Muse has the potential to be the behemoth winner in this space, and it won't just win in the personal agent space. It has the potential to knock the frontier models well back on their heels.
And they're already pushing hard into small businesses.
Yeah. Enterprises are really—it's a really tough sale because they have all kinds of other moving parts that you have to move around before these things can be adopted. But talk to my plumber and he's like, 'Yeah, we did put this in last Tuesday, we started using this. It's great.' There is no long process. And of course now he's out telling all of his buddies in plumbing, 'Rather than hiring an assistant or having to spend all day like reading your texts and emails and trying to keep track of bids and things you need to do, this thing does it and it's free.'
Welcome back to the Dick and Paul Show. Today—sorry, I always forget to introduce myself. I'm Dick Costolo, former Twitter CEO, entrepreneur, founder of several companies, and currently venture capitalist at 01 Advisors.
Did you mention that you're Dick Costolo?
I did, I think so. Did I not mention I'm Dick Costolo? Dick Costolo.
I'm the other one. I'm the one who's not Dick Costolo. I'm Paul.
Podcasts, yeah, yeah, yeah. I'm venture capitalist, blah, blah, blah, blah. All that stuff.
Don't belittle the profession. Come on.
I've been reading—there's this Twitter feed from a guy. Did I send this to you? It's called, uh, I don't know, like basically dunking on venture capitalists.
Yeah, it's been around for a while. It's very good. You're just discovering this?
No. VCs congratulating themselves. VCs congratulating themselves. It's been around forever, but I actually forgot it existed and then someone sent me something from it the other day where—I'll just completely de-genderify everything—but someone was at a gym and saying things about how they were going somewhere else to raise money.
And they just handled it tremendously. Like it was this idea that everyone should be so excited that you're going to Abu Dhabi to raise money. And of course the meta comment on it was, 'And then the gym, you know, everyone fell down applauding this bold maneuver.' It's just—venture capitalists are ridiculous.
All right. Well, on that—[laughter] That's right. Was that the topic?
No, that's not the topic. Today's topic is personal AI assistants. These digital assistants such as Muse, Instinct, Haven, Town, and—
Dot. Just to name—Google has one, I don't know what the name of it is.
Which, by the way, is one of the problems we will discuss about Google during the course of this conversation.
AI has one, but you're not allowed to have it.
Yeah, you can't have it though.
What are you talking about?
It's on like an enterprise plan.
Oh, I see. You can't have a person—you can't actually have a personal AI assistant.
If you're a person, you cannot have it.
If you're a person, you can't have a personal AI assistant, Dot. You have to have an enterprise—
Just the people who don't want it, only they can have it.
Yeah, only the people who don't want it or need it because it's superfluous to what they're already doing, they can pay for one.
They want an impersonal assistant.
So, one of the interesting things that we can start with about—you and I have discussed about these personal AI assistants—where they really run into trouble is the concept of Chesterton's fence. Chesterton's fence is the concept, and I'll talk about the origination of it. Chesterton's fence is the concept that if you don't understand why a law or an institution or some physical structure or custom was originally put in place, then you probably shouldn't change it or remove it without first—
understanding why it's there in the first place.
And it comes from the British author, British G.K. Chesterton.
Author of the Father Brown mysteries.
He—I think it was his book called The Thing. He posits: imagine a man comes across a fence in the middle of a pasture, Chesterton's fence, and says, 'Well, there's—' looks around like there's absolutely no reason for this fence to be here. Um, you know, now that I own this property, let's just take this fence—
Or—and the thought being, Chesterton's proposal being, 'Look, it may have been there to like prevent the cows from escaping. It may have been there to prevent plants from being damaged.' Like if you can propose to me why you think it might have been put there, then we can discuss taking it down. So, how does this apply to AI systems? Well, in all sorts of ways. So, it turns out there are all sorts of reasons that companies and governments and others institute some layer of friction to accomplish a task, whether it's making and canceling a reservation, whether it's moving my savings account to a higher yield savings account for some short period of—
Entering a lawsuit in small claims court.
And we will get to that. Entering a lawsuit in small claims court. Um, there are all sorts of reasons there's a certain amount of friction associated with that. And one of the things you see in the early days of these personal AI assistants is people jumping for joy that—you're not going to believe this—I was able to have Instinct or Muse or whatever file a lawsuit in small claims court for me, and just all I had to do is tell it to go file a suit in small claims court and it was done.
So, the first derivative consequence of this is obviously, well, that's not going to be the way this works for very long because the small claims courts will be inundated with lawsuits and claims that were never—you know, 'I want $48 back from Comcast.' This is going to go away shortly. And we can go on and on about a number of these other cases.
Yeah, and then they're endless. And I think—I mean, and it is, as you said, it's kind of an example of a friction. I mean, the other one that came out, we were just talking about on Friday, I think it was B of A announced that you can no longer use an agentic tool, one of these personal AIs, to access your banking information.
And it's not just B of A. There's a few others now have come in and blocked them.
I think JP Morgan, B of A, and Citi have all said you can't do this. Now, the immediate response from the agentic crowd is to say, 'Oh, those bastards,' because, you know, they're preventing me from being one with my money and agentically, right, moving it around. And you know, fine, you can take that point of view, but the problem, of course, is—imagine why—turning it around, how do banks prove that they're not about to go broke? One of the ways they sort of demonstrate they're not about to go broke is by having deposits.
And if your deposits can go flowing in and out whimsically, like, 'Yeah, you know, I'll go put over here, over there,' all of a sudden—
And be slingshotted around in sort of a—in a fraction of a second.
Fraction of a second. And multiple times a day, swapping between high yield accounts.
Right, right. This is hugely problematic for these companies. And so, friction, back to Chesterton's fence, in the context of banks is retail deposits, you and I versus, you know, some giant organization. They rely a little bit on you and I not moving your money constantly all over the place because it creates a sort of a safe deposit base for them, right? So, you can look at it and say, 'Oh, they're just bastard predators, right? They're just relying on me not moving my money.' And now that I have an agentic tool, I can move it freely and suddenly I escape the evil clutches of these giant banks. But turning it around, what's good for you isn't very good for the system.
Because now all of a sudden you have all of these deposits flowing around like water inside the system, and this creates all kinds of liquidity and insolvency problems at scale when deposits are moving overnight, and all of a sudden banks don't know from day to day like how permanent their deposit base is. So, what's that going to lead to? Less lending, right? It's going to be less money in the system because now I have to hoard even more, right? And I have to hoard even more at a higher rate—percentage because this money's being slingshotted around the system.
So, a lot of these agentic things sound great as an individual, but then as soon as you extrapolate and then say, you know, restaurant example.
Yeah. I'll give you my restaurant example. The restaurant example we've been discussing. And by the way, these things are great if I'm the only one using it.
That's it. Right. [laughter] Right.
And of course, that's not what happens. The restaurant example is, as many people understand, the way a lot of restaurant reservation systems work is you can book a table and if you cancel, you know, just over 24 hours in advance, there's no charge. Well, that was when it was hard to go reserve a table unless you had, you know, where you had a time written and had an assistant who you beat into the ground—you wouldn't ever go reserve 19 tables just in case Thursday night.
And then I'm going to decide later on which one I want to do. And then my agent is going to tell me 26 hours beforehand which of the 19—which 18 of the 19 tables do you want to cancel. It's only—that only happens for a very short period of time before the restaurants either say, 'All right, it's now $50 to book a table or more or whatever.'
Right. And there's no cancellation.
Yeah, and there's no cancel—or the cancellation period is now a week and a half, 10 days. So, all by way of saying, one way of viewing the early days of these personal AI assistants is it's going to be a case of this is why we can't have nice things. And everybody else, all the rest of us, are going to be forced into this world of you can no longer book a table here and cancel 24 hours in advance because this thing has happened to us. You actually can only move your money from X account to Y account once over the course—or whatever happens. There will be some other—there will be a next step beyond just blocking these things that say, 'All right, well, the new rule is X,' and it will be basically less fun for everybody than it is now.
But the example, and you gave this earlier, was that let's go away from ones where it breaks so obviously. Amazon's a good example, right?
Yes. So, Amazon's a great example.
How does it—you know, the response from the agent crowd is, 'Wait a minute. I was using Amazon or I was using my tool to shop on Amazon and enter bids and so on. How does it hurt them? Look, I'm doing business on Amazon. Why are they blocking me?' And we'll go to another example after that, but I mean Amazon's a good one.
Amazon's a great one. In fact, Elon posted on Twitter, 'Well, that's ridiculous. Maybe we should build a competitor to Amazon.' Well, it turns out that when you go to Amazon, they make some very large percentage of their money now from ads. When you go to search for, you know, I don't know, a hair dryer, there are a bunch of sponsored results at the top and those are ads, and, you know, they get paid when someone clicks on one of those on a cost per click or cost per, you know, CPA basis. And if the agents are just coming bypassing all that and never seeing—obviously never seeing ads, Amazon's like, 'Well, we're not going to just give up 50% of our business.' Now, the response from the 'Hey, you're blocking my agent' crowd is like, 'Well, that's—you know, I don't think they understand what's going on here.' And even if they do, they're like, 'Well, that's stupid. We'll just go do this now somewhere else.' But what will happen, of course, is—I mean, Amazon's not stupid. They'll go do a deal with the company that's running the agents that can afford to create the most value for them.
And it'll—and that'll be it.
That'll be it. That'll be it.
And that will most—more than likely be Muse because Facebook has almost infinite resources and can throw 200 billion against that and no one else can.
You see, and what I find interesting is the response is they just—it's the usual, they just don't get it.
Yeah, they don't get it. In fact, I was at dinner with an investor and a CEO of an AI company, and their response was, 'Amazon just doesn't get it.'
They're just—they're living in the old days. They want it to happen this way. It's not going to happen this way anymore. It's going to happen this way. And the problem is, of course, I get where that's good for you as the agent, not so much for me. And I'm much larger than you are, so why should I facilitate whatever you're doing? And this is the problem. You see this over and over. Like the Amazon example is great, but the one that I think is the most interesting and has seen almost no conversation is this idea that one of the primary cases for how people use this stuff—like leaving aside how individual consumers use it, but if you talk to small businesses or others, because I was talking to my plumber the other day. He said, 'You ever heard of this Muse thing? It's really great. I use it to respond to incoming requests and all that.' But he said, 'The best thing about it is it reads my email.'
