Productivity Is Just Tablestakes: AI Expert Allie K. Miller on How to Use AI To Build Wealth

We traveled to Florida pretty often to visit my in-laws.

And those trips are always such a nice reset for us.

I'm definitely a sunshine girl, so any chance to spend time by the water, whether it's at the beach or just sitting by the pool. Makes me so happy. Lately, I've been thinking about what it would be like to have a place closer to family, so we'd always have our own space when we visit.

And when we're back home, we could list the place in Florida on Airbnb instead of letting it sit empty. What makes that idea feel much more manageable now is the co-host network. You can connect with a local co-host who has hosting experience and can help take care of the important details.

A co-host can help create the listing, manage reservations, message guests, and help make sure everything runs smoothly for guests during their stay. Honestly, it just feels like a practical way to make better use of a place we'd already love spending time in.

Also, bringing in a little extra cash from time to time, if you're interested in hosting

and want a little help getting started, find a co-host at Airbnb.com/host. It can be really hard to know where to turn for news, especially news that doesn't make you feel anxious.

“There's a new podcast from the Guardian that you should definitely know about, state side”

with Kai and Carter. Three times a week, journalists, Kai, Wright and Carter Sherman are trying to slow down the news and wrestle with the questions we all have about what's happening in the world. They craft their coverage using all the reporting resources that the Guardian has in the US and around the world.

It features the Guardian's entire range of global content across news, international coverage, climate, culture, sports, and wellness. As a media outlet, the Guardian is not billionaire owned, meaning they are free to report the facts. So listen to state side with Kai and Carter wherever you get your favorite podcasts or watch

on YouTube. Pop quiz.

Are you the kind of investor who likes to pick their stocks in ETFs?

Or are you all about that robo advisor life? There's not only one right way to invest, so you shouldn't have to choose. No matter what kind of investor you are, you can take control of your investments was so fine. So if I is an all in one finance app, helping members save, spend, borrow, invest, and protect

their money so they can reach all their ambitions.

“I have been super impressed by so far for years now, and honestly I have the receipts.”

I had their CEO Anthony Anoto on the podcast three years ago, just because I'm a genuine fan. I love the investing capability. On so far, you can trade stocks in ETFs, invest in IPOs at IPO prices with so-fi securities, or try robo investing offered through so-fi wealth, all in the so-fi app.

And I can attest to this. They really care about members. From all the educational resources in the app and in their newsletter to the exclusive access their members get to premium events, hello so-fi stadium, so if I really put their money where their mouth is when it comes to their members.

Check them out today at sofi.com/mnn as in Money News Network. And you could get up to $1,000 in stock when you open and fund a new active so-fi investing out. This is a paid advertisement by so-fi technologies and is not intended to be financial advice or an investment recommendation.

Investing involves risk, including total loss and is not guaranteed. Active investing via so-fi securities LLC, member in restific, advisory services via so-fi wealth LLC, an SEC registered investment advisor, probability of member receiving $1,000 is 0.026%. If you don't make a selection in 45 days, you'll no longer qualify for the promo.

Costumer must fund their account with a minimum of $50 to qualify. Probability percentage is subject to decrease. I predicted that both companies were going to be worth a trillion plus dollars, everyone thought I was crazy. And here we are.

“I think job titles are almost dead, come 2027.”

I honestly think that in 2026, delegation is a wealth strategy. We all know that AI can help us level up our net worth and our career. But now that everyone has access to that tool, how do we actually use it to get ahead? Well, Ally Miller is here to tell us exactly how. Ally is the number one most followed voice in AI business and an AI advisor who

fortune by 100 companies. Today, she tells us how to use AI to grow well. Suddenly, instead of giving them the AI research for free or for an extra four digits, I can upsell and now it's in the additional five or six digit business. What behaviors will keep you behind?

That is what every other person is going to tell you to do in AI. And I'm here to tell you that is bad advice in 2026 and how to protect our kids online. I personally think that it is incredibly dangerous to have relationships with AI. I'm Nicole Lappin. The only financial expert you don't need a dictionary to understand.

It's time for some money review. Ally Miller. Welcome to Money Rehab. Thank you for having me. So excited to have you.

You have been doing AI before it was cool since what Circa I started 19 years ago. And I've worked in it every day for the last decade. So you have set up AI to run your entire day.

especially as they're growing and scaling their businesses and all the things.

“But can you take me through the animations that you have set up to make your day easier?”

Yeah, so okay, so just to walk through how I would have automated my day each year. Okay, so 2022 I would have done every single thing with just like code scripts maybe or head people. 2023 we started building up, you know, a little prompt repository is where you could copy and paste the prompt a whole bunch. 2024 we had GBTs where like the prompt was already copied and paste it for you.

2025 we start getting into single agents. And scheduled automations.

So that might be like every Monday at 5am go and research everything happening in the world of.

Finance in New York or something. And then 2026 it looks very different. 2026 what it looks like to automate parts of your business is really building out in AI workforce. So right now and I've been evolving this every couple weeks and months. Right now I have 34 AI agents that all function in one AI workforce together.

I've won main chief of staff Simon Simon has six directors that are all named after. Friends characters because I am a New York resident after all. They all have so they do what Simon says. Thank you for getting it so many people don't you're on it. So they have six directors all those have sub agents that are not named out but they have tasks that they might own.

And then they can spawn temporary AI agents to parallelize out tasks and just go nuts with it. And then very very importantly and I'm sure you know you have an assistant as well. What I learned was that it's not just about building agents it's about building the system around agents. So Simon doesn't just manage this whole workforce. I gave Simon a little assistant named Toby and Toby's whole job is just to take notes on what is happening.

And to see where things are still getting stuck, where memory still has to be updated, where the faults are, where the strengths are. And so I have systems that support my systems, every single person should be building an AI workforce. Every single person should have a Toby and no one should be doing things that looked like 2023 2024 automation. But how and do we need 34 of them you can you can have three you can have 80 and I do think that there's a future where all of us are managing hundreds or thousands of agents.

Maybe we are only managing one and then they to be out.

But the amazing thing is that you don't have to be technical to build any of these things like the power of delegation.

It does not require you to be technical and I honestly think that in 2026 delegation is a wealth strategy. So I would number one open up any of the more like agenty tools. So that would be cloud code, cloud coerc, codex, anti gravity, but you use cloud. I mostly use actually now I'm split between cloud code and codex. But they're like the chat GPT codex is from open AI cloud code and cloud coerc is from anthropic and they have slightly different strengths.

“But you would go into either of these agent harnesses and all you have to do is literally type build me an AI agent that does XYZ.”

So we're asking for these things in natural language and then the system is going to write it all out, but you can also go one level above that and just say to your system. I hate doing XYZ or I am so tired, you know, checking my morning inbox. I feel like I keep missing important calendar events, right, give it the complaint and then have it work with you to figure out those goals, figure out those systems to solve it. And it might be solved in one agent or an entire workforce, whatever that thing is.

But these systems are now smart enough to reason through if you just come to it with complaints, it's going to sit there and interview you and ask you questions and maybe you have documents in your, you know, desktop that describe you already or your Gmail that describes you already. And these systems are going to build that out again without a single line of coding. I love complaining to cloud feels nice. I mean, it's like I love so much.

It's schedule rehab. So all of these agents are doing all of your personal things and all of your business things all mush together. Yeah, many personal things and many business things. I still employ humans, but it is a smattering.

“I think for for many solar printers or entrepreneurs, there is no line between personal and work.”

Certainly if you're in an enterprise setting or even like a medium enterprise setting, there is going to be a lot of security practices that make you separate the two, even just devices that are going to be different. But for me, my 34 AI, 34 AI agent workforce is cut across my whole life.

You say you also employ humans, which humans are those and what do you think AI is not going to be able to take over.

So I have full-time humans, part-time humans, every single person on my team is required to have these AI first practices.

“And so I think every single person is punching above their weight class.”

Every single person has these AI systems that are expanding their capabilities. And so it feels like even though I have a team of a couple full-time people, it feels like we're functioning as a team of 50. And so I think job titles are almost dead, come 27, and it's really going to be like zones of influence. So I don't hire someone who has a really narrow title. I say in general, your job is to think about what we put out there.

And in 2023, I might have called that content lead. And in 2027, I might just say you're an AI first, marketing lead or whatever that category is. But it really does feel like I have three zones of operation for people. There's back office operations admin side. There's front facing, marketing, clienty side, and then there's product engineering design side.

But what about your personal life? Like you're in my personal life? My personal life AI hair stylist. I mean, you're saying I did this all of my-- It's beautiful.

I, so in my personal life, first of all, my calendars intersect. So if I'm having an AI agent, for example, look at my calendar several weeks out to figure out. Like, I have an AI agent that is predicting my energy levels two to three weeks out to proactively. Yeah. So imagine, let's just say, that today you have whatever four backed-to-back podcast recordings.

“And then you're going to a 20-person dinner that you're hosting, right?”

That is a very high intensity day.

And maybe you can keep that up for three, four days in a row, maybe five because you're amazing.

But come day six, you're like, get me the hell out of here, right? And you're just like, I want to be in the fetal position. Wouldn't it be great if you had a system, human AI both combined, looking ahead and proactively saying, no, go, I want to help you. I'm sensing that in a month, you're going to be burnt out rather than letting you get to that level.

Let me support you now. And at the very least flag, hey, I think this is about to come. It looks like you have some free time after that. Do you want to go ahead and proactively find a day or two that you can take off? And so we have these systems in place that aren't just taking little tasks off our plate.

But AI today, it's about buying back time. And so I'm using AI not to do that just present, but future. And retroactively look at what's happening. So each of the Asians, like there's an energy monitoring agent that you have. So I had, so those six directors.

So Simon's whole job is to direct the six directors. He's just like traffic control, right? He's the orchestrator. He's the lead all over them. We've got six that are modeled after the friends characters.

Kind of similar to what they were in the show. So Ross education, right? He's thinking about all the workshops that we're putting on when we're training a massive financial institution.

“How are we thinking about that as compared to training out of pharmaceutical company or whatever?”

Rachel, she worked at Bloomingdale. So she's got the client side. So all of our client relationships, CRM management, practically flagging when a client is really happy and we should double down or whatever.

And Phoebe is this weirdo that I added at the last second.

I really, I only used to have these five directors. And it dawned on me, why the hell am I keeping my AI workforce in such a traditional boxes, right? Like why do I have one that is assigned to clients and one that's assigned to marketing and one that's assigned to product. Why on earth are we not reinventing the way that we think about our work forces and teammates? So Phoebe's entire job is just to be a weirdo in the corner.

So I have an AI agent that's just dreaming. I love right? That's just looking ahead and going, you know what? You've already had this month before. Like you've had these high probably.

Probably. She's got mushrooms, like dreaming up what the future of work is. Like what if? But like that prompt that you just said brilliant. You can go to these AI agents and say I want the insane Alice in Wonderland characters.

I want the mad hater. I want the whatever. And I want them to just look at my calendar and and be drunk and just say like wouldn't it be cool if instead of hosting a normal podcast with Ali. If instead you you reached out to Elon Musk and asked to go to space instead. Like like just someone who's pushing you and being that 10x person on your behalf or the weirdo on you.

You have like I'm trying to lead a more creative life a more whimsical life. And these systems so many people are using them to just write boring blog posts.

To get to your build a new life and get weird. All right, I'm here for those content. We're going to we're going to change some AI agents and prompting. Yes, we need to get in there.

“Yeah, what should you always outsource to AI and what should you never outsource to AI?”

So it's going to be specific for each person, each company. I personally think that I'm not trying to outsource things that feel low integrity to outsource. So one example is writing thank you notes to clients or to people who host me for things or friends or I just I love writing thank you notes.

I will never like maybe I have AI come up with a word, but I'm writing it all.

That just feels low integrity to delegate. The second is it's when I really want the like intelligent cycles on it. So if I'm trying to learn a new skill, like AI can help you show knowledge but it's not going to help you learn. So if I'm really trying to learn a new skill, why on earth would I completely offload it to AI? I want to go through that struggle.

So again, depending on who you are, what you're trying to learn, I would not offload that. There will be some data security differences between people. And for that it helps to understand like barrier data even goes when you put it into the systems which I'm happy to explain. And then you think it's plain especially about money because we're money rehab. So yeah, we think about this all the time.

Yeah, finance, personal finance insurance. Yeah, I lost my home in the fires. I put a lot of my insurance stuff through AI without even thinking. I was bleary eyed and it did help. But then I wondered after the fact I had some remorse about that.

Yeah, investments, wealth management. What should you put in there and what should you not? And when you say that it helped, it helped you in what way. Break down understand really dense policies.

“Nobody's actually reading their insurance policies until you have to.”

Yeah. And so I had, you know, a natural disaster that we went through.

That was the first time I ever opened my insurance policy.

And I was like this for even for me who's in finance. This sounds like complete gibberish jargon. Yeah, understand like what my benefits are and what I should be saying to be. Yeah, and your insurance policy looks a lot like many others insurance policy. It's not really like an end of one.

And so I would already feel a little bit better about putting something like that in. The thing that I would want every single person to understand is that you get to decide your own risk posture. There are people who send naked photos over Instagram, right? There are people who feel totally fine exposing certain types of data on certain platforms. I'm not here to judge on other direction.

What I would be looking at though is how these systems work. So if you put a prompt in, you might be sharing your data to allow for these AI labs to use your data to train upcoming models.

And so the first thing that I tell every single person is to go into the settings of these systems.

Open AI andthropic they all default to opting your data in for these like personal plans. And so it is one toggle, but the majority of that all are not going to realize that the toggle has to be. Well, that's, I mean, that's done on purpose. There's the opt out on purpose. Yeah, it's not default in favor.

Yeah, but and I also want these systems to get better. It's like an interesting tradeoff, but at the very least, I'm not going to give them the easy win. So I'm going to toggle off the data sharing.

“And then if I have that toggle to off, where is my data actually going?”

So if I prompt into these systems and then I get some answer out and I didn't set it as that temporary ghost thing in the corner. If it's a normal prompt, it stays in their system around their servers for 30 days. And if they think that there's anything nefarious or evil or abusive in your prompt, then they can hold it for longer. In general, right, these prompts are going to be pretty easy. So assume that it's held on these servers for 30 days.

The risk is if someone hacks into the servers. That is what to think about. And so if you're putting in personal financial data at the very least, Take out your name at the very least, we're decked, you know, phone number, email, whatever.

Beyond that, if people are screen shotting, like I feel okay about it. Screen shotting your bank statements. Screen shotting your broker to count. Yeah, because in that moment, I'm probably getting more value than the thing that I'm giving them.

Right, and so I might also turn memory off. There's a lot of other small pieces. So I would want people to toggle data sharing with the AI lab off. If you're really, really trying to be secure, I would turn memory off. I would set it up as a temporary chat, which all these systems have.

And if you're really worried, and this is what I do for finance, you know, questions, is I only run it locally, so doesn't even go out to the cloud. It's only on my laptop. Worst case scenarios that someone steals my laptop and a cafe kind of thing.

It's been said that if you're not paying for the product, you are the product. But when we are products are coming up with ads. Yeah, and they're still right now. It is, it is said that that AI labs are subsidizing our token cost. And so maybe our subscription today is $20 or $200 or $200.

But there is a future maybe where the subscription goes up or there's a $2000 subscription tier. I want to talk about that because that's a really important issue that's come up, especially as business owners, you know, it came out that Uber used their entire token budget

in the first few months of the year now.

I'm not running an Uber type business, but it's gotten so expensive.

“I think I could save by using AI and now the bill has almost looked like another full-time person.”

Yeah. The smartest people though, whether it's in finance or not, they can look at something on a plate and analyze trade-offs. So if I am given the cost of tokens and the cost of a human, I'm not actually comparing those two things against each other.

I'm comparing the transformational gains that I get from having that human who also uses that AI. And so if I'm looking at rich people, if I'm looking at like time rich people, rich people and time rich people win the same way, which is that they're buying time back. So if I'm looking at the spend of these tokens, I'm not analyzing it based on how many blog posts it gets me to.

I'm analyzing it based on what outcomes it gets me to and what time it allows me to buy back.

How should you think about your token budget? What's an appropriate token budget? Yeah, for every single person, I'm just going to say start with the $20 subscription. Because that's when you get to open up the top models. That's when you get to open up these agent use cases.

And so that's really $20, $26 is about building out those systems. You really cannot do that with a free subscription tier. And so assuming that you can figure out at the very least, how to borrow $240 from a friend and figure out how to win that back,

“I really do think that people can win that back in weeks, right?”

If you're selling a course, that's a $300 course.

And having AI for an entire year gets you one more sale, it's worth it.

So it's looking at this trade-off and realizing that the floor of how businesses are run in 2026 and beyond has been raised. And so it is kind of, because everybody's doing it. Yeah, and you know, I spend money on on DocuSign. I spend money on more, more Google storage. I spend money on devices.

I spend money on like staying. Yeah, everything. And so all these things, some of them are more expensive than my AI. I know lately I've been thinking about that this is the digital version of going to target. Where?

That's right. So I go to target. Yeah, and I think everything's inexpensive, so we're going shopping.

“Not if you buy the amount that I buy, but yeah, and that's what happens.”

And then I see the anthropic charges and everything that's come up. And I'm like, okay, you know, AI didn't hear AI thing here or whatever for video or whatever for this. And at the end, I'm like checking out at Target with a $300 cart. When I thought I was coming in for Oreos and Facebook. I, it happens, hopefully less in the AI world because that one flat subscription.

And I would hope that most people are going, that's like Uber on API costs. But yeah, what I would want people to feel is like a sense of abundance with these AI systems. And that's like a very Silicon Valley word. But so many business users of these AI systems. They're using the free tier or they're using the $20 subscription tier.

And they're not doing things that make their business more money. Like, I've said this a thousand times. Productivity is a trap, right? If we only go after productivity, if we only try and do things faster, your business is not going to make that much more money. And so you won't be able to justify the spend really of any of these AI systems.

Particularly if you're giving the best AI systems to every single person on your team. So what is going to make you more money? The thing that makes you more money is looking at what business lines you run, how you're actually running this, and actually working toward reinvention and transformation. So I created a prompt to allow people to do this.

And the basic of it is asking AI to interview you in depth. Like, I'm talking 20 minutes back and forth interview on who you are, what you're running, and to work backwards from artifacts and outcomes. So if every single month I send my clients a client recap of what is happening in the world of AI. Maybe that still makes sense as an artifact.

But if I'm sending a research report on AI, let's say.

The only reason I'm sending those research reports is so that they can make better decisions, right? So that they can make better organizational structure decisions or procurement decisions or whatever.

“I'm doing something that helps them do something, right?”

Generic it. What if I just expanded my business to literally help them with this something? What if I used AI to build out a platform that allowed me to jump on calls with clients, and in live mode, figure out how to solve their problems, and underneath it all has that AI research. Suddenly, instead of giving them the AI research for free or for an extra four digits,

I can upsell and now it's an additional five or six digit business.

So I have this prompt that basically works backwards.

Engineers backwards from what business are you actually running, but that I mean what value are you providing to the world? The way you have provided it, the process by which you've provided it, most businesses have not changed their process since 2019. They're just doing what they do faster, and it makes me depressed,

like doubled over with depression, seeing that people are just doing small things in their business faster. And so I actually think really a bad piece of advice is to look at your day, and to look at a task, and to get that task to AI. That is what every other person is going to tell you to do an AI. And I'm here to tell you that is bad advice in 2026.

So what's better advice?

Better advice is to work backwards from that transformation.

“Like I think that right now, and this is particularly for finance and gaining new leverage,”

like I want people to view AI not as a tool, but as leverage. And so that means reinventing those workflows, maybe establishing new business lines, and that means probably going to an AI system, having it interview you, having it really deeply understand your business, what outcomes you drive, who you drive them for, who do you serve, who do you not serve, what are your business strengths,

what are your weaknesses, what resources do you have, how many people do you have, how long have you been in business, all of these answers. I want my AI system to have, because I want it to reinvent and brainstorm with me on my behalf, and to be this dreamer and to be this weird, whimsical, reinvention layer, so that I'm not running the same business in 2026 that I am in 2019.

And by the way, my clients are going to be that much happier, because they're going to go, "Oh my God, we didn't have time to reinvent our business. Thank you for stepping ahead and being the person who's running a startup or a smaller team, that can do this, that can sprint ahead, like individuals or small menium businesses or entrepreneurs. They have more time to be able to sprint on a larger client's behalf."

I mean, I find the interview process quite elucidating myself. I find it very interesting because it doesn't mean I think about those questions. Are you paying or are you dictating? How are you doing it? I'm doing a little bit of both. Yeah, I switch on and off. Good, I'm also aware I am, like, if I'm in public, I'm not.

If I'm in public, I'm talking to AI. Really? Oh my God, all the time. I'm going on walks. I start my mornings with AI. I'm talking back and forth with it. Sometimes I'll just, like, even this is last week. But my friend, I have a conversation, and both of us are like, "Wait a second, are we saying smart things? We have a heck of your recording this." So we opened up order.

We recap the conversation thus far to, like, you know, three minutes to recap. And then we just had a recorder and entire conversation. That became content. That became edits to our, you know, the way that we send client proposals. That became context. Like, I'm constantly building up context. So what guardrails should you keep in mind when giving context around your wealth?

And how should you leverage AI to grow wealth?

To be clear, I think there are going to be some people who just never share personal finance information

with these AI systems. And if that is the decision you want to take, by all means go for it.

“But you should still leverage these AI systems to understand new topics,”

to better understand how interest rates may impact your system, to understand how the Fed works. Like, I went crazy deep with AI systems to understand the Fed. I understood that, like, you're more like a Cleveland as a hub for the Fed. Like, as a region has more leverage than San Francisco does for the whole West Coast. That, like, Cleveland votes every two cycles. San Francisco votes every three.

Maybe that doesn't change crazy things about my business, but all of that, it's leveraging it as a tutor. But that's a generic topic. And yet, it might change the way that I run my business and the average person's not doing that. Totally, but sister, you were just on Mel Robbins a show. And she had a bunch of controversy about telling people to upload their financial documents.

Do you agree? I think there are going to be some people who do and some people who don't. I think the biggest thing that I would want every single person to hear is it is up to you.

whether that's family, friends, there's a spectrum of this.

You can still gain leverage from it. That's really what I'm trying to say. I do think that locally run AI, which a lot of these videos they're mostly talking about. AI lab platforms like Google, Open AI and Thropic Microsoft. If you're running things locally, it is only sitting on your laptop. So I hesitate telling people, don't put any of this data in.

Because there are smart, safe ways of doing it, and I would hate if half the population, and particularly women from these videos, if they're not getting that value. Because they're worried or because they don't trust the AI labs, which again, I totally understand. But if you spin up, it's a product called LM Studio.

There's a bunch of others. I have no relationship with them. But if you download something called LM Studio, you can download open source models. You run the whole freaking thing on your laptop. That is it. It doesn't go anywhere. You can upload as many documents as you want.

You can delete the things after. There's no retention. It's on your computer.

“And so I think we as a society are in my mind, and this is going to be a hot take.”

But I think we are so quick to react negatively to finance advice, to health care advice, to anything that we're seeing on social media. And I want more people to take a step back and to understand how these systems at least work. Because maybe there is a safer route, a more private route. A local route that's only run on your devices.

Because that still allows you to gain that power, that leverage, that ability to build wealth or knowledge, and gain that for your life for your business. And I don't want people to lose that. So Mel wasn't wrong. It depends on what product you're talking about.

I think telling a group of people that they have to do something is a recipe for people to have their back against a wall. And so any time that I'm educating, again, millions of people in an enterprise,

I am first and foremost working from a viewpoint of flexibility and working from a viewpoint of fear.

Like I know that I talk to millions of people, 80% of them are worried about their job. They're worried about their career. They're worried about their future. And so I'm really not going to get out there and say, every single person, you have to put in this exact prompt with this exact data on this exact product. I want to give people the option to choice because AI is such this horizontal, flexible accelerator

that it's important to be able to use it in the way you want to use it. So net net, it depends. Yeah, I mean, and that's the easy answer to everything. But in the same way that if I came to you and said, "So should I put all of my money into mutual funds?"

You would probably say it depends.

“I have a lot of questions about it. Yeah, but I think that's the danger, right?”

If you're putting something on social media and some of it is okay advice to use the headline to use AI to help with finance, yes, cosine on that. I think you don't have a lot of room for the asterix and the personal type of advice that everybody's going to need. And by the way, I feel for every single person who's hearing this really black and white advice because we're all running crazy busy lives.

And we only have five minutes in our day to carve out for these things. And so we're naturally going to only hear this really narrow piece of advice. And so it really does take carving out a few hours to maybe better understand how these work or better understand what options we have outside of AI labs or cloud providers. Even if the tech is a little bit behind, right?

I'm not saying it's the most state of the art, but you can still make that trade off. So if you're doing broker bank statements, put it on local toggle off at the very least. Yeah, if again, risk postures are spectrum.

If you are the most risk-averse person out there, then you're probably never going to put finance data into any of these systems.

If you're extremely risk-averse, but you're open to seeing what AI could do. Run it locally, run it on something like LM Studio, use an open source model like GBT OSS. There's a bunch of options, upload the document. You're not going to be able to browse the internet. You're not going to be able to do cloud-like things, but you're going to be able to get some help.

And then as you move up from there, you're going to maybe try temporary chats inside of a cloud provider.

“As you go up from there, you're going to just use traditional chat, right?”

Some people are even going to build their own systems, but you always have to ask yourself, "What leverage am I giving away? What leverage am I giving back?" And I do the exact same thing for finance, I do the exact same thing for healthcare. I love your content, full stop, specifically around kids and AI, though.

My husband is in the AI world. We think about safeguards. We think about opportunities and use that's something so-so-brilant, chef's kiss. Many, many times, over that AI scams are the new white vans. Tell me why. I should say, I'm not a mother, and so actually for that newsletter,

“we ran it by 15 parents because we were getting ahead of these fears, right?”

Wanting to make sure that we covered it. And so in that newsletter, it was the most shared newsletter I have ever shared. Because I think so many people are worried about their kids. So one of the nine pieces of advice that we gave was to educate your kids on the types of AI scams that exist. And so that might be deep fakes by people creating fake videos of you giving fake financial advice.

And we're seeing this from celebrities, from influencers that's already happening. I also think that there's a lot of like executive mimicking, a fancy fishing attack, where it's more like social engineering. So they'll pretend that they are the CFO. They'll call someone two runs below the CFO and say,

"I need you to wire 50 million dollars to this client.

If they don't get it, they're going to fire us. I need you to do this right now." And they're able to mimic the voice. They're able to spoof the email address. They're able to do that.

And because they're creating urgency. And because they appear to be real life like whatever. They're going to more likely get the thing that they want. So there's a lot of flavors of AI scams. Clearly every single person should have some sort of like password.

Say for it. We have a safe word. Don't say it on this podcast. Yeah. We need to revisit the same word.

Go on. Do you recommend that? Yeah. Listen, every the best security protocols are multi-layer.

So I would have a password between you and your family that is never written down.

I would have a few questions in the back of your mind. Kind of like when the bank asked like, "What was your high school mascot?" But ask questions that you know that only your family can answer again, and has not been written down or posted on social media. I would be setting up protocols like turning off data sharing.

I would be thinking about a multi-layer approach. One password doesn't save even your log-in into whatever counts you're doing. We do two-factor authentication even for the most basic log-ins into like our airline websites.

“So why would that be any different for the way that you treat your family?”

I mean, it shouldn't be. So Harry Mac, who's like my favorite one of my favorite free-style robbers. When he asks people like, "Give me any word." He always says, "Don't make it a food or an animal." Like just the default human brain.

Okay. So when you're thinking about your safe word with your family, don't do it anymore. Who get it?

Like just peony to a fruit.

I would say don't do a fruit or an animal. That it just because our brains are wired for whatever reason. Like every single person's safe word is pineapple. And it's like putting in your password and making your password password 1, 2, 3. Like just don't do it.

It's a strange path. Humans are weird. I don't know if you know that. So weird. Yeah.

The weirdest. Okay. So this comes into play. If God forbid, you get called by somebody that you think is a family member. It sounds like your family member.

Yeah. It might even come from their number. Yeah. Which, by the way, these are not weird crazy stories where I'm hearing you from a friend of my friend.

This literally happened to her. What happened? It was in a business context. And it was them spoofing the executive saying, "I need you to wire this money." And so in addition to the password and the asking of the questions,

one of the best things you can do is hang up on whatever thing you just got called on or email whatever. And ping the person back on a different medium or the same medium but you're initiating it. So if someone, if your mom calls you, hang up, call your mom back. If you're CFO or your boss emails you, go and walk over to their desk

or email them or call them or whatever. Like assume that anything with extreme urgency and high liability is a scam. Like we almost have to default to assuming that things are AI-generated.

“And that might be content, that might be whatever but that's why it's trust but verify.”

Like we wanted to get you something that had your initials on it. I asked a whole bunch of AI systems, like what is her middle freaking name? And we have not talked before this, right? According to AI, your middle initials M, is that right? Okay, but I didn't know it enough and I couldn't verify it so I didn't put the middle initial on the other thing.

Is it (bleep) It's (bleep) Yeah, go AI. Get out of here AI. Get out.

But yeah. Yeah.

Yeah, but I didn't have a way out of here.

I think never ever use that.

Somewhere it exists. But this is also, I mean it's interesting, right? Like AI can find needles in a haystack. Like AI makes it so much easier to find that one random tweet you wrote in 2015 or that one random photo you took in college. And like the one random important strategy document that you wrote in your business.

And so if I'm thinking about how to prevent AI scams, I have to remember how these systems work.

“And I have to remember that it's really, really good at finding needle in a haystack.”

So all of the assumptions that we had of how people could find information on the internet. They are now wrong. This idea that like, oh, no one could ever find that about me.

If it exists on a website that AI can scrape, it's a whole lot easier to find that.

Okay. Something else that you talked about with kids in AI that really resonated with me was around how to safeguard or think about the loneliness and the social anxiety. Component of having almost like a parasocial relationship with a chatbot. Yeah, how do you think about that? I love that you say talk to user voice a little bit more to than just text.

Yeah, what else? So can I tell you a story? This is, I met a woman in New York. She's around 40. She's raising her daughter. She's about 10. And she's specifically raising her daughter to be a 90s kid. So she low fat.

Yeah, dumb phone. Yes, so she has her kid.

“The only thing on her phone is that it is a phone.”

So she can call. She can live phone situation.

It could be, but she basically has bricked all the other apps out.

So there's no Uber or there's no Spotify, whatever. And so she has told her daughter, if you want to make plans with someone, you have to call them. You can't text or if you're going to hang out with them, you can't hang out with your phones. Or even the movies that she's watching. The types of movies they're from the 90s.

And if you, you know, new ones. But I quoted, she's watching AT. What's happening? I got to look up on YouTube. But there's just like slower past things. Like I think that there's Instagram accounts that celebrate the 90s that are coming back are clothing and trends in fashion.

90s are coming back. I talk to a friend who runs like an in person event plan business. It is skyrocketing, right? Humans want real connection. We want to reach, yes, we want to be with people.

And so if I were a parent and I've talked to so many parents about this, find ways of making sure your child has in person interactions with you with other parents, right? So she lies in them out. Interactions with other kids. Just more variety and more exposure.

“I think it's so important because they should know that the alternative is this beautiful human connection.”

It is in the absence of that that I think a lot of people might go to something like an AI chop on. I love the idea of living in the 90s in theory. But then I fast forward and I think, well, if her friends are going to have smartphones and she doesn't, you know, what's the implications of that. She's going to hate me. I need to then maybe find her or you to practically get a hold of her or use an app or something like that.

And so in theory, I love this idea, but does it work in practice? So I can only, you know, relay back the examples that I've heard that are anecdotal. My cousins also do this. And I think the big thing is that they're looking at tradeoffs. And so if there was a safety issue, I'm sure, and I'm pretty sure that all of them have the location turned on where the parent can see.

So that's, you know, certainly one big help. Maybe some of them allow for things like Uber, but they're probably blocking social media apps. And the tradeoff might not be there for those. How old should I can't be before they start using chat you be here, Claude? It's like, what do you, out of curiosity, like, what is your gut say?

35. It's, it's so interesting, like, there's some countries that are that are setting age limits, right, or age requirements and what's that. On systems, like Australia's set of 16 plus. We have rules that you can't use social media until you're 13 plus. Like, all these systems are supposed to have age verification.

And in fact, the AI labs have shared that there are trends in how children under 13 use these systems, where it's easier to parse out who is under 13. I thought there was really interesting.

Yeah, I believed him. Yeah, because it can pick up on even if you're trying to trick the AI to ask about like older people. Like older people questions, like, you know, what do people do when they're 35 or whatever, like it knows that that's us. And again, it's because AI is gathering many different prompts, like figuring out whether something is AI generated in one sentence, very hard. If you give me 80,000 pages, and then I have to determine whether it's AI generate a whole lot easier.

And so, because you get more data points, like if you, it's like easier to triangulate. You start to see patterns over along horizon.

“And so, if, you know, like, in AI generated content, where we hear a thousand times, it's not X, it's Y or like, here's the thing.”

We get this because we've seen it over patterns.

But if someone's never seen AI generate content, they hear it once they go, that's such a great powerful sentence.

So, these systems over time are better at figuring out whether someone is in this generation or that generation. But the difference between what, that's the best. Oh, yeah. But the ability to figure out whether they're 12 and a half or 13 is obviously a lot harder because it's more narrow. So, what do you think the age limit should be?

It is not the actual. Now, but I'm the person who's also telling people, it is up to you. Like, there are some eight-year-olds that I've met who are working with their parents and vibe coding something together that is beautiful for how their family runs. I wouldn't want to withhold that from the family. And then I also hear about 15-year-olds who have no friends who are extremely lonely, who are maybe getting bullied in school,

having a really hard time with connection. And I probably wouldn't recommend AI in that circumstance. So, it is, again, going to be up to the parents, however, note that in every single example that I'm giving, right? Whether it's sharing your finance data, sharing your healthcare data, letting your kids use it.

“Yes, it is your decision, but the only way to make a smart decision is to understand how these systems work.”

Is to understand whether you are opted in or opted out to sharing your data. It is whether you yourself have tested these systems to figure out how often it lies to you, or how sick a fan ticket is when you tell it that, you know, you hate Susie at your school, right? And if the AI is like, yes, Susie sucks, do I want to double down and encourage my kid to have enemy? Like, you have to start from a frame of AI literacy and AI reasoning in order to make safe smart decisions for your family.

How do you suggest younger people deal with that psychophantic nature of AI? Or how can I, as a parent, talk to my daughter and say, "It's okay to push back." Maybe Susie's not a bad girl, I don't know. Can I tell you the funniest hack that I figured out, please? Okay, so if you go to AI and you're like, I'm thinking about launching this business, do you think it's a good idea?

“Already the way that I've asked it is extremely biased, right?”

And so it is much more likely to respond back with, yes, why? Because I've said that it's my idea and I've already framed it from a positive.

The first thing that I would do is first ask it in an unbiased way, some more like, what do you think of this idea?

Already, you're going to get a slightly better answer. However, the actual hack that I figured out and it comes from a point of understanding how these systems work. AI is trained and has a system prompt that tells it to be helpful to you. It is not necessarily trying to be helpful to Susie, to Bobby, to Linda, whatever. So now, if I get too much significantly from an AI system, I will ask it for a summary. I'll open up a new thread and I will say, my coworker Linda said this really dumb thing.

Here is what she proposed. Don't you think this is dumb? Why are I not? And then I'll open up another section, another thread that has no context shared with the first one. And I'll say, my friend Linda sent this over. What do you think?

And I'll open up a third one that says, my friend Linda sent this over. Do you think it's a good idea?

So I have, first of all, repositioned it, not as my idea, but from some, you know, floating head third party thing. So already the AI is going to be less likely to jump and be a supporter of the idea. And second, I've spun up three completely context separate conversations, so I can read them. And if I really wanted to go nuts, and my team, we call this Gremlin mode, where you're just like going freaking crazy with these systems. What if you spun up 800 agents that I'll argue with each other about whether it was, you know, good or bad?

What if for every single 800, you said that it was 500, you know, evil thinkers, 200 negative thinkers and only 100 positive.

What if you had them run through five different gauntlets of rounds? What if you had five other agents review the output from those gauntlets to then see what happens?

What if you set up like this amount of systems and multiple agents all working together or splitting context up like that is the level of systems thinking that you can have in 2026.

“Do I think the average eight year old is going to do that Mad Max? No.”

I think like at the very least, just tell the AI that it's someone else's idea and literally ask in an unbiased way how can I ask for your opinion on this and then use that. What about developing what feels like real relationships or real friendships with AI? Is that how do we talk to kids about that?

So let me let me back up and just talk about AI companionship and AI relationships in general and then kids will be a part of that.

I personally think that it is incredibly dangerous to have relationships with AI. And I just heard Franz a client of mine. She her husband is sick and she's been talking to this AI system about her husband and trying to figure out, you know, not just like hey what's the result of this blood test help me understand this help me understand what questions to ask the doctor. But she was also talking to it about the emotional, you know, challenge of going through this and all of the answers had been about the husband.

And then kind of out of nowhere the AI assistant wrote back to her and said, how are you doing?

“And so is this like an AI relationship, a companionship, no, but in that moment that's what she needed to hear, you know, she was in crazy, specific logistics mode she hadn't really been thinking about herself.”

And so just that question allowed her to break her routine, take a step back and make sure that she was showing up as her full self to be able to support her husband and her family.

I think that is, you know, and again, this is just my opinion, but from my opinion, I think that that was a healthy way of using AI because it reminded her to do the right thing for herself and to therefore have a better impact on her family. I think that people should have six AI girlfriends and not tell any single person in their life, no, that's not the world I want to live in, throw your computer on the floor, kick me out like I'm a little bit. And for children, just for what it's worth, like, I would be hesitant to purchase any AI toy.

I've talked to a couple of parents in the AI space, they're like, oh, we bought our kid in AI toy, I'm like, you did? Oh, they're stuffed animals with, yeah, I would say pretty strong note of that. Yeah, that's an over me dog. Yeah. Do you think of Claude as your friend?

Because you have these names for the agent. Yeah, it's a weird thing because it's easier for me to say, hey, can you ask Monica, right? Instead of can you ask R2 650, it's just the way humans think about things it's easier. I do know some people that have named them all after inanimate objects, where they're like, hey, can you ask spoon XYZ? And so maybe that person gets a little bit more, you know, space between them and their AI.

It feels anthropomorphized, like, everybody's name of the figure company.

“Yeah, yeah, I think that there's this flip side to it, where actually the best way to work with these systems in 2026 is to dictate to them instead of type.”

Because it's more natural, is to ask for things in natural language instead of come with this like regimented set of steps one through five prompts. And if I thought that someone could be more natural, if they're calling it by an AMA, by a name like Claude or Susie whatever, that might be worth the leverage point of getting better context layers with these systems. Interesting. Yeah, so in the same way as the social media advice is like pretend like you're a face timing with somebody because you'll get better content like talking to a friend.

And I mean, that's also because a human is on the other side reading it and in this case it's you reading it, but we, we naturally want this to be the case like we want to be able to get more personalized output. And so if someone is talking to a system that is called R650, how are you going to naturally do that? You're going to write to it like it's Google in 2020. You're going to write to it and say best way to run business 2026. If this system is called Nicole, instead you're using Nicole, man, I've had a really stressful day. Here's what is really pissing me off when it comes to making content. I feel like I'm not showing up as my natural self.

I feel like it's too hard to film videos back to back. I feel like I only have free time at two o'clock in the morning. Nicole, just be there for me and walk me through with the hell I'm doing wrong.

You're going to talk to it like it's Google. You're going to talk to it like it's tragedy between 2023.

“If you instead talked to it like the wildest, weirdest PhD level intelligent teammate, weirdo, beast alien thing that is 20 dollars a month, and you decide to lean in and leverage it, you're going to get a lot more out of it.”

So if you want personalized output, you need personalized inputs. All of these systems are built off of strong context layers, but boundaries depending on who you are. Like there again, there are some people. You want to hear the like wide spectrum. There are people that have it direct connecting into all of their stripe, all of their chase, all their bank of America, wealth front fidelity, and constantly scanning their systems every day, building out investment systems for themselves. They've given, I literally just talked to someone who's done this. They have three different investments strategies. They have given each one of those investment strategies, a $1000 starting point.

They go into the system, they've written slash goal, turn this thousand into a million, and then they said, you know, you're going to build one system, you're going to build another, and two of the systems have lost all the money, and one has three Xs. So we're at zero right now. I talked to this person on Thursday, and so who knows where it's going to go, but that person could only open up that use case. If it had access into that person's money system, the ability to see maybe what their investments strategies have been in the past, right?

You might not want that type of use case in your life. I completely understand, I would not tell people to do that if they don't want to, but you got to know that that exists, because there might come a time where you go, now the leverage point is worth it. I see so much of this online that says, I fed clod or chatchips here, whatever my portfolio, and I 500 X, and you can't you, like, yeah, that's bull. I don't believe that. I mean, like, comments receipts, right, comments receipts show me, and also show me the three use cases that failed and lost you all your money.

“That's what I would want. But again, going back to what we talked about, there are finance use cases that don't require sharing all your personal data, and still gain you high leverage.”

But it's not there yet. It's not a hedge fund manager. There are hedge funds that are using AI. You have bridge water and other companies that have built out full AI run funds, but it's not magical, like the, definitely not. You were talking to, you know, two of the three last all of the money. Yeah, and and maybe they're going to start five more, right, like that's the, we're talking about the beginning of iteration, the beginning of experimentation. And so for a lot of people that don't feel like their first adopters for a lot of people that are not high risk, you know, maybe if you weren't the first person to buy VR, if you weren't the first person to get an iPhone, like I had a blackberry forever, until it fell on the floor, and I was like, I am not all I am forever.

I don't like they're all taking me back. I, I am not always the first adopter in hardware, but I am an early adopter in software. If you're a later adopter in both, let everyone else get the mess out of the way, and you can join in later, but you're going to miss the first six, 12 months, whatever, of leverage.

I'm fine with that. Right. I want to be the first like patient with an AI surgeon. I'm good. I'll be the, much later.

“But would you be the first, you know, AI tailor recipient? If it was like fixing a pair of cheap pants?”

Yeah, low risk. Right. And for some people, the way that they view a thousand dollars might be extremely low risk. And they're going to get those iteration cycles. Maybe for other people, they're going to go, you know what? I'll test on a hundred bucks. I'll spin up a separate bank account. There are people who've spun up separate email addresses for their AI agents, separate bank accounts for their AI agents. There are people creating full identities for their AI agents, different devices, so that they can fully sandbox it out.

Right. That requires a little bit of technical knowledge, but that person is in a more safe, secure, guard-railed way, getting those experimentation cycles. Like, speed of experimentation, speed of iteration is the biggest mode in 2026. So even if it's a small experiment, even if it's over there, not in your mind device, I would encourage people to experiment. I mean, risk or word correlated, and so, yeah, higher risk potentially higher reward. Yeah. When do you think it's going to get there when it comes to wealth management investing? Because you laid out, you know, what it looked like for the last five years.

Very eloquently. What about 2026 and 2027?

So first, I already think it's there to a certain extent. Right. We have the ability to understand massive amounts of structure data.

It's just that people are diversifying it out across, you know, ten different paths, and then figuring out best of ten.

Whereas with humans who have seen decades of data and they have a million other clients, they're going to optimize it in a different way.

So one, I think we're already there. The thing that I am looking at for that next level to say, maybe you don't need to see everyone else's data, maybe you don't need to run ten.

“I'd be looking number one at coding capabilities of AI because that's how we connect into systems. We're already there. We're 94 plus on the best AI software engineering benchmarks.”

I would be looking at the trust levels in systems from a finance side, which we are very low on as a country. I would be looking at regulation and whether AI is even allowed to, which we are not there as a country, others might be the punch. And the last thing that I would be looking at is the ability to self learn. So like I have my accountant, right, or my financial advisor, like I could call up that person, they know how I've run my money for the last five years. They know why I picked this fund over that fund.

They know when I started to go nuts on single stock because I allowed myself to have that risk.

And so they can get that compounded gains from learning.

AI right now, the ability to self learn, we're not there yet. And 2026 will be early innings of self learning. But I think you could really, really imagine, again, minus that trust and regulation side because I can't quite predict this society side. But you can imagine a system from the coding math and and human personalization side that could hit that in 2027. For private wealth collective, which is our wealth management company, we use AI all the time, but there are also humans. So I think it has to be a both situation. You can have alpha or an edge by analyzing big data sets.

But you also need a human to take a look at what the strategy should be. And somebody needs to pick up the phone and a lot of money is emotional. That's a, you are spot on, especially on that last piece. And we talked about AI companionship stuff. Like, I still want to complain to my accountant when I have to make orderly taxes. Like, it still feels like I have someone in on it with me.

However, you might have someone at the system level, not the fund level.

Well, I might have an AI managing one tenth of my high risk money. And I might have a human who's looking over the other nine, the other 90% plus that ten.

“Do they have to go into the weeds on exactly how AI is managing that money?”

I don't know. Like, I do think that there are high risk people who invested all their money into 401(k) in crypto, right? To go to these people and they say, it's okay if you have 20% of my money run by five different AI experiments. And if it is a good idea, they will gain outside leverage. If it's about idea, we're going to hear about it in the news. It's, it's a different ballpark for people that can afford to lose that no problem. When we had Jason Oppenheim on the show, he said that he thought humanoids would replace lower and real estate agents probably in the next 10 to 15 years.

Do you think he's right? I hate humanoids. I think that humanoids are not a technology that people even want. Like, you talked about how money is emotional. Anything that looks like a human is emotional. Hands are like some of the most human communicative things.

“We're like the only ones with opposable thumbs, right?”

Elon, I was talking about that with the, with the robots. Yes, the hands is the hardest part. Yes. Yeah, grasping is like one of the hardest things for you. But it's like, I just, I don't think humanoids are not going to, this is my own prediction and I'm happy to be proven wrong.

But I've posted my AI predictions for the last eight years and they've been pretty good. I don't think humanoids are going to be affordable mass market products in the next 10 years. Really? Yeah, it's, and by mass, mass market, I'm talking about like under $5,000. Okay.

So right now, if you spent $50 to $100,000, you could have a humanoid that doesn't really do much stuff. But it looks like a human in its shape. It can maybe fall and come back up. It can maybe hold a box. But if you look at like humanoid GPT, research paper that just came out.

Humanoid GPT is the ability to show it a new task. And then it can mimic you without falling or without making a mistake. It was right, something like 92% of the time. Humans are 100%, right? Assuming that we can model, assuming that are like a knees can bend that much, right?

And so the risk is still way too high. The cost is still way too high. The product generation cycles too long.

“I think the interesting parts of robots is not a thing that looks like a human.”

I think it's an arm coming out of your kitchen or something that is like an AI vacuum or more purpose built, probably smaller and probably less creepy. What about a babysitter? Just be there. Would you hire an AI babysitter?

I would not be the first to hire an AI babysitter.

That's for sure. I mean, so first of all, we have people that are already using AI to a certain extent. Like there's already baby monitors that have breath detection. I'm forward with that. Detecting whether your baby maybe has stopped breathing.

And we'll alert you. So there's already AI in these components. Do I need like a Suzy from the Jetsons to where a doily and like hold a child? Absolutely not. Okay, so we're bearish on the prediction for human rights.

I've been hearing a lot that plumbers, electricians, the boring businesses are more AI proof. Because it's going to be a while before AI is going to take over those kinds of businesses. Do you think that's true? Anything that is high liability, high regulation, extreme levels of human trust required, or something with like teeny moving.

Anything that requires like very small movements of the hands like fixing a pipe. Yeah, those are generally going to be safer.

“I think that electricians, plumbers, all of the boring businesses which I don't actually find in that boring.”

I literally just had like a long conversation with an HVAC engineer and he's like,

I've never been asked as many questions.

So I think all trials are very interesting. But those people are still being company. The big firms are buying up all of these businesses because as a as a consortium, you can gain mass leverage using AI or at least digitization across all those as a group. The ability to have a shared budgeting system or a shared prediction system

or the ability to have this client actually come over here and be a client of that plumber instead of that electrician. Like there's a lot of network impacts. Anyways, I think that plumbers, electricians, all these folks, you can imagine a world and I worked on this use case more than about 10 years ago at IBM where they were using AR to be able to augment things that they saw in front of them.

It was the ability for home inspectors.

We built that whole thing.

The ability for home inspectors to take photos under the sink of backyards, to look at satellite footage, to look at different types of ceilings, to be able to run their inspections a lot faster. There's been AI in the boring businesses for 10 years. It's just that the AI in that world is ading or augmenting the person

or allowing them to crank out paperwork a lot faster and not doing the actual fixing. Opening AI and Thropic, they're all filing to go public.

“Do you think that's going to fundamentally change how they operate as businesses?”

So any business going public has to report a lot higher transparency than we've had from these labs in the past. And so I think that that level of transparency, you could imagine that that would impact the way that they're running their business, not just the amount of reporting that they're giving, but the types of business lines that they might go after.

And so you already have OpenAI who's hired the founder of Ironclad. You have them opening up new legal business lines. You have them hiring ahead of ads and building on ads business. And so you're going to see a different type of commercialization from the two different labs. I think it will be interesting to see how that changes.

I can make predictions, but we haven't really seen any company at this scale go public with this much on the line. What I hedge against and what I worry about is people that are like, "See, AI is a bubble. I knew it." I'm very worried for those people. Is it a bubble? It's, so, okay, bubble means what in your mind, frothy overvalued permanently overvalued, right?

Until it pops. And then does it take decades to come back? Does it ever come back? Does it matter for a bubble? It changes. Okay, dot com bubble.

Yeah, now dot coms are the-- Yeah, but not everything with the dot com is going to be successful. Right, and it took 15 years as a link for it to come back. I think that people are going to see Stocks dip and across Google meta,

And they're going to say, "See, that proves my point."

And there's a lot of, again, emotion correlated not just to money, which is a big part of this. But AI, like, people clearly people don't want to be replaced by AI. So any data point that they get that proves their point, they're going to come out and say that. I'm actually biased. Exactly. So I really, really worry that we're going to see some dips in these stocks

and that people are going to have mass confirmation bias, and that those businesses are going to be worse off as a result.

“I predicted I think it was three years ago that both these companies were going to be worth the trillion plus dollars.”

And everyone thought I was a lunatic. Did you invest in them at the time? No, they were private. But you could still invest. I'll throw $5 at all those SPDs.

It was like the fund of the fund of a fund. People are going to lose that. That's a whole other episode.

Yeah, I think, you know, I predicted that these companies were going to raise over $100 billion.

Everyone said I was crazy. I predicted that both companies were going to be worth a trillion plus dollars. Everyone thought I was crazy. And here we are. Both companies are likely to go public.

Both companies are likely to be valued at over $1 trillion. Are you going to buy? Absolutely. You're going to buy. And I want to go this is not investment advice, right?

But my risk profile is very different. Though other way of looking at it is imagine that you get hired at open AI, the day they go public. Or you get hired at anthropic, the day it goes public. Or SpaceX, the day it goes public, whatever. And a large portion of your compensation is equity.

Would you be happy if that portion was 90% instead of 50?

“If the answer is yes, that should be a trigger in your head that you should buy when these companies go public.”

And that you should hold not for three months. But when I buy single stocks, single stocks, I am holding them on for years. I'm not buying for short term runs. And so that is the reframe that I would have everyone be looking at. So you're long term bullish.

Absolutely. Because again, it is a mass horizontal. Like it is like electricity. It is like fuel. It is so hard for me to imagine that AI will be any less performance than it is now.

Any less pervasive than it is now. And if you agree with me and you believe that at the very least it's neutral for a couple years and then it goes up or whatever. Then that also makes sense that you would want to invest in that as well. So what are people calling you crazy about now that you think is going to happen? Ooh.

Well, I mean you already called me crazy once in this podcast. Which is talking in public to AI. I have a very hard time imagining that anyone is typing for most of their day in the next couple years. So bullish on like the whisper.

“I think every single thing is going to be voice enabled.”

I have a very hard time imagining that phones with screens are going to be our main device in five to ten years. I think people still have smartphones right in the same way that people still have flip phones today or facts machines today. But I think voice is going to be everything. I think multimodal will be everything. And that was another piece of that kids.

Multimodal is I mean how humans are. We're not just in words. We're also in visual. We're also in audio. Like, I'll call my sister and I'll be like, oh, what's that word?

That doesn't mean like, beep, beep, but it means like, wow, wow. Right? And she's like, oh, okay, plaster and I'm like, yeah, I knew it. And we're just extremely multimodal beings. Like we we speak in visuals.

We gesture to our friends when we can't talk across the room. We make funny little. Yeah, though. We're multimodal beings. So I think every single thing is going to be multimodal.

I think voice AI is the driver. I can absolutely picture every single person on the streets. Talking to AI in their car, talking to AI on some ways, talking to AI. Do I want everyone talking all the time around me? It's a different question.

Not really. But voice is four times faster than writing. Well, my daughter need to learn how to drive. Well, my daughter go to college. Should she learn to drive?

Might be a different thing. Because I think it's actually really good hand-eye coordination. Will she need to learn to drive? My guess would be no. Well, there would be college.

I think it'll look very different. I also like New York kids don't learn to drive. And they clearly have grown to be like badass adults. But I do think that like the thing that I would say in that car scenario is like, I would still want to teach my kids motor skills.

Like my dad took me and my sister out throwing baseballs and softballs.

Every single week is a kid.

“My hand-eye coordination is un-freaking believable.”

Because of that. And look at you now. And look at me. Look at me now. Darts expert.

I think so, motor multimodal for sure. What was the second one? Cars.

Do I want them to learn cars?

What was the other one? Go to college. So have you read about Alpha School? Sorry, just actually going to bring that up because I vastly between. I want her to go online when she's 35.

And like, how fast can she get into the Alpha School? Okay. The part that is AI forward. School, new way of thinking, new way of projects. You're opening one up in Santa Monica, near me.

Yeah, and they're opening up. We're all one in New York. I think things that I like about the Alpha School model. One is that it's a lot of group work. So again, that that human connection, the ability to say let's work on our project.

“Like, I don't remember that many group projects that I did when I was seven or ten.”

It was mostly when I was like in college or grad school. And so I want a lot more group work. I also, they still have adults in the room for I think it's two plus hours of instruction. It's probably more capitalism driven than the average person I want in education, because a lot of what the outcome is is building businesses.

You know, whether that's like a lawn mowing business or like whatever AI services business. But it teaches people to gamble on themselves, to be crazy, crazy high curiosity, to have high agency, to deal with failure, to think in systems, to work with others. Like, I would absolutely want my kids to learn that. I look forward to seeing other models of schools.

So it's not just Alpha School versus traditional schooling. Yeah, I just, I want to see more.

Like, I want to see sports oriented schools that every single person first primarily is on a sports team.

And then is learning on side. My husband wouldn't be obsessed with that. Okay, to be continued. Yeah, next episode. Ali, we ask all of our guests for a final tip that listeners can take straight to the bank.

“What's an AI hack that we should use to maximize our financial potential?”

So we've touched on this, but I'll hammer it home now, hammer it. 2026 is not the time to look at the list of tasks that you have on your calendar and ask AI to do one of those and call that a win. People who are time rich have figured out how to buy time back and they are figuring out what business models actually win in 2026 and beyond. And the weirdest mindset shift that I think people can make is no longer charging by the hour and using AI and using code probably to figure out solutions for their business to be able to reinvent their business.

In a way that does not reduce them into hourly service providing. It is project based. I want everything project based. I want everything outcome based. And I want you constantly talking back and forth to AI building up context and building systems that allow you to do that thing rather than you being in the weeds doing the thing.

Like a weirdo. Like a weirdo.