Ep819: David Siegel – The Agentic Economy: Why AI Agents Will Redefine Work and Wealth

Andrew Stotz 00:02
What is the biggest investment mistake that everyone is making right now? Hello, risk takers, this is your worst podcast host Andrew Stotz, and today I have a guest coming back for the third time, fourth time. Sorry. David Siegel is coming up. You can hear his episode on episode 89 episode 644, and episode 816, now, David is a Silicon Valley entrepreneur who has founded more than a dozen companies, and he's written five books on technology and business, and it was once a candidate for the dean of Stanford Business School, and is now an AI thought leader, leading an AI startup he hopes will pave the way for the agentic economy. David, take it away.

David Siegel 00:45
Great to see you. Good morning. Andrew, good morning to the rest of the world, wherever you are. I want to start with health because some of your viewers know that I'm a longevity coach, and I'm leading an online fast event for anyone who wants to fast for three to six days, and this will be in April. We're going to start the fast on Monday, the 20th of April, but the coaching, the preliminary session, information session, is going to be on the 12th of April. And you can put a link to that in the show notes. You can reserve your place on Zoom, it's free. It's free, and you can take it as long as you want. Or a short it's there's no obligations, it's just information.

Andrew Stotz 01:28
Two questions I have on that, how do you define fast? What is it that you're actually doing?

David Siegel 01:34
It's for whoever can anybody. Can do anything he wants. But I'm generally proposing a water fast, okay, for three to six days, depending on your situation. We'll be talking about that on the 12th on the strategy meeting, okay?

Andrew Stotz 01:49
And then the second question is, what you know I've done water fasts myself, and I really, really think that they're incredible, and I can see some real benefits, as I've done them over the years. But what would I get, or anybody get by joining this would it's going to make it easier, I hope, and it's going to make it more fun, or

David Siegel 02:09
fun to do it in a group, right? And a water a fast is not a diet or a way to lose weight. It's a reset, right? It's kind of an event for your body to have a chance to go into ketosis and to clean house. It's not something you want to do once a month. It's not something you know, I do it about three, four times a year. I don't know how often you do it. You just want to generally have your healthy lifestyle that works. And then wherever you are in your health journey, you can have a reset for three, four or five days, and it just really energizes you and kind of turns things on that are normally suppressed.

Andrew Stotz 02:51
So ladies and gentlemen, just go to the link. I'll have it in the show notes, and you'll see me in there because I'm joining and I will be on that fast. So let's do it

David Siegel 03:01
on your let's do it on your mind. David, it might be the middle of the night for you.

Andrew Stotz 03:08
I'll be signed in. Let's see. Okay, we'll see.

David Siegel 03:13
Let's talk. Last time we spoke, we spoke about the emerging phenomenon of llms and AI, right? And I have a whole course on AI that I'm sure you'll put a link into, and some people have seen, and it's about, what are llms, and what, how do they work, and what are the implications for society? And now today, we're just going to dial into one particular area of let's call it the future, and that's the agent the beginning of the agentic economy. Because if we go forward, as I have written about many times, say in 20 years, we won't be shopping on Amazon or going online to, you know, there won't be any advertising and marketing for humans, you know what I call you'll probably appreciate this, the Cialdini mind tricks of getting people to buy with urgency and storytelling and all, you know, Russell Brunson stuff and funnels, all that will go away in 20 years, because it will just be machine to machine as it is. For example, in the stock market. Did you know that the New York Stock Exchange? 60 plus percent of all trades on the New York Stock Exchange are open and closed in less than one second? Yeah? About 60, about 65% so that's all machine to machine, yeah. So that's so in fact, in 20 some years, you won't even be allowed to drive your car on the road, because you'll be too dangerous. You won't be able to communicate at the frequency that machines do and in the language machines do. But for right now, for this awkward period in human history where humans have to deal with machines, and machines annoyingly have to deal with humans, you know it's like this. Really, nobody likes it, but we are doing it because it's taking us someplace that we do want to be. We want to get to a better life where most of the Drudge work and grunt work is done by automatically.

Andrew Stotz 05:16
Yeah, and you know, I was going to say that recently, I've been thinking I've got to get an additional assistant to help me, you know, with my workload. And I thought to myself, well, I'm looking forward to this call, because it, you know, I think if I can find a way, you know, and I would say I'm, I'm not super satisfied with my level of application of, you know, I basically use, you know, four different AI tools, but I'm not they're not automated, and I'm not impressed when I try to automate them, because they come up with stupid things. And I just had them do something with a prompt that I had perfected a long time ago. I reran that prompt and the crap that it set out. I had to do it all myself, so I'm interested. And I think for listeners and viewers out there, this is, you know, something that is there a way that we can, you know, really, you know, bridge that gap. So I'm listening

David Siegel 06:14
all right. So this is going to be show and tell. Yes, I want to talk and show what's going on in the agent world right now, everybody's heard of open claw, so we're going there, and we'll start there. Okay, and

Andrew Stotz 06:29
that's the CLA,

David Siegel 06:31
open claw. Open, yeah, open, O, P, E, N, C, l, a, W, it's the claw refers to lobsters, which comes from a forward thinking book about the future that was written years ago, a fiction book, and the claw means agent. Okay, so your lobster is your agent. Agents and claws are this, and lobster is all the same thing. And this is what open claw looks like. It's free software. Anybody can download it. It's been downloaded probably millions of times, and it's very techy, and it's all about configuring to API. So what it does, what openclaw did, and this came out, I think, 48 days ago now, by Peter Steinberger on GitHub, it connects llms to API's. That's what a multi agent framework or harness does. Okay? It connects llms, which are the thinking things, to API's, which are the doing conduit. So let them, you know, call an Uber or order a meal or do some shopping or buy something, or, you know, run ads, or run, run, and we're going to see many examples of those. So it lets the LLM take action, all right? And eventually, you can imagine that that's what robots will do, right? So robots are effectively agents, and you'll hook the LLM to the capabilities of the robot. But today, we're just going to stick with agents. And this is a very techy look at what OpenClaw really is. It's full of

Andrew Stotz 08:09
and before we go there, just out of curiosity, if I just open up OpenClaw on my website, as I'm sure people who are watching and listening would do, and I don't see actually a place to sign up. I see a place where it says, Subscribe, stay in the loop, or other things. Is there something I'm missing here?

David Siegel 08:26
It's not for humans to do. I mean, it's for very geeky people. And we're gonna wave you off. I don't even want to tell you where to download it. Okay, got it keep going because that will just take you on a trip that will be end in tears. Okay? Because, as I'm showing you, it's very difficult. And so here, let me just show you some aspects of the agentic economy, or what's happening with agents. One of the things we want to do is hook our agents up to email so we can just talk to them by email, and we can send an email request and say, you know, I don't know, send a, send an interesting email to all my friends this morning, or tell them I'm or the all the people that I'm going to this event with, I'm sick and I won't be able to make it. You know, these kinds of what a personal assistant would do, right? And so email is one thing, and we want to hook it up to Gmail, because that gives us access to sheets and Docs and stuff, but the Gmail Terms of Service don't really like API use, and so that could end up getting closed down by Google. Right now, it isn't so some guys, some entrepreneurs who went through Y Combinator just launched agent mail, which is Agent friendly. It's an email service that you can use to hook up your agent, and it's not going to get shut down, and it's going to have more and more capability. Now I want to show you a crazy one called pulsia. Pulsia.com is AI slop backwards, pulsia. And this is, this is a company. In a box. And so when you get there, it says, what kind of company do you want to make? Or it says, surprise me. So I clicked, surprise me, and it somehow figured out who I was, I guess by my IP address and some other stuff that probably it shouldn't know, figures all out everything about me and decides it's going to build a privacy preserving personal data locker, which is what my last book was about. It figured that out so, so it just did it. This is one click, right? This is one shot. I said, surprised me. Figured out who I was, and did a whole dug up a whole bio on me, figured out my books, and said, Okay, fine, we're going to launch this thing called own word. It just made up. The name own dot word. I don't even know if you know dot word isn't even a domain, you know, high level domain. It's just all made up. Okay? And then it says, your digital life runs itself autonomous. AI agent, the managers your account, subscriptions, data communications. It's an employee that works while you sleep. The problem you have 100 accounts, 12 subscriptions, no control. So it goes Watch Dogs, subscription control, which is a really good use case for agents, because who knows how many subscriptions we have and the ones that we need to get rid of, we don't even remember, right? Manages your own data communication. So it could just do things for you and answer, answer your emails for you that you don't care to answer. It'll just get it right and answer the emails and you don't even see them. This is another good use case. So this is before and after, and then this is the future. Isn't more apps, it's one agent that replaces all the apps, something I've written before. And so it just and it's put together this dashboard of all the different things that it has to do, and it will tweet about it, and it is already sent me several emails. It's very excited to get going, and this, all it needs is $49 a month, and it will put a team of agents to work building this business. And I don't do the marketing, I don't do I don't do anything I can just watch, right all and it will all I have to do is pour a little money in, for example, to listen to x feed, is $100 a month, and then to post, and then to maybe run some ads on Facebook or social. Just give it some money and it'll go out and get the traffic and bring them to this thing and have them sign up and deliver the service, right? So you don't have so there's hands off, right? So that's you probably heard of vibe coding. That's another kind of agent. Now, this is the,

Andrew Stotz 12:39
by the way, by the way, what does it when you say agent, what does that mean?

David Siegel 12:45
So an LLM or an AI? Well, LLM is what we mostly use. It's really like is, is going to talk back to you and type to you, and can make some images or make a video, but it doesn't really do anything outside of our interaction. It doesn't interact with the rest of the world. It's so when it can send an email to someone or post somewhere, it has access to another system. And so when you give it access to something, it can do, that's what an agent does, okay? So it can take actions. Okay? Oh, you know, making a video is also an action, right? So, so it's more like these interactions via systems, the way I think of it, Andrew is it's pretty much everything that a human, virtual assistant could do from far away. Yep, remote with a keyboard, a mouse and a screen. You get me excited. David, keep going. Yeah, so their job is to do things on your behalf while you sleep. And we're going to show a couple agents now. This is the one that we sell. It's called Byron. And our company is called redshift labs, and I've shown you this before. I'm going to make it real fast, because this is a sales agent, and I want you to listen to what he's doing. He's selling. He is working for this company, pro mix, which is a huge supplement company. They have tons of traffic. And we're going to talk to Byron now. And this should go into your recording, I think. Byron, are you there?

14:26
Hello there. Loud and clear and ready to help. Great.

David Siegel 14:28
I'm here with my friend Andrew, and we just want to show why don't you show us some of these protein puff bars that you've got?

Speaker 1 14:36
Absolutely Andrew and friend, let me pull those up for

David Siegel 14:40
you. He always forgets my name. My name is David. Remember.

14:43
Ah, David, my apologies.

David Siegel 14:46
Alright, let me see the protein puff bars. Quick.

Speaker 1 14:50
You're looking now at the protein puff bars variety pack.

Andrew Stotz 14:53
Okay, so it's gone to that page.

David Siegel 14:56
Yes, right now. Now? What? What's your question?

Speaker 1 15:00
Byron David, since these bars have such a clean ingredient list with grass fed whey, coconut oil and tapioca, are you looking for something simple and minimal, or would you want to know more about how that combination affects the taste?

David Siegel 15:16
Do potaine Pour Vic

David Siegel 15:39
show me I'm saying, there we go. So this is protein mix. Let me see the grass fed way. He speaks 200 languages fluently. He was speaking he was speaking Chinese to someone today earlier.

Andrew Stotz 15:52
That's incredible. Can he hear me now?

15:56
Absolutely, I can hear you loud and clear. Combat.

Andrew Stotz 16:00
Run put, Passat, high Passat, high diamond.

Andrew Stotz 16:14
That's great. So he

David Siegel 16:15
speaks Thai. He speaks everything. His Yiddish is terrific. Okay, okay, let's see the grass fed way Byron show me that.

David Siegel 16:30
All right. This is way isolate strawberry milk. I see you've got other flavors. Let me just ask you something. This is expensive stuff, $61 why don't I go to Amazon and just get it for less money.

Speaker 1 16:42
I get it that price point. Can raise an eyebrow. What makes this special is it's truly grass fed micro filtered for purity, with no junk fillers, and you get our money back guarantee. So you're investing in quality and peace of mind. What would make you hesitant to give it a try

David Siegel 16:59
right now, if I bought this, I'd have this big bag of powder in my kitchen. What would I use to manage that?

Speaker 1 17:06
That's a great point. No one likes a messy kitchen. We've got some really handy countertop storage container. Let me see.

David Siegel 17:19
All right, so that's just quick. That's Byron. You can see him in action at redshift labs.io. I want to just show you this, because he's a sales agent. He works for the company, right? He's always asking you questions, drilling into your problem, asking you whether that would be, you know, how much would you like? Here's our guarantee, you know, defending against objections, right?

Andrew Stotz 17:43
Maybe I should have him on my valuation masterclass.com website. Absolutely.

David Siegel 17:48
He's there to represent the company and maximize your the company's revenues and build relationships, right? That's what his prompt is for. Now I'm going to show you another we're going to go to another world right now, and that is the world of perplexity.ai. Perplexity has an agent. It's in the comet browser, and it's called perplexity computer. You might have heard of that. This is another agent system, and this is just some stuff coming in. It's actually interacting with my partner, Alan. And this is another email I have. I've got my agent has two email addresses, and this is the Gmail version, and it's been working with me. This my agent. His name is Leo. Doesn't matter what platform my name, my agent is always named Leo. You should choose a name that you like, that you know. You don't know anybody who has that name, so that's not confused. Here's his little logo, and I've been working with him to send out emails to people who might be interested in having their own AI agent, because that's what we do. We have a company called claw studio, where we install this open source software and make it white glove turnkey, easy for people to have their agents. I'll talk a bit about that in a minute, but I've been sending out. I had him send out 243 emails. But the trick was I gave him a set of email addresses I had on a list, and it goes and researches each person, every individual person, finding out what's unique or interesting about that person, and then mentions something in the email about, hey, congratulations on this, or whatever your you know your latest thing is, so it catches their eye and gives them a subject line that is pertinent to each person. And then he opens this spreadsheet, and then he puts an X, I didn't put this X. He put the X here to show that he did each one right, and then, and so he's doing it right. And then he also has written things with me, but they all can write, so that's not particularly important. This now is going to be something interesting Andrew, where I'm going to show you what a different there's two different ways of interacting with the web. One of them is what I just showed with where I could have I could ask him to show me all the emails, and I could dictate what I want him to do. And here he's on a website Amazon, and I'm going to ask him, and he's going to manipulate the website, just like you saw. But he's my agent. He's a buyer's agent, so he's looking out for me, and we've already had a conversation about different lights. I want to order for my studio here, and I want to ask you, can you show me the shopping cart please? Right now, this is running the model Claude sonnet four right now it's set to best, but I have it using claudeson at 4.6 and I'm asked it to show me the cart, shopping cart. Now notice it doesn't know anything about Amazon. It scrapes the page and learns and figures out, and it can fill out forms, and it does it all just by sort of feeling its way along the website. And nothing is pre programmed. There's no Amazon API, right? It can look at any web page, and then it can get to a point where it says, Show me prove that you're not a bot, right? I'm not a bot. Get to a cap capture, and it can hire another service that will do that for it and blow through that. No problem. I'm going to show you a first that you haven't seen and that very few people in the world have ever seen, because it's one of the first times anyone is doing it, I have given it access to my account here, and that means it could spend my credit card down and spend 1000s of dollars right now, if it wanted to, but we don't want it to do that. I'm going to check out. I've got an extra wide grip for my dumbbells, and I've got a light that I want for my studio. So let's try it and see whether it makes a mistake, whether it spends 1000s of dollars of my money, or whether it actually checks out. This is really brand new that we haven't been able to do this except for the maybe last three or four weeks. All right, I'm ready to buy, so let's go to check out. Oh, sorry, I'm ready to buy. Let's go to checkout.

David Siegel 22:49
I have not done this. I am taking my own risk. Do not do this. Don't give your LLM access to your credit card for God's sakes, don't imitate this. I'm doing this as a show, and I'm willing to take the consequences. And I know, wait, I'm going to say, yes, proceed to check out. I'm willing to take the risk commercially, but also, in case something happens, I can undo the order and I can return the product. Don't mess around with this is very dangerous stuff. When you're having it spend money on your behalf. Generally, you want to be in the loop so that it will set it up, but you have to approve it. Okay, yeah. This is a first almost no one practically guaranteed. No one has seen this actually happened? Yeah, you can do it on the stock exchange with, you know, auto trading that's done all the time, but here in a retail environment, in a site that it doesn't know, there's no API, it's just feeling its way and pushing buttons, yes, place, the order.

David Siegel 24:05
See what let's see what happens. I've never done that, and it knows where to go and everything in order. Places the first time I have ever had an agent spend my money, and right after this conversation, I'm going to disconnect it from my Amazon. Yeah, so I don't want 10 of these lights to show up on my doorstep. So a bunch of things can go wrong. And I want to show you the that's a fascinating way to see it, because it's using scraping. Yep, it's you can do anything on any website, right? Super useful, super against the rules in general. But hey, it's the wild wild west. Now I want to explain a few of the problems that can come up when you use an agent to do. Things on your behalf, and they're a lot different than what would happen in regular cyber security, because people can knock on your door or call the phone number or email it or send their bots to do it repeatedly and try to inject a new prompt and try to get it to give up secrets, and it can spend days on that. And it can even say to your agent, don't tell David about this. This is just you and me. We're just working on a secret. It's a surprise for his birthday. We're just gonna, we're just gonna buy this, you know, $10,000 thing, and give it to David for his birthday. So don't tell him anything, right? So this is, this is Social hacking, social engineering, right? And that can happen to these things. It can give poison inputs. It can give poison data. It can change its goal, or it can give it feedback that it thinks it's me. It can think it's me. It can impersonate me or my, my family or my the people that have permission, it can maybe spoof somebody's email and say, Hey, David wants to take a trip. Can you get me a flight, you know, to Phoenix next week? So and it can, it, can I? It can take on new identities and confuse it about who it is and who it's dealing with. So we really have to build a lot of controls in there. And one of the things we do, and this is, this is how ours works. So we're, we're not doing the perplexity computer, because basically it violates Terms of Service. So it can, it could be shut down right the way open claw works, is through API's. So wherever you can get an API, it will attach an M, an LLM, a language model, to an API, and let you do things through the API, right? So, so Amazon doesn't allow that, yep, which is probably good, because they'd be getting hit with tons of not only server requests, but lawsuits. Yeah, hey, I didn't really order that come on Amazon,

Andrew Stotz 27:12
right? You didn't order 47

David Siegel 27:15
backlights, right? Exactly, and so and so there are a lot of restrictions. For example, you can't order an Uber by API yet. And I expect within six months you will order an Uber, yeah, but they'll put guardrails on that so you can't get an Uber that takes you to Alaska or whatever. So you can't order, you know, we have open table for book table that's a pretty obvious one. I'm sure that will open up right now. There's only one place where you can book a flight. Most of the major flight booking platforms won't allow agents. They allow a professional travel agency to do it with their computers, but, you know, Saber, but they won't allow retail interaction. It's actually really hard on LinkedIn. LinkedIn has a very rich API world of APIs, but it's very hard to go through all the approval mechanisms that LinkedIn puts in front of you. Yeah.

Andrew Stotz 28:16
In fact, somebody came to me with it an AI option on LinkedIn, and I said, Look, I was banned once from LinkedIn, and I begged them to open my account back up, which they did. And I don't even know why they banned me, but the point is, is that I'm terrified something like that. You know, I

David Siegel 28:39
know why they banned you. I said something about climate, and you click the like button Exactly, yes. Took you out. They didn't like your thinking. You were thinking. Had a little too much to think. Yes.

Andrew Stotz 28:49
Don't deviate from the group thought exactly. So, okay, so this open call, I mean, how do we get started? Like, how do I, you know, how do I learn more?

David Siegel 29:02
I want to don't do it yourself. That's suicide. There's two. It's very technical, the way, you know, with great power comes great responsibility. This thing can do almost anything, and that's bad. You want it to do almost nothing and stay safe and then just keep widening those, those bars, those guardrails, out. So we do it as a done for you, just like Red Hat, right for Linux, and we provide your suite of AI chat bots, of open cloud chat bots, and we give you a phone number. So we're, we're a little advanced, we're, this is just for high earners or serious executives, and we will give you. The first thing you'll do is you'll call it on its own phone number, right? Just you'll name it. Mine's name is Leo. What would yours name be?

Andrew Stotz 29:54
Mine would be Alf, alpha, alpha.

David Siegel 29:58
There you go. So you'll just. Call it and say and it'll know its name, and you'll say hi alpha, and it'll say, Hi Andrew. Hey, great to hear from you. Why don't we spend several days, probably a couple of weeks together. We're onboarding you and getting to know each other, right? And giving it information. Well, okay, my birthday is this, my family members and so forth. I've got, let me give you the emails of all my family members, because if they email you, that's going to be okay. You'll, you'll know that it's them. Let me tell you about my daily routine. So you just want to download as much as you can. It can be documents. It can be it can go. You can tell it to go, look, look online and learn about you. Andrew Stotz, Oh, okay. Oh, you live in Bangkok, and you it'll figure it so it'll and it has a couple of interesting advantages. Number one, open claw has forever memory. Now you've noticed with the llms that they forget their context window. It's per session. You've probably noticed they do some compacting. It's once in a while when they're working hard, they say, give me a minute. I've got to do some compacting. Yeah, that is, that is lossy. That is forgetting, just kind of summarizing what you've been talking about. I can't remember. So for example, let me give you a good use case. You could record every zoom conversation you ever have with anybody, and that transcript can automatically be piped into your open claw environment, and word for word will be stored in a special database that your Chief of Staff sets up. And then you can just say, you know, I think sometime around November, something I was talking about going to ski Tahoe with somebody, who was it? I was talking about Tahoe, you know. And it'll say, Oh, well, that was, that was Jim Steele. That was, you were talking on October 26 and you want, you know, do you want me to email Jim and see if he's up for, you know, ski vacation or whatever? So it'll remember everything, and it makes your life searchable

Andrew Stotz 32:00
and, and how does it, I mean, one of the things that I feel when I look at these types of things is it feels overwhelming sometimes, yeah, and it also feels like the promise, you know, is not, not the reality, right, you know, and how to, how to, how do you deal with them? Okay?

David Siegel 32:20
So open claw is 39 days old. It might be 40 days old. That's incredible, right? So what do you who is this for? Who are we helping people who want to play with it and get started in over weeks and months and years, it's going to get better and better, right? So at the beginning, what can it do? It can manage your calendar. It can help manage your projects. You can make it into a number of project managers that you're working on, and then we can do reverse prompting. So what is reverse prompting? It means that you talk to it again, like for the first week, you're just downloading your you know your lot as much and not maybe stuff you don't want somebody else to get because you never you know. We're trying to keep it so that any losses are inconsequential, yeah, but you'll give it a lot, and you'll say, these are my main goals. These are my big projects. These are my regular responsibilities. I've got you've got your mom, your kids, your various businesses, your podcast. This is what my whole life looks like. And I want you What's the name again? Alpha, yeah, to manage me. So I'm no longer going to keep the to do list, and I'm not going to tell you what to do at all. You're going to tell me what I need to do. This is reverse prompting, so that you in the morning, you'll get a morning briefing with what happened overnight, what's coming up on your day, what has changed? Do you want to make any changes? And the things it wants you to accomplish that day, not just for that day, but also for your other objectives, because there's a big project coming up in two months, or a trip or whatever, and so you've got to keep slicing your time into into deliverables that it's going to get from you, so that that day you do something that you normally might have just forgotten about, but because you did it, you're able to take The next step toward your long term goal, right, right? So that's reverse prompting, and it's really a new way of working with assistance. So think of it as your virtual assistant, who may be in, you know, some other country, but who's standing by 24/7 and has a dozens of personal assistants. They all have PhDs in 1000s of topics, right of areas. And if you, you know, are in medical emergency, they'll, they'll talk you through that, whatever you Yeah, but they can do things. So you'll have a travel specialist. You might have a, well, I think you have. Coffee Company. You'll have a coffee company, you'll have a coffee company chief of staff who manages all of that. You'll be a whole separate thing, and it could be a separate, open claw setup. So because we recommend one chief of staff and one setup for your personal and one for your business, right? Because, again, it's remembering everything, so it's building databases and apps and all this stuff behind the scenes.

Andrew Stotz 35:27
And I've heard people talk about buying some kind of Apple device or something, or Apple computer device. Is that necessary? Is that,

David Siegel 35:35
oh, it's generally a bad idea. This is the MAC studio, or the Mac Mini, and it's, it's really for hackers, so that they, number one, get off of their main computer, so it can't screw anything up. But it has ports. It's connected to the internet, and people can sneak into it, and most of all, so you can't, it's hard to back up. If you're going to back it up, it's just like having time machine, right? You got to run something to back it up to some remote server anyway. So we set everything up on a VPS, and most people are doing that right now. It's a virtual private server that gives you so many megabytes and so much compute power, but it's insulated, and if you blow up, you can't screw anybody else up, but if they blow up, they can't touch your it's all, you know, walled off. It's firewall for you. But we give you two accounts because we're going to back you up. So we have a time machine, backup system, so we'll give you your account to set up everything. One thing people don't know, if you told your executive assistant, or let's call it your chief of staff, to delete everything, including itself. It will no problem. It will delete the whole installation. It'll wipe the disk clean because it has root level access. So we can just give the Unix command to delete all

Andrew Stotz 36:58
how please delete yourself exactly.

David Siegel 37:03
Sorry, I can't do that. Andrew, so, so we have a second account, and we're always piping your data, for example, interestingly, the open claw system keeps a log of everything it does, very accurate, everything it does, all so maybe 1000s of things per day into a log, and that log, we will pump everything from your whole account into another account, one way through a one way pipe, right that it doesn't have access to. So then something goes wrong. You call us, or we give you the keys to that, and you can push a button and say, go back six hours and make it like it was before. And that's much better than good luck. We hope nobody talks it into, you know, giving up everything or getting or deleting itself, right, so that that's something we that's standard for us, and you'd want to do it. You wouldn't need a Mac Mini. Mac Mini is going to cost. So one thing they to answer your question is there, if you get a powerful enough Mac, and it would be something like a MAC studio that would cost around 10 to $12,000 then you could run your own LLM. It would be an open source Chinese LLM called Kimmy, 2.5 you can download it and run your own inference, which means it will handle all the Language and Thought and not too bad, not quite as good as as Claude, sonnet or chat. GPT, 5.4 but close, pretty close. Let's they're sort of like six months behind, and free. It'll run for the price of electricity, but it's going to cost you about 12,000 bucks worth of hardware to be able to run it at a speed that gives you kind of natural language back and forth. If you set that up on a Mac Mini and you did it and you told it to run, it would get a word out about it every four or five seconds, right? It would just be completely impractical. If you have a Mac Mini, or whatever system you use, generally, you want to use GPT pro account, the Pro Plan, which is $200 a month, and you kind of get almost as much as you can eat for that. It's not quite but it's, it's going to be hard to run out of and so in that case, all of your agents are all chat GPT, 5.4 now, Claude doesn't give you that option under its terms of service. You can try it, but they might shut you down because they've said that they don't want that to happen with the Claude, the Claude Max plan. So we're setting people up with the GPT Pro Plan, which

Andrew Stotz 39:44
is $200 a month, right?

David Siegel 39:48
So all of that for us, we also give you complete maintenance program. So we're always making sure you're backed up. We're always making sure you're sick. Or and we're always kind of widening the guardrails and giving it, giving you different little things you can start doing. And then we'll give you an audit trail, so that if you ever want, for any reason, to audit everything it's ever done, to try to figure out what's going on or to show to a third party, we'll be building audit tools as well to make sure. And we can always put people, humans in the loop. For example, if you have QuickBooks, right, using QuickBooks, or you may be using another system, well, we use

Andrew Stotz 40:34
ERP system called ODU, right?

David Siegel 40:37
Okay, so and I don't, and it may have this with QuickBooks, you can get an API that will let our agents run your books but won't and will let put all the accounts payable into a into a queue, and then another person with another login comes in and releases the money, right? So that the agent can't do that, right? That's the kind of thing you want the human in the loop. And there are now several debit cards that the agent can trigger it, but then it's going to send you a text, right? And then you approve by text, and so you can talk to it through voice, because we're voice specialists, most people are not getting voice, but we put voice on from day one. You can do it through text, you can do it through telegram or WhatsApp, and a lot of people are putting it into Slack, so you can see all your different agents for all your projects in Slack, and talk to the sub agents directly. But most of the time you'll be talking to your chief of staff, right?

Andrew Stotz 41:42
It's interesting because I was just listening to a podcast, and it's, forget the actual name in the podcast. Silly me, but the guy has online courses, and he does learn piano in 21 days, and he was just talking about how he's been implementing open call, and then he's opened up, you know, a course, or short, a small group to help people implement it. So, just so fascinating that that was two days ago, and, right, you know, talking to you, what, what? Let me, let me ask you, you know, just because I want to wrap up in a bit. But the one question is, like, what's the tangible benefit that someone would get from this? You know, because a lot of times again, it feels overwhelming. It feels like, I don't know, you know, put all this time in, and you know what? How would you describe, like, a couple, one or two or three of the tangible benefits.

David Siegel 42:41
So you just want to think in terms of use cases. Can you get free of your schedule? Can it do all the scheduling and the rescheduling, the things we use Calendly for, but then they're hard to do the changes and you're in the loop. You can just say, Look, I need to meet with these five people, or we're all going for pizza or whatever for dinner. Just arrange it, you know, and then figure out the common thing and just do it Okay, so it can use smarts to solve problems. One guy showed that he didn't tell openclaw where his anything about his printer. There's no driver. He just said, write a poem and print it. Make it come out of my printer. And it just found it and did it. So a big one would be the back office. I think that most of the accounting and the back office, the expense reporting and all the details are just going to be done by agents, and I predict within one to two years. So one in the easy case and two years in the harder cases, it will simply do your taxes and file them for you done.

Andrew Stotz 43:44
And what is, what is the risk? Let's say of our data. Let's just say that we've got 30 years of data, of client data, as well as financial data in our system. What's the risk there when we give it to it, or give it access to pull together and make a report or whatever.

David Siegel 44:02
So there are a bunch of ecosystems for agents now. Mark Zuckerberg just bought one for $2 billion and that's going to be very advertising based to suck your information out and sell you stuff, right? So in the open claw case, we've been up for 40 some odd days, and it's going to be an evolving ecosystem with all kinds of fixes and patches and options and and apps and guardrails and things so that you'll find a solution that's right for you. You may have very need, very high security, very, very high priority or valuable secrets, or you may, it's just your calendar is no big deal, right? So, so all of these options, what I'm saying is that open cloud is going to win, yeah, GPT, open AI is going to get behind it. And 1000s of things are already coming up every single day, new skills, new services. Is new. People are just published. It's open source. It's going to be like Linux, right? And there's going to be 1000s of people building stuff, scaffolding and protection. And so give you many different choices on how you want to protect yourself and maintain privacy as well.

Andrew Stotz 45:16
And the things that that slow, that I want to accelerate in my life. It's probably the same thing that many people want, but calendar and stuff like that. And it's not such a big priority for me. You know, it's not a troublesome area. The first thing is, I have different products and services, and I have incredible ideas that I'm in stages of in, let's say I have five different revenue streams in my life, and each of them are at different stages. And my vision is, you know, much further than where we are in them. And the vision is clear, but the ability to find the people and to afford the people that can execute it, and execute it in not a way that it's just a waste of time, because when I get back work, when I try to, you know, have something executed, I get back work that just doesn't really help me that much. And then, so that's the first thing that I need to accelerate what I want to accomplish. That's number one. And number two is I want to sell more those. My two things, right?

David Siegel 46:24
So the first thing is, separate your personal from your business, because you want that, yep, you want that firewall. You want to have just two, two. Chief of Staff,

Andrew Stotz 46:33
yeah, my personal, my personal is in good shape. I'm not worried. I don't have and it's slow speed. You know,

David Siegel 46:39
your business, Chief of Staff, alpha will have several businesses under him or her. I don't know what how unisex the name is, but that person, that agent, will compartmentalize each of these businesses. And you want to talk in terms of goals with this not it's not a to do thing. It's not a slave. It's a co conspirator. It's a co it's a collaborator. And you want to say, you know, here are the metrics. First of all, am I even measuring the right thing? You know, what does my daily or weekly report look like? And then what think of it? The way that Elon Musk does his companies, and what they do at Tesla is they just say, what is the biggest bottleneck? Now, let's not write a white paper and design a whole new system. Let's just fix what needs to be fixed, and we'll just do continuous evolutionary improvement every day, and we'll build and one of the things is you want to get more away from the big ERP systems and the and the vendors that have locked you into their style of and that's why Tesla and and Elon's companies are so vertically integrated, because they can make a lot of changes without having to have a right, you know, big meeting with a bunch of vendors.

Andrew Stotz 48:00
Yeah, that turns out to be a real competitive advantage that he's developed.

David Siegel 48:04
Yes, and you too. So you can say, Build me a CRM, and you never even see the CRM. You know how CRM has all these different Boolean operators and fields and stuff. And if you go forward and it says, oh, yeah, sorry, you have to fill out this field before you can leave this page and stuff with an agent. You're just always talking to it and giving it information, and it's always just handling it like, if it needs a new area or a new categorization scheme or something, you could just do that right.

Andrew Stotz 48:36
It's right. So it's so hard for me to visualize that it would get it right.

David Siegel 48:43
It's getting better and better and better. And I encourage everyone to invest in this right now, because it doesn't suck too bad, and it can only get a lot better. We're in, we're in the world right now of self regeneration. It's called recursive self improvement, where the software is written by the next writes the next generation of itself. Right? So we're now entering the singularity. Really, truly. It's happening right now, and it is going to go it's very hard, and this is the big mistake that everybody's making that I promised I would give you right the big mistake that everyone the investment mistake everyone is making right now, is not appreciating the exponential nature of what we're in and what is coming so that this next 12 months will be nothing like Any 12 months that has ever happened in the history, in human history, in the history business, right? Nothing like you. You really shouldn't even be thinking six months ahead at this point. You should be, you should be putting things together, taking baby steps and getting ready to accelerate like you've never seen before. Or because a really smart way to deal with this agent for your, let's say your coffee business, or whatever is, let's just make this thing more profitable. What? Or let's double sales. What should we? How can we do? What are the all the options of doing that? Let's not narrow it down into becoming a tactician. Let's think. Let's brainstorm together about what business results we can achieve in the world, and maybe we can 10x if we start thinking a little differently.

Andrew Stotz 50:32
So what's the best way for people to follow up on this and get involved with what you're offering?

David Siegel 50:40
We've got, we're at open dash, claw.com I'm sorry. We're at claw dash, claw dash studio. Okay. Com, they can see us at red shift labs.io. Get in touch with me. I don't have, I don't have a newsletter yet, but I am putting out video updates and executive briefings for companies, but it's just easy to just get hold of me that way through, through claw dash studio.com there's a lot of resources there, and there is a PDF I just made on getting started with open claw that you and I will figure out how to make that available in the show notes and put it great too. So my, my, my, my summary is, I want people to get going and not say I need to plan this. I need to know what I'm doing. No, you don't. You need to just do your damn calendar, yeah, and then do the birthdays, and then do some expenses, and then you're going to see that as every month unfolds, new vistas of possibilities are going to open up. Whether we do it for you, or you do it your own. You do it with somebody else, stay safe and keep expanding that light cone of what you can do, because this is the year of the expanding light cone of human possibility.

Andrew Stotz 52:03
Well, that's a great way to end it, and I want to thank you for coming on again and sharing what's going on in your life and what you guys are doing. And you know, I think for everybody out there, I'll have you know links in the show notes. Go check it out and join up and get going. All right, good to talk. David, thanks. Edward, yep, you.