for example, is from OpenAI. They have a Gemini Veo model, I think is what it's called. It's really good at generating content. And just the other day, I saw a 14-minute video generated by one of these tools, stitched together with ElevenLabs, which does the text-to-speech. It gives you that real voice. They had done all the AI tools, and I think even used Suno to make AI music. So the whole gamut of AI tools, and the video was pretty good. I watched like seven minutes out of 15. It was pretty engaging because I was curious about it, but it also kind of hooked me in. So from that, I got this idea, we're this close where someone can stitch together these 10-second clips. And by no means is it perfect, but in the future, I'm going with a full-length film AI model. Two hours of high-quality video, production grade, something that maybe you would put into your DVD player back when we had those and you think, this is good. Again, you have to write a good script and all that, but it's high-quality video that tracks characters. There are no facial changes or hair color changes, which I did see in this 14-minute video by the way. Something that can retain characters over time with a good script and it will be as if it could be released on Netflix.

Okay, cool. I like it. Yeah. Um, just on a related note, Ben Affleck and Matt Damon were on Joe Rogan's podcast. I didn't listen to the whole podcast, but there was a viral clip of Ben Affleck talking about AI in the context of making movies. It's pretty interesting. So check it out.

Okay, so my first one on my list is humanoid robotics will be in high-income households. And I'm going to say it'll be around 500 units by the end of the year. I don't know if this has already taken place, so if it has, I'll just give myself the win. Wait, why 500? That's so small. This is going to roll out slowly. I thought about this when there was an announcement around a company that released an announcement for a humanoid robot. It was shaped like a human, it did laundry. I don't know if you remember that video or not. Yeah, yeah, definitely. Yeah. So I think they were going to sell for like $30,000 or $40,000. And so that's obviously a high price tag for most people. And I think generally people that are going to be interested in something like that will be kind of more on the tech-forward side of things. So I don't think it'll have that many, but I think they'll start to be rolling out a bit more in households. Because for me, like I think about it from my standpoint. I hate doing the dishes and doing laundry. No, that makes sense. Yeah. I would love something like that, but I don't want something walking around my house. I don't want...

Yeah, yeah. I mean, there's the pros and the cons. But while Ben was mentioning that, I just Googled "humanoid robot" and Google Search has a little sponsored product section. And in the sponsored product section is a carousel with three robots. One that's $21,000, one that's $73,000 as a flagship humanoid robot. And one that's $128,000 that will be delivered by 2/4. So February 4th. I have no idea if this is legit. It's called robostore.com. So I don't imagine anyone's going to this website and checking out for 130 grand. But hey, I think robots are coming. AI models are getting really darn good. We're in the text space, but I think we're moving to the general world model, which a lot of companies are pivoting towards. And I see it happening. 500 feels low, so I'm thinking you're going to get this one.

Yeah, and I'm talking about humanoid. I'm not talking about like the dog robots. Like I think those are interesting too, but I'm talking about like the humanoid where it's like that creepy little dude walking around in your house. Because what I guess is interesting is, in order for those to work right now, they basically record... there's a time period where they just record your house. Then a human in a VR headset is doing the movements. Exactly. Which is even weirder. So, no thanks. But again, some people will want that. So I'm going to go with that one. That's a good one.

Yeah. My next one is an AI operating system. So I think Microsoft has tried to get Copilot, their AI assistant, into Windows and maybe put it in too many places to some backlash. But to me, with the advent of AI, it brings these opportunities to reimagine things. So I don't know exactly what this would look like. But with Claude Code becoming so popular, managing your computer just feels like talking to it. I imagine maybe somehow, some way, Claude or other AI agents can manage your browser, manage your files, do pretty much anything you could do. So it's going to be an AI operating system where you kind of boot up your computer. Maybe there's some beautiful UI that morphs to whatever you need to do. But essentially it's text-in, text-out. You're not using an IDE to write code. You're not using Microsoft Word. You're just talking to an AI agent and it's doing that for you. I don't have too many details, but it feels like there's an opportunity to do something here. I don't know if it'll be fully loved, but I imagine some people will adopt it, given a potential productivity boost. So I'm going to coin it the AI operating system.

Okay, interesting. That actually is going to tie into something I have later, different though. So one will be true, one will not. We'll see which one's which. Cool.

Awesome. My next one is personal AI assistants go mainstream. And so I think, you know, you and I and I think the people that you tend to deal with are very much all about AI right now and are tech-savvy. But the general population is still very much like ChatGPT, like the general population. I'm talking like everybody. And you know, I think Google made an announcement last week or the week before about how they created a universal agent commerce protocol. So basically how your agent can go shopping on your behalf and they partnered with Shopify, they partnered with all these different companies in the e-commerce space. But I think beyond just shopping, like planning, whether you're planning for a trip or planning for certain investment goals, I think you're going to see these assistants become way more mainstream. And I do think part of that will be the broader mainstream adoption of things like Claude Code and Claude Co-work. I think that's going to open up the door to actual AI assistants. Because right now, you and I would agree that you can make AI assistants right now on Claude and, you know, all these open-source tools, like Codex, right? But I think it's not mainstream because it's kind of in the terminal or it's very kind of make-it-yourself. But I think these companies are going to introduce more and more stuff that's going to make it easier for this to be more mainstream. A bit nebulous, you know, I don't know how to measure that one, you know, say "go mainstream," but I'll leave it open-ended. We'll see how that looks at the end of the year.

Yeah, I think that one's a solid one. And it even feels like this might go on the 2025 bingo. So you're hitting two very probable ones. Cool, I'll take it. I realize you're also up next too, so I'll let you hit it. Yeah, check the notes to make sure we have the right number. But I'll go next and we'll just keep it rolling.

I think QuickBooks loses market share. So it wouldn't be an accounting theme for me without talking about QuickBooks. But they've really kind of made some interesting choices as of late, or maybe even just non-choices. And others have been making those choices for them. So like there's been a couple more AI-first competitors like Digits, like Campfire. I think Xero has done some really cool stuff with AI and MCP. They were the first kind of accounting ledgers that I saw use MCP. And they were pretty early on it too. So I think all in all that's going to lead to accountants and users looking elsewhere. Not only has QuickBooks not made those pushes, but the moves they have made haven't been well received. They just recently too, made some UI changes, not small changes, like big changes, and people universally hated it. And they kind of position themselves in a weird battle where they want accountants to shill the product, but then they will email customers saying we can do your taxes for you. You don't need your CPA, that kind of stuff. So they just, they've held on to it for so long. They spend millions and millions of dollars on Super Bowl ads every year. I think they're going to lose some market share. We're going to call it 3% or more by the end of 2026.

Okay. I support that. I think with the AI shakeup, there should be some competition. If something better comes out of that, sounds great. My next one, since you got the double whammy, is a $1 trillion AI IPO. So I'm not a tech expert in terms of IPOs, but that feels like a large number. I think I've seen it touted as there hasn't been a trillion dollar IPO. I'm not sure if that's correct, but I'm going to run with it. And I think given the AI valuations and these companies' revenue and costs both skyrocketing for the foreseeable future, and the confusing landscape of which money flows, I think in this space, we're going to see a knockout $1 trillion IPO. Whether that's for the leading AI model labs or a smaller company up and coming, it feels like 2026, if not 2027. But I'm going to jump the gun and say 2026, quick get-to-market. I even think xAI was talking about IPOing too, which I thought they were owned by Twitter or whatever, but who knows. Long story short, $1 trillion IPO. I think it would be a landmark tech IPO and give people the confidence that AI is here to stay. Since I've seen so many articles talk about both the impressive upsides and the crazy downside: cash-on-fire situations. No idea where that will land. I'm not super in the weeds on that, but I'm thinking one of these is going to happen and I'm going to pitch it for 2026.

Yeah, cool. Yeah, I was just looking at the data and it looks like there's a big gap to get to that $1 trillion. I think the highest one ever was $25.6 billion, if I'm looking at Statista right now. Really? Wow. Yeah. So hey, but you put it out there, so good luck. It's happening. I could see it. I mean, look, they're spending a lot of money on getting nuclear data centers all set up for energy. They're getting all the RAM, they're stealing all the RAM from everybody. So I could see it. It looks crazy, but you know, we're in a crazy world.

Okay, cool. So my next one is advanced biometrics for almost everything. And so I think one, just the fact that AI voice recognition, like text-to-speech is so good now, like it can pretty much impersonate anybody. We've seen those clips of it being President Trump or it being President Biden or Joe Rogan or whoever and it sounds pretty good. Like it sounds pretty good. Like yeah, you might be able to catch it in certain words, but like, you know. And then not only just voice, but then now also video. And I'm sure you've seen the last couple days those videos where it's like a mocap AI influencer. Oh yeah. Yeah, yeah, I've heard about this. It is pretty crazy. Like it's kind of uncanny. Like it's wild. So for those that haven't seen the video, it's like I can be on a webcam, moving my hands, moving my arms, but then what is shown is like it can be a woman or like a different person. It could be like a famous person. You know, they will do all my exact movements and mouth movements and all that stuff. So I think security, you know, and thinking about like interviewing. Like interviewing, that could change interviewing all of a sudden because like you don't even have to really be that person to get hired, you know what I mean? Like you could scam companies, all that kind of stuff. So I think long story long is I think biometrics will be more and more important. I think it'll be Face ID for almost everything, or like fingerprint for almost everything by the end of the year. Like any kind of banking login, you know, right now you have to do a lot of like two-factor authentication. I think it's going to move you to some kind of biometric thing even on your computer as a requirement, not like an option.

I like that. That's one of my favorite ones so far. I think you really pitched a compelling case there. And for folks listening who are looking to make a business, sounds like an area in which you could make a big dent. I know calling into some of the banks will do a voice authentication. So maybe Schwab or Fidelity, I've done this before where they ask me to say a phrase, I say a phrase and then it unlocks. I'm thinking with AI, are they prepared to handle a world in which I have many podcast hours out there and they could take my voice and run with it? I don't know, and you sound different on the phone and there's different characteristics of that. But at the end of the day, it does seem like there's an opportunity there for more security. So I love it.

Cool. Awesome. My next one is a $10,000 a month AI plan. And again, don't have a great definition on this, but we're in the era of Claude Code. Everyone's paying $200 a month for Claude Code. Then Claude Code got shut off for various applications, which was a little bit of drama over the past few weeks about other AI coding tools using Claude Code's subscription plan as a login. So they were basically emulating being Claude Code, but they weren't Claude Code. And then I believe Anthropic went and shut those folks off and said, "Hey, you're paying $200 a month, and that's for Claude." But anyways, what that feels like is that people are getting more money out of that plan than the $200, which is not news to anybody. It seems like they're definitely losing money in that regard. But if they want to become profitable and these new models have crazy high intelligence, I want someone to come out with an extremely high-priced, you know, X dollars per month AI plan that gives you a ton of credits, a very smart model, maybe even fast too. Which we've yet to see with a smart model, but I'm going to pitch this as a $10,000 a month AI plan. Which by most accounts would be extremely expensive. But for the right task, could be worth it. Math, research, finance, quant trading, crazy coding stuff, who knows. But this would be really unique. And to get subscribers on something like this is a really high bar. I want to see people innovate because we're seeing $10, $20, $200, and we kind of stopped there. I think maybe a few months ago, we saw a $2,000 a month plan from a different AI company, but $10,000 is 5x that. What can you get with that? I don't know, but hopefully someone creates it and delivers a really great model and a great experience. So $10,000 a month AI plan.

Yeah, I like it. I think Jason Staats, my accounting influencer fellow, he talked about these expensive plans way back last year, that they could be coming. I think there was rumors that one of the big, you know, AI companies were talking about like a $10,000 a month, maybe even a $20,000 a month for like the PhD level research like you said. So I think it's got legs. I think it wouldn't be surprising, which is crazy to say, but yeah.

Yeah. And it looks like I'm up for the double whammy. So my next one is a GitHub competitor for AI. As I've talked with folks at the Agents Anonymous conference last week, there is a shift from opening your IDE and writing the code versus talking to an AI agent like Claude or Codex. That whole journey of you talking to the AI agent is not very well captured in the final code. It's almost as if you had spent so much time talking to the AI agent to get the code that you wanted. But when you get to commit time and code review time, all of that context is tossed out the window. So to make it more modern and to understand how people are writing code now, it feels like there's an opportunity for a GitHub competitor such that there's a prompt history attached to all changes. There's model-attributed lines, so you can say this line was written by Opus 4.5, this line was written by Sonnet 3.5, whatever. It really gives you a better understanding of the intent of the changes. Whereas sometimes looking at the final code is a little bit harder to understand. Maybe there's a shift to a more modern GitHub competitor to have AI built-in from the get-go. I would love to see it. I think GitHub has long stood as the champion of open source, which is great, and they've done a good job. There's been a few bumps and bruises, but along the way, they've learned, they've adapted, they've shipped a pretty solid product that's used by everybody, which makes it easier. But I'm going to go with a GitHub competitor for AI. And I'm going to add this, which will not be on the bingo card, but I'm going to say Cursor is going to strike with this competitor. So Cursor is building web browsers. How about they just pitch my idea, you know, build a GitHub but AI-focused, have that run for three weeks and go ship that in like six months after they polish it up. So I'm going to go GitHub competitor for AI and a little asterisk, maybe a double bingo bonanza, Cursor ships that product.

I like it. Very specific too. Cool. My next one will be a quick one, but I think it'll be self-explanatory. 2026 midterms coming up. I think UBI will be mentioned. I think it'll be mentioned. I think, and this will tie into my last bingo item, so you're getting a little sneak peek, but I think it'll be on people's minds. And you know, Andrew Yang ran on UBI back in 2020. And you know, he never really had a chance, but... A little early. Yeah, but I think he was early. And I think I put on my Twitter, and it got like five views, but it was like, I wonder if Andrew Yang would have fared better in a more recent cycle or like an upcoming cycle. Because it certainly feels like now, you go, oh, this, there's something that needs to be addressed, you know, like something does. I don't know, but like in 2020, that was kind of still a little woo-woo talk. Like people, especially people that like maybe, I think maybe the most frontier and like ahead of the curve people kind of maybe saw something like that coming. I think Elon had for a long time, even when Andrew Yang was talking about it, had mentioned that UBI was on the cards and would need to be done. But like I think the everyday person was like, no, AI sucks. You know, 2020 ChatGPT wasn't around. And so like, but now I could see it being a thing and I think people will make it part of their pitch to people. And I think the first time we'll see it back in action in earnest will be during the 2026 midterms in certain areas.

I like it. That sounds very timely and I think with the AI advancements, it's a natural conversation. So I like it. Cool. My next one is a widely adopted flagship AI hardware device. So we've seen a lot of these. We've seen a lot as prototypes, marketing campaigns. Not a lot has shipped. But what I'm envisioning is something that is on you every day, is kind of recording, tracking your location, doing data collection, kind of like an Apple Watch. But the biggest thing I think is seeing what you see. So maybe it, I'm thinking like Meta Ray-Bans that you wear all the time that are recording everything. Whereas I think Meta Ray-Bans, for the most part, you have to click a button to record. So this is something that's always recording. You can talk to it. It's recording both audio and video, which is a tall task to have it last, you know, a long day with those kind of peripherals. But to me, the human memory is good but not great. And if we had an AI hardware device that could do better, that would be a step in the right direction. But with that comes all the privacy concerns and legal concerns that people don't want to be recorded. Are you even allowed to record? So to me, I would love to see a device that is able to hit the right marks to have accessibility, privacy, legality built in and be able to consume information that would help improve my life. How they figure that out and how they deliver it, unsure. So I'm going to go 1 million units sold for some widely adopted AI wearable device. A million units sold in 2026? It's a stretch. It's a stretch. Hey, we put it on there. We're making bets. We're making big bets.

That sounds horrible. I mean, I'm not saying your guess, but like that sounds horrible. Like you're supposed to forget things. Like come on. I don't want to remember every single thing that ever happened to me. Like get out of here. It is a, if you've watched Black Mirror, they do have an episode that kind of has that. The story of you, I think, is what it's called. The entire story of you. Yeah, I don't remember the exact title, but it's very similar to this. I mean, again, it's a tough problem. I think it's useful in the right context. That's all I'll say.

Yeah. Nah, then you'll remember things you're supposed to forget. Like, "Oh, that person did this," you know, hold a grudge forever. Yeah, uneasy. But I could see it. A million is aggressive, but I could see it.

Okay, I have the next double whammy, I guess. My first of the double whammies is I think there will be a top 10 AI artist song on Spotify during the year. And I'm going to say this and I'm going to brace myself for any heat we get or it'll be good for the views, Brad. But I mean, like some of those K-pop songs are basically AI songs. I'm just going to say it. Like they're basically AI songs. So like, look, I can't comment, don't come after me, don't come after me. Not that they're not good songs, but like it's just, it's just... You love Golden? Golden's good. Golden's good. But like my daughters listen to Blackpink. Like, "How You Like That?" I don't get it, just personally, I don't get it. So I think it'll be here soon. And I think 2026, there will be a top 10 song that's purely just an AI artist. Maybe we should submit the Breakeven Brothers podcast intro because that is a banger. That is a banger. Yeah. I agree. Maybe that's the one. I'll release the full song out there on our YouTube channel. Yeah, please do. Please do.

Cool. Okay, my next double whammy, and this one is basically the one I mentioned to you where it's kind of similar to your AI OS. But I think that Apple does something big in 2026. And when I wrote this down, I wrote the desktop robot, which I think is something that they've announced that they're working on, or rumored. It's rumored, heavily rumored, I would say. But I do think they recently did that partnership with Google with Gemini where they can basically license Gemini for their own AI. Because Apple Intelligence is an oxymoron at this point. But I think if they hit it right, they could use Gemini with Apple Intelligence. And I just think about the Claude Co-work demo video that they did, where they asked Claude, "Hey, rearrange my files." There's no reason why Apple Intelligence couldn't just do that natively. Like, you know, if it had AI built in that was actually good and functioning, you know. So I think there's a tremendous opportunity, and it kind of relates to your AI OS. I think Apple will have it. I think this is the year they kind of come back. I'm rooting for them. And I think that getting Gemini in their Apple Intelligence will be a big piece of it. And then I think they're going to come out with a big hardware release with some kind of, not a robot, like a humanoid robot, but like a a device, they call it like a desktop robot that's going to just kind of help you. Camera, kind of like Meta had that thing where they could track you around. Oh, I know exactly what you're talking about. The camera. Yeah, we got one as employees at the time. You used to work there, yeah, you should know. Yeah, yeah. They discontinued it. Yeah, I'm going to have to look it up while you're talking next time. Yeah. But so I think, I think I'm kind of hedging my bets with that one, but I think Apple will have a big year in 2026. And I think part of it will be the Apple Intelligence built into their OS where they can do everything that Claude Co-work and all that kind of file organization can do. And I think they're going to have a big hardware release. Portal. I see it. Portal. Yeah. Okay.

Yeah. Yeah. I like it. Apple needs to do something, so I'm hoping this is something they take on just for their own sake. Did you get the did you get the sock? Did you buy the sock yet? No. Oh, you didn't buy the sock? Their $200 sock they released. Apple has socks? No, it's a it's a phone sock. Oh, I yeah, I heard about that. Yeah, that was crazy. It's like a giant long thing? Yeah. Thinking outside the box, I guess. Should have put that on the bingo card last year. I would have never guessed that. Never guessed that.

Uh, cool. My next one is a new AI-specific coding language. So AI is writing a ton of code, people aren't opening IDEs. Is there a way to become AI native on the programming language side? Such that AI is extremely well versed in the language and there's testing built in. There's UI built in so it can look at things, test things, get performance benchmarks from things that are just easy to do in this new programming language. So humans can prompt it, AI will deliver. The code base is extremely AI forward. What that means, I think constantly evolves, so I think it's a bit of a moving target. But people have done crazy stuff with AI. They're building large, complex projects like a web browser autonomously. To me, this feels like it's a must. I just don't know if it'll be widely adopted for someone to just say, "Hey, why not? Let's try this." So yeah, a new AI programming language that's tailored to AI and in various ways just makes the AI agents love it and write great code with it.

Yeah, that'd be interesting. I think that's an interesting one because, you know, at least my understanding of like AI is it's... there's training data. It's all built off of the training data that it has. So I guess in order for that to happen, someone would need to make that language, AI would be trained up on it. Or like would AI come up with its own language, you know, and go about that way? Those are open questions.

Yeah. I'll let someone figure that out, but you're correct. That is how it works. And I'm hoping someone will have the wherewithal to make this a reality.

I like it. Okay, cool. My last one, and this goes back to the UBI one, is I think that we're going to see some small market, like meaning like, you know, where I live, Phoenix. Or some other smaller markets, let's call it Kansas City, let's call it Nashville, let's call it Raleigh Durham, Denver. I'm putting Denver on the map too. Sorry, Denver. I think there'll be some small market recessions/corrections of things because I think there will be some AI-induced brain drain. And I think while that goes on, I think you will have a flight of talent into the major markets: SF, New York. I'd probably stop there. I don't know if you could say Seattle. Yeah, Seattle. I hope they lose against the Rams next weekend. But um, Seattle and like maybe LA. LA's kind of... LA needs to figure out what they're going to be because they're kind of, I think LA's lost, but that's a different conversation. Um, but you'll see like the K-curve. That's a thing in econometrics where, you know, basically the rich get richer, the poor get poorer. I think the smaller markets will be looking like that, and then the larger, more technology-focused markets will be going up that way as long as California doesn't pass that wealth tax. But my point still stands. I'm bullish on SF, bearish on Phoenix. Sorry.

Wow, that's something I would not expect to see from you, bullish on SF. I'm going to frame that one. Just from like a market perspective. I'm not saying like a living place to live. I'm just saying like I think people in Phoenix will feel... People in Phoenix and Nashville and Denver will feel the pinch of AI before people in SF will, I think.

That makes sense. That makes sense. I like it. Cool. Well, my last one is the reintroduction of human typing. And how to put this plainly is every single social media post that's by a popular figure, a high-up at the company, it feels like it makes its way into AI at some point. So talking about something, chuck it into an AI chat, polish it up, clean it up, no spelling mistakes at all. And what I'm thinking here is, you know, I write blog posts once a month on my blog. And as I wrote it, I thought, should I send this to an AI chat to make it better? Oftentimes I do, but then I actually had this correction in my head that, you know, maybe the way I say things, although very far from perfect, with typos, with not perfect grammar, is the way to be human again. So I think AI is so front and center for everybody, it almost feels like a failsafe in terms of writing. Because I'm not an excellent writer, but I think I have clear thoughts and get my point across. But when you have such a smart backup, it feels like, "Oh, I can write 80%, chuck it for the last 20%, it rewrites, and I accept." And I think for these blog posts I wrote recently, it felt like I wanted to keep in typos, wanted to keep in poor grammar, only to signal the fact that I have a brain and I have a preference and I have a style. And you know, whether or not AI would write better than me, I think is a different story. But I'm going to think there's going to be a resurgence in 2026 of people writing things without going to AI for the final draft or for the starting draft. I know people do that as well of give me a content outline and then I fill it in. I'm saying raw human thoughts to get you from start to finish of writing something and showcasing that you have thoughts and it's not just AI. So maybe Twitter will have a new feature of, "Hey, there's two typos here, must be legit." And what I really don't hope happens, which can happen, is AI will add fake typos. I hope it doesn't happen, but hey, maybe that's a step in the right direction. So hopefully more human-written content by people we know and love.

Yeah, I like it. I post on LinkedIn, which I don't know why because more than half of it is AI now. But when I post on LinkedIn, I just I write it all out and I just post it. And I've gone back and looked at my posts and there's like plenty of typos or like just grammar errors, mostly because I'm just writing on my phone. I'm not like sitting down and sipping coffee and writing and typing it all out. I'm just like spitfiring, you know, so. But yeah, I agree. It does, when you see a really well-written post, especially with the emojis and stuff, you just know. You're just like, "I don't give a shit what this says." So yeah, yeah.

Cool. All right, well, let's wrap it up there with the bingo card. 2026 is locked in. I think we have some good guesses here. I think this is actually... Yeah, should we recap? Just give the headlines for every... I'll go first. Just give my my quick recap. You take it, run it through all of them.

Yeah. Okay. So, my first one, full-length AI film model. Second one, AI OS from me. Third one, $1 trillion IPO from an AI company. Fourth one, $10,000 a month AI plan. Fifth one, GitHub competitor for AI, hopefully Cursor. Sixth one, widely adopted flagship AI hardware device, asterisk, 1 million units sold, which to be fair, I think I maybe should lower that, but we'll keep it. Seven, a new AI-specific programming language. And eight, the reintroduction of human typing. Pretty solid. We love it.

And Ben's, first one, humanoid robots. So we talked about 500 units. Feels like a low number, but hey, maybe we're already there, who knows? We'll have to check in with that. Second one, personal AI assistants. Obviously is on the rise, and I think this feels like it must happen or 2026 feels like a bust. QuickBooks losing market share is number three, so a 3% decrease in 2026. I welcome competition, let's go. Four, advanced biometrics. So for anyone listening to the original idea, there's a business opportunity. Please take it up and when you do and you're rich, let us know. Five, 2026 midterms will mention UBI. Very timely, totally makes sense. Six, top AI artist song during the year. We pitched the Breakeven Brothers podcast song, which many of you have reached out and talked about how much you loved it. So maybe we'll release the full version at some point. Seven, the Apple desktop robot, you know, kind of something that sits on your desk and will help you do important things similar to the portal. Would love Apple to make a step in that direction. And eight, the final one, the small market recession due to AI-induced brain drain, kind of shifting people from smaller markets to larger markets. So yeah, overall, pretty solid spread. I think we have a bit of tech, a bit of money, a bit of finance, a bit of accounting. Overall, really solid. I think some are out there, you know, a million units, that's a lot. But some are pretty in there. So we'll be excited to kind of wrap this one up end of this year and see how far we get.

Absolutely. Cool. All right, let's move to bookmarks then. Uh, you want me to go first?

Yeah, let's do it.

Okay. So I found a cool one. In the agent's era, we are talking a lot about skills. So Claude Code has created skills. These are just some markdown files describing how to do something. What I've seen recently is a trend of really smart people or companies distilling their knowledge into a skill and sharing that with others, either for free or paid. One that was out recently was a quote-unquote React Native production experience skill, which is nine plus years of production React Native experience from a company called Callstack, distilled into a skill. And this is publicly available, so you can go to Claude Code, do `/plugin marketplace add`, their marketplace name, and then `/plugin install react-native-best-practices`. What that does is it allows Claude to get professional expertise where it wouldn't necessarily know how to wield React Native. I haven't tried it myself, but I'm sensing the trend that this will be here to stay and companies and individuals see a large opportunity in distilling their experience and expertise into these skills files and sharing that with others. So if you do React Native, as I've done and launched my mobile apps with it, definitely take a look at this. It should help and kind of bridge the gap between Claude being good at React Native and Claude being great at React Native if it works well. So maybe I'll do that soon and kind of report back, but that one should be pretty solid for folks who work on React Native.

Cool. I like it. My bookmark is actually a YouTube video from a YouTube channel. I've watched some of his stuff. I was just turned onto it a couple of days ago and I've been binging it, I guess. His name is Indie Dev Dan. Not sure if you're familiar with him. Does not sound familiar. He does great videos. They're really cool too. He has like his own way of like, I don't know, his own like production style. It's really cool and unique. I hadn't seen anything like it. But the reason I was watching it was because I've been ramping up on Claude Code. I know, crazy, because if you listen to the podcast, I was not a fan of the CLI, but that was for coding. And I've actually been using Claude Code not for coding, it's a bit more like kind of just personal assistant type stuff. Like just kind of information gathering. Got him. Yeah. So you got me. So I've been trying to play catchup basically on knowing how to use skills and all that kind of stuff. And I have a really good start to something, but anyways, his videos are really good because he goes into Claude Code. And the video that I bookmarked is for hooks, what's kind of like what you need to know about hooks essentially, because using it for accounting, what we're kind of envisioning doing is like building out... I think he focuses on the four core. It's like model, context, tools, and gosh, there's one more. So I'm forgetting what the fourth one is. But anyways, it's just kind of how do we bring that core four into using it for accounting. And so like we have certain skills where we can get data from the database. We have certain skills where we can prepare the Excel file. And so you can really put it all together with something like Claude. It doesn't have to be Claude, it can be open code, any of those CLI type interfaces where like it can interact with your files and run bash scripts and stuff like that. I think it can be really powerful for some accounting workflows. So, long story short, hooks are important because we want to be able to log like tool usage and log like events so that we can see like an audit trail and have reproducibility. And so hooks were something that I've like tried to wrap my head around a bunch. And it's like every now and then I come across like some kind of like documentation or code where I'm like, I just don't get it. Like no matter what I try and do, I just don't get it. And like hooks is that for me right now. I'm just like, you could explain it to me three times like I'm five and I'd still be like, "I don't get it." So I need to get through that video and like really absorb it because I think there's a tremendous opportunity there for accountants. So excited for that one.

Sounds like the age of AI, just topic after topic being a slap in the face with something new and trying to figure out which is useful and also to overcome some of the challenges and understanding how these things all work together because I've felt that so many times in the AI space.

Yeah. Well like and then Anthropic is just like merging commands and skills. And not to derail us, but I'm like, well, I just learned what the differences are. Now you're going to merge them together. Like, come on now. I can't keep up. You're too late. I can't keep up. But cool. All right, well, we got the bookmarks done and we'll wrap it there. I guess one quick note for our video, Brad, you want to give us a fill us in on what we should chat about or kind of mention here?

Yeah, so this is our first episode or second episode of 2026. Going to take a medium-sized break and figure things out. So we'll be back in a few weeks. But in the meantime, we'll be around, we'll be learning. So stay tuned for our next update. Should be a lot to cover like always, but yeah, we'll be gone for a small break just to figure things out.

Yeah, I think we normally post every two weeks or so. Give it like three or four, I think. And if we're earlier, then wonderful. But give it three or four before you do any wellness checks on us because we'll be busy.

Yeah. We're learning. We're figuring things out. Cool. Awesome. We'll end it there. Good stuff. See you next time. See you next time. [Theme Music]