You Are the Circumstances

When Seth Godin was last here, in 2023, he said you can’t argue people out of the fear that comes with AI, but you can build something cheaper, more convenient, or more powerful—and watch them move anyway. When the rules change, he said, people need a handbook. His lands today: The Knot: Problems Can Be Solved. Seth is back to talk about his new book, which argues that holding two incompatible wants at the same time, creates entanglements. The Knot helps you distinguish problems you can solve from situations you need to accept, freeing teams to create the conditions for change rather than just hoping for it. Welcome to Invisible Machines. I’m Josh Tyson, joined as always by Robb Wilson—co-founder and CEO of OneReach. Robb and I are co-authors of Age of Invisible Machines, the first bestselling book about agentic AI, and this is the podcast where we explore the future of intelligence through conversations with the people shaping it. Today, as we unpack entanglements with Seth, we also discuss bad demos as bad compression, constraints as a gift, why OpenAI sold you a screwdriver instead of a solution, and why the companies that refuse to choose are Kodak. Bottom line: if problems are bad, AI is bad. If problems are opportunities, this is the change agent you’ve been waiting for. Let’s hear more from Seth Godin.

Seth, it really is great to see you again. we like Rob and I were so excited to talk to you. Actually, we've been kind of talking about talking to you all morning long. And ⁓ one thing that was like really interesting to us is as we look across your body of work, you know, from like like Purple Cow through Song of Significance and now on to The Knot It feels like a lot of what you've been doing is almost like compression. You're looking at the way people make decisions and you're compressing it. And with each book, you get closer to sort of that raw truth where you're just now to the structure of it. And it's really interesting because we realized that with really good compression, like if you compress something properly, the structure remains and you can hydrate it with any data and it still functions. Whereas if you have bad compression, or you're telling a bad story, you have the wrong identity, all those things. Like when you add someone else's water or like different water than you started with, the structure dissolves. does that feel accurate to you in terms of like how we're describing your work

That's such a... first of all, it's very kind, probably more than I deserve, but it's a brilliant insight. know, so much of what passes for insider punditry is simply ⁓ juxtaposing convenient explanations on unexplained phenomenon, right? Like, this company succeeded because X, Y, and Z. Well, yeah, but other companies that have X, and Z didn't succeed. So no, that's not... I was talking to a guy about Da Vinci earlier today. You can come up with all these reasons why Da Vinci was successful. But the biggest reason is he was really lucky. I mean, he showed up the way lots of other people showed up, but they didn't have their paintings stolen from the Louvre in 1912. And, you know, what I'm trying to do is teach people and be strict with myself about first principles, because if it doesn't apply and it's a special case, then I haven't added any value.

Yeah, maybe that's why the not feels particularly useful and relevant and exciting right now because AI adoption is such a thorny tangled thing for most companies to deal with. so having this like really reliable framework, I think is a great service, right? And even starting with this idea that you have to come up with an identity, you have to figure out what you actually are. You know, often that's like coded in branding or some sort of jargon, what you're talking about is something much deeper. And it seems like that sort of is almost job one in terms of how am I going to bring AI into my organization? Well, what even is my organization to begin with?

Right. There are some tricky words here. The trickiest one is purpose. And so I don't start there. I start with what's this work you're doing for and who is it for? And, ⁓ you know, my friend Eric was talking about ⁓ fake mission statements and ⁓ Johnson and Johnson cares so much about their mission statement that they have engraved it in granite 10 feet tall in front of their headquarters that they exist to serve patients. doctors and nurses, employees, and then shareholders. Yeah, except they're the same company that got caught putting asbestos into baby powder. So their real purpose is to make a profit and enrich the executives. But they're afraid to put that in giant letters in front of their building, even though that's what gets rewarded. That's what we're paying attention to. So if we can be clear about what this is even for, how do we know if we're on the right path? What did we sign up for? So many decisions about AI and other things become easy because, we're here to serve these people and we can use this tool. Then of course the answer is yes. If we're here to fully employ artisans, maybe not.

Hmm. There's a phenomenon in AI, particularly you can look at it through LLMs, but it's called catastrophic forgetting. it's this well-known concept that like, if you train a system on code and medical terminology, and then you start optimizing for code, you start losing something else. And you don't know where you're going to lose it, right? And there's a recent study that came out where they input 17,00 languages and they lost like two thirds, if not more of the languages. But what was interesting was which ones were retained and why. And the ones that were retained had more of a core grammatic formula that where all of the words essentially act as a reinforcement to the core. So the languages that had a core, wasn't the order it was ingested. It was the languages that had more of a core grammar were the ones that persisted. And it just reminded me of that.

Yeah, no, it makes perfect sense. you know, pick a trivial example because grammar is so complicated. The word xebec is a great Scrabble word X-E-B-E-C and it's a three-masted schooner. And the reason you two don't know that word is because A, it's not structured like other words in English and B, there are no three-masted schooners. And so because it's the special case in two ways that serves no function other than for Scrabble players, it's going to fade away.

I just wrote that down so I can use it next time I'm playing Scrabble with my wife, so thanks for that. It there's this interesting thing that we're seeing with AI too that I think your book touches on in interesting you mentioned the power in someone within an organization kind of revealing what's possible. But I feel like that gets a little fraught with AI demos, right? Because there's so many, that's like the majority of what's out in the marketplace are these demos that like kind of we're talking about. if you think of a demo as compression, their bad compression because it's a demo that works on a very fixed set of data. But then if you hydrate it with new data, it completely falls apart. it is very powerful to show people what's possible with AI, but if you're showing them the wrong thing, you're actually jamming another pin into the center of the knot,

yeah. Well, so let's let's try for a little bit of historical context. I believe A.I. is the biggest change in our world since electricity. But if we go back to electricity. First, Edison fought against AC power by electrocuting elephants in public. The catastrophic noise around electricity was enormous. Most companies said, we will adopt electricity last. We're going to be, we're not worried about this. We'll just keep doing what we were doing. And washing machines were one of the first home appliances and they were notoriously unbalanced. Also in those days, there were no electrical outlets. to unscrew the light bulb and screw in thing. so washing machines were killing people. because they were marching around the room, wrapping around people's necks, and it was bad. So I bring up all of this because you didn't see that at the demo of the washing machine. The demo of the washing machine was perfect. And it took us decades to get good and safe and productive with electricity. And the stuff we do with electricity now wasn't on the top 100 things that Edison was hoping for. And so we create all of these sedimentary layers. So now AI shows up and people say, well, why isn't it perfect at every job? Well, I'm not surprised at all. I'm surprised it's perfect at any job. But when a CEO looks at it, if she's saying it has to be perfect or I'm going to ignore it, that's a mistake. And if he's saying, ⁓ it's going to do this right now forever, that's wrong, too. It's. this paradigm shift to use the cliche, but it really is a paradigm shift.

you know, in our conversation with Josh, like, kind of thinking of you as like the Einstein of decision making. And, and I'm like, okay, he's coming up with the formula for how we make decisions, right? But what's impressive about him is not just that he could convert into math, something in the real world that hadn't been converted to math, but that he could take a long formula. and compress it to a simple one E = MC squared. So now it's like communicatable, I guess. ⁓ And then I'd give you extra credit for like, and then you'll name it better you'll call it something like the purple cow and his is just the theory of relativity. So I give you extra credit for like coming up with a better name for sure. ⁓ But underlying it. I couldn't help but think how scientists hold on to their theories. Like how much they fight, even if they know that someone disproves it, like how much their identity is caught up into their theory being right. And how much that just reminded me of what you were talking about in your book if, companies are just formulas that we think work yesterday, and now we're saying those formulas may not work tomorrow and your identity is like, but I invented E. equals MC squared, then the kid that came in and said, wait, I found a flaw in your formula because the ground just shifted and physics changed. You're like resisting it because that's who you are.

Yeah. Well, again, back to the purpose. So my understanding is that it took 20 years after scientists proved that ulcers were caused by bacteria for gastroenterologists to stop saying it was caused by pastrami sandwiches. Twenty years. And why would that be true? Because we like being right. We like being sure and we're attached to a certain way of thinking. But if we get back to first principles, the purpose of my interaction with this patient is not to make me feel smart. The purpose is for them to feel better. And surgeons miss this all the time. Lots of doctors miss it all the time because the only person who's in the room with you every time you go to see a patient is you. So you're thinking a lot about how it feels to be you. What does it feel like to admit? that AOL isn't going to be the center of the internet anymore, or Yahoo isn't going to be the center of the internet anymore. I was there when Yahoo stopped being the center of the internet. And the 10 people who ran the company couldn't deal with that. And so they kept acting like it was the center of the universe.

does it take a different type of clarity to cut through? Because I feel like your book does such a good job of talking kind of like how to cut through bureaucracy and find maybe what's really true. And then that that becomes the pin that you can remove. But there's also with AI, especially there's this layer of confusion created by software, Like so much of the systems within an organization, and the culture of the organization are defined. by software and the constraints of software or the limitations of software. And so does that become like a separate task almost? Like there's the clarity of through some of the human operational aspects. And then there's also untangling them or seeing if they're separate or connected to the hobbled software that we've been using for so long that now, you know, can effectively be rewritten, but only if you figure out the bureaucracy first maybe.

Well, it's complicated. And one of reasons it's complicated is because people in organizations, particularly senior leadership, pretend they know how things work. They don't even know how a toaster works. They don't know how a refrigerator works. They certainly don't know how software works. But until recently, it's been deterministic. The toaster works every single time exactly the same way. Software written in COBAL works the same way every single time over and over. But nobody knows how AI works, even the people who built it. And it's inconsistent. Now, what's interesting is six months after it starts doing a particular task, it starts getting really consistent and we now take that for granted. But the frontier is always nutty. And so people who are afraid and people who are seeking the reassurance of the status quo. and people who don't want to be early adopters have no trouble what abouting AI because there's always something on the frontier that's really funky and weird. So for me, we keep coming back to entanglement. And this is where Einstein shows up in my book. ⁓ What Einstein said about quantum mechanics that happened to be incorrect is that spooky action at a distance can't possibly be true. And it turns out that ⁓ tiny particles are entangled. But that's what I want to bring up is that we are entangled. We are entangled with multiple feelings and fears and obligations and goals that we don't want to say out loud. And when change shows up, when decisions have to be made, we freak out because We can't talk about our entanglements, but they are right there. And those entanglements make us stuck. so decision making goes haywire because we make decisions the wrong way. And forward motion becomes challenging and we get trauma instead of resilience because we refuse to deal with the spooky action at a distance with the magic invisible entanglements in our lives.

Yeah, that entanglement thing is so fascinating I was actually just having this conversation with my wife last night. She was excited that I was talking to you also. But she she's come up on this podcast before. She's a she's a hairstylist, but I consider her to be a pretty powerful empath. And she was talking about this kind of cohort of women she sees them all as powerful in that way. And they're all really good at helping one another recognize their own power. But then they all also kind of struggle internally with not feeling like they are powerful enough or not recognizing their own power. And as we were talking, we were just sort of realizing That's sort of what makes it special, right? like self mastery, right? That's almost a contradictory statement because you, you you become a master of your own emotions through connecting with others. It's not something you do alone in a room, there's something about that, that give and take between people when, when there's real vulnerability there, I think that's kind of what you were hinting at, right? Like it's weird to think of an organization as being vulnerable, but that, that is sort of what's required. have to, you have to open up and let something in and it's scary.

Yeah. Yeah. So the indoctrination is real. Misogyny is real, but we're all indoctrinated to be cogs in the system. We're indoctrinated to believe that people who make more money are smarter than people who don't. We're indoctrinated to believe that shareholder value is primary. We're indoctrinated to believe that ⁓ our boss ⁓ is scheming behind us, behind our backs. We better obey them. All of these things create ripples of unspoken entanglement and Because fear is one of the three big levers and the indoctrination keeps pressing the fear button. This is where the phrase, will it be on the test comes up. ⁓ Telling your parents you got a C- These are fear based cycles. And when we are in community. The community has a chance to help us expose these entanglements, expose these fears and reassure each other as we figure out how to make forward motion. So a simple. easy observation is San Francisco is significantly more AI friendly than Topeka, Kansas. The people aren't smarter. It's just they have scaffolding and peer support. It's easier to fit in by being AI friendly in San Francisco. It's hard to fit in doing that in Topeka.

Yeah, we had this conversation recently with Marina Nitze, who wrote a book called Crisis Engineering, or she co-authored a book called Crisis Engineering. And she was telling this story about a call center, this idea of a call center existing. I think it was in the Department of Employment in California, and it was during the pandemic. And everyone was wondering, like, what's going to happen, what are we going to do? There's all these people out of work. And everyone kept saying, well, don't worry, we've got the call center. And so finally she and her team actually like stood up and walked to the call center and found a room with one guy in it who was like really confused as to why they were calling it a call center. He's like, have, I have people who process claims that I can forward calls to, but this is not a call center. that I think touches on something you talk about in the book. I think you call it proximal empathy and this idea that You can't be a tourist, right? If you're really going to solve these problems, and it's interesting, especially in this age of technology, that it's human work. It's like getting up and walking places and touching things and talking to other people. Like that is actually the pathway to opening a door to this technology is through human interaction and finding these entanglements that exist and figuring out how to remove the pin.

Well, you just said something that we should have said earlier, which is the word problem. AI is a problem. AI is a problem because it's a change agent. AI is changing the way your competitors show up. It's changing what your customers expect. It's changing all sorts of things outside of your control. It's a problem and it creates a decision for you. So if problems are bad, AI is bad. If problems are opportunities, then AI is a change agent that you've been waiting for because it is an opportunity to solve problems. And if you can solve problems that create value, you get to do it again. And this posture of problems can be solved and they are opportunities gives organizations the chance to put their fear aside. and go do what they got set up to do in the first place, which is to go solve a problem.

It seems like somewhere in here fits this idea that there are people who benefit from the status quo and there are people who benefit from change and most of the resources are held by the people who benefit from the status quo. Or least it's more centralized. I should say that.

There are certain categories of benefits that are definitely held by the status quo, but the flywheel of technology is not one of them. So the flywheel of technology, which continues to speed up my entire life, is the reason the status quo is back on its heels. And if it weren't for that, we would be living in 1960. But Moore's law times 20 other things means that it doesn't matter what the status quo wants because technology is on the other side.

Yeah, there's some formula here around more resources and fewer hands means the other side has to be better organized because you have to collect more people. But once you do, then they have more power. So that technology has to organize.

Mm-hmm. Yeah, mean, right. And the organization is it's interesting because Timothy Leary's expression, find the others is critical here. The forces of change and particularly progressive thinking have been terrible at traditional organizing. Short attention spans lots of loud voices in the room that don't get in sync bad at following instructions over the long haul. But and If you can create a culture, culture is resilient, culture amplifies. so technology and change have been accompanied for 50 or 60 years by a culture that has a set of rules and standards and expectations. And that's what we're dealing with now is a head on confrontation between the two cultures, the culture of status quo and the culture of change.

I almost imagine that the people in power have this magic wand, you know? Or at least everyone below perceives they have a magic wand. And they say, can you change our culture? And they go, okay, our culture is changing. And then everyone's like, it didn't change, do it again. You're like, do it again. And then at some point people realize it's not magic. maybe the culture changes. because they realized it's going to have to happen at the bottom.

Yeah. Yeah, it's not the bottom. It's the foundation. It's the thing that holds the whole thing up. And it's interesting, you know, if you look at. Yeah, friendly, if you look at organizations that have ⁓ visible.

Or frontline, yeah.

legends, it's easier for them to keep that going for a while. FedEx had it going for a long time until they willfully tore it down. The Costco hot dog is a symbol and a signal. the idea that Howard Schultz embodied in one person a mindset of a certain kind of quality and that when he shut down every Starbucks in the United States, he was sending a signal back to try to get a culture back on track. But even these giant organizations, they veer from that because people like us do things like this is the definition of culture. And there are a lot of people. So people like us is a deadweight or an anchor or an elevator that keeps it going forward, you just got to figure out who the people are and keep that on track.

Yeah, Rob's been having a lot of conversations recently with the CTO of Patagonia. And it's funny because they're like a company that I would say is probably like some mostly averse to AI. but they also are a company that has made Earth their primary stakeholder. And so they're no longer bound by this idea that they have to make profit. And while that might not necessarily change the hierarchy within the company in terms of humans, it... really unbinds them to do a lot of interesting things. before those conversations even began, we had always kind of looked at them as an example of what companies need to look like in the future. And it's funny that there are companies right now racing to adopt AI or implement AI, and they're they're not doing it, or they're failing to do it because they're so bound by this this pressure to to make money, essentially to put profit over everything else, because that That works against the need for systemic change. It fights it.

Yeah. So if we get back to purpose, if Patagonia is not careful, their loss of a founder with a very idiosyncratic vision is going to turn the place into a political mess. Because if you can't find a proxy for the work you're doing, a proxy that can be applied regularly, then it's going to be about personalities. And the good news. about a company that's driven just by its stock prices, the proxy is very clear. So you don't have to have a lot of conversation about what are we here for and what are we measuring. And so the companies that are racing to do partnerships with OpenAI understand that issuing the press release makes the stock price go up. So it matches their purpose. And the challenge, the opportunity, and I'm guessing that many of the people who listen to this don't work. for Fortune 500 companies, we get to make choices one more time about what's the purpose of our work. Why did we get out of bed today? Right? That if you have the time to listen to a podcast, you're not working to pay for the food you're going to eat today. You have some space in your life. What are you going to use it on and why? What is, what is the world going to get out of this? And how does that match what you said your purpose was in doing this work? And I think AI is confrontational because it forces us to think very deeply about what we're up to. you know, in my case, it's read three million words I've to read all my books without a lot of trouble. I could train Claude to write all my blog posts for next year and then I'd be done in like three days. So what is the purpose of my blog? What is the purpose of me going to the office? If I can't get clear about that, then I don't know what to do with AI.

Yeah, and one thing I love about the book is it brings up the fact that there's like the difference between, you know, sort of being able to codify and then the reluctance to codify because you're lying. And codifying means like now you have to be truthful with your company. Because if you want machines to do the work of your company, it needs to know the truth. If you lie to AI, it will do what you said. You lie to a person and they go, I know it's the stock price. You can stop telling me we're here for a bigger purpose. But AI is just going to believe you. It's like, oh, we're here for patients And the next thing you know, you're like, what happened to our stock price?

Yeah, exactly.

Yeah, well, think to your point about individuals having this choice, right? Like we're seeing companies really drag their feet on AI adoption and consumers are absolutely running with it. And I wonder too, like a lot of them they're unpinning their own Gordian knots, right? In their own lives. Like they're figuring out how to use this technology to unbind themselves in all these magnificent ways. And some of it's even occurring within places they work, whether it's through sanctioned LLM use or just private LLM use. I wonder what kind of pressure that applies to organizations as well, because that that is a significant change, right? That's, know, in a sense, that's more agency at the foundational level. But it's also there's opportunity there. Like if you have a proper ecosystem in place, you can actually, if people are creating their own automations, you can you can bring that into your operations. But but that's like a very difficult position to reach, know, especially if you're clinging to old software and old systems and old ideas and old identities.

Yeah, mean, if we're looking at this at a macroeconomics versus microeconomics analogy, if I had a day job, there's no way I wouldn't use AI to save myself time doing my job, right? That if I could figure out how to have a hundred people working in a different time zone making shoes for, you know, whatever those guys were, the elves who made the shoes, and I could just hand it in at the end of the day. Well, of course your employees are going to do that. So that's at the micro level. We're going to see that over and over again. It starts with high school students, but now it's everywhere. But at the macro level, most companies aren't making the right thing. Most companies have the wrong business model. They think they do something for a living and it worked really well 10 years ago, but it doesn't make sense for them to do it now. So how do you take a deep breath? and deal with that. So if you were running a furniture company in the United States 40 years ago, as soon as container ships make it very clear that you can import furniture for a quarter of the price that you can make it yourself, you should reconfigure what you do for a living and say, we don't make furniture, we design and sell furniture. And the companies that did that did great. And the companies that didn't do that struggled and then went away. This is that times a thousand. And the people who aren't making that decision are blaming the status quo and inertia and the system, but they're Kodak. They're going to go away. And they're going to say, under the circumstances, I did my best. And I'm going to say, you are the circumstances. So you're not allowed to say that the same way Kodak deprived a lot of hardworking people of their careers because they couldn't confront the fact. the world was going to change in a way that they weren't comfortable with.

Yeah, and like who says the new world's bad? I think lot of people don't understand like the output of Nvidia is like a data file. Somebody else makes the chip. Nvidia puts out a data file. That's what their whole company does. They create a data file and then they email it. Like the what? trillion dollars in value to make a PDF. somehow somebody decided that if you don't make the furniture yourself and all you do is design it, that that's not valuable. You have to make it. But then they would be missing that Nvidia is making PDFs and doing quite well at it.

Yeah, and I'm in favor of making. think that we need organizations whose purpose is to employ human beings to do work they're proud of. And if that's what you're doing and it's working, I have nothing to contribute. If you are looking at this fork in the road and saying, what should I do now? And you want to simultaneously, this is we get to the but in the end, you want to simultaneously make the stock price go up and make as much money as you can. but also be able to say, do it the old way, ⁓ we make things, we have a factory. I'm not sure you can do both of those things at the same time right now. So you should choose.

I think that kind of applies to a lot of the fear around AI, right? Like that's more of like a situation, like people's fear around AI could be viewed as a situation and don't necessarily cut through all the cords in the knot with a sword. That's counterproductive. But what you can do show them how it can improve their lives, like what they're doing. And that might even run up into the upper echelons of a company, right? If there's... resistance there, but then there's all of sudden this realization if I had this, if this tool is real and I have this, like, imagine what I can do. Imagine what we can become.

Yeah. And why isn't this happening? It's not happening at a corporate level yet because number one, no one's in charge of AI. It's not a single product. Number two, the biggest early movers are tool makers. They're not solution providers. And so when a solution provider shows up, their job is to tell the story of a whole solution, right? Then when you absolutely positively have to get it there overnight, for $12 FedEx will make this envelope appear over there. And they have salespeople who try to persuade an entire company that this changes things for them. Whereas Anthropic and OpenAI and the others have said, here's all these cool screwdrivers and wrenches and saws, figure out what to do with it. And it's going to lead to trauma and disruption because they aren't selling a whole product solution. They're selling features.

Yeah, and I think like AI,

Yeah, I...

have this ability too to sort of dissolve maxims that have been around for a while. Like the triple constraint Better, faster, cheaper pick two. Like that, that doesn't really exist anymore. We talked to Dan Golden, the former chief administrator at NASA and he called it horse pucky. So he he was his term. But like, but that's just becoming more and more true. I guess, are those like invented constraints at this point, useful constraints. And then there's like these sort of made up constraints that are kind of there to keep these stories in place that hold up these flawed structures, I guess.

Yeah, no, I mean, we've been trained to not get in trouble at work. And the easiest way to not get into trouble is to go to a lot of meetings, say as little as possible and do what you're told. And none of those things are revolutionary. And so when a revolution shows up, the easiest thing to do is to ask, what about this and what about that? But mostly to stay on the sidelines and ask for compromises.

Josh and I, like we went back and forth as we were talking about this conversation today, because, you know, we talked about the, say the hydrated version, you know, of decision making and you being a decision scientist. ⁓ And, and then, me being fascinated just by realizing that what you do at a meta layer, like almost if there's a first principle, it's a zero principle, like before first principle, which is your ability to codify things that are in the world. And realizing like now that's a skill every organization needs to have the things you're codifying for us, which is super interesting and read all your books. But then there's like your ability to codify things that I'm even more interested in. How do you do that? And how could you teach that to other people so that they could do that within their organization? Cause that's, that's going to be key to instructing machines to do this is it needs to be codified.

Yeah. And the technique is super simple. you know, if you're physically fit, the technique of juggling three balls is not hard. The hard part is being comfortable, dropping balls for a very long time. If you're okay with that, I can teach you how to juggle in less than five minutes, but you're not okay with that. And that's why it appears to be difficult. So, Noticing and explaining. That's what I do. Notice why something is that people currently aren't explaining that doesn't seem to make sense, that acts in ways that are counter to expectations. So I don't have to explain gravity because nobody is waiting for it. It's like not surprising how gravity works. So notice the things that are changing or surprising and then hazard an explanation. If the explanation doesn't hold up. come up with another one and keep going until you have an explanation that feels universal. So inventing email marketing, anyone could have done that. I didn't do it technically. I just looked at what's going to happen to attention and trust and what will marketers do when they can send direct mail with free stamps, right? And you just work your way down the math. It doesn't take very long. So what I find myself doing is wondering sometimes at personal expense, cause I ask it out loud, why is, why is this your job? And why, why did they do that? So they used to be one of those, ⁓ mid market fake steak houses nearby 20, 30 years ago. And when you went, no matter what, how busy the place was, if you had a reservation, they couldn't seat you. And they said, we're almost ready for you. Please have a seat in the bar. And then 10 or 15 minutes later, your table would be ready. And I stopped going once I figured out what they were doing, which is people went to the bar and they were bored. So they got a drink because the purpose of the restaurant wasn't for you to help you have the best time. The purpose of the restaurant was to help them make the most money. And so making a promise that you can eat at seven o'clock because we said you could eat at seven o'clock wasn't as important as telling a story. about why you had to sit at the bar for 10 minutes. And if you're always noticing these things and trying to come up with an explanation, life gets pretty frustrating because you see how many systems are working against the purposes that we think they have. But it also becomes very useful because you stop being a superstitious pigeon and you start understanding what makes the food come out of the slot.

Yeah, feel like constraints to ⁓ you talk about constraints as a gift. And that that feels extremely true when we're talking about LLMs in particular, right? Like the the hallucination problem is really a constraint problem. And to what Rob was talking about earlier, if you if you have these sort of these core formulas at the at the center of your organization, they can

Correct.

provide a lot of these constraints that LLMs need. But then again, of course, it becomes this problem of being honest about what those core things are gonna be, right? Because like Rob was saying, if you say it's patients above all else, and that's like very much at the center of the knowledge model that your LLM is, referencing on a minute-to-minute basis, then those decisions are gonna reflect that. So... Yeah, again, I guess it's it's more of this identity seeking thing. But but it's it's like taking constraints in the real world that are gifts like you described beautifully how Apple was able to take this constraint of not being able to sell their stuff the way they wanted to in retail outlets to just like making their own retail outlets. And this feels like a mixture of that too, right? There's like the real constraints. that every AI is going to have to butt up against and that every organization faces. But then there's all these internal constraints that you actually almost need to create yourself. You need to figure out what they are. They're very personal constraints to each organization.

Right. And once more, we're entangled and we don't want to say the constraints out loud. So you can make an entire hysterically funny movie highlighting people who refuse to accept constraints. This is what it means to turn your amplifier up to 11. That spinal tap is about nothing but three airheaded people who refuse to accept constraints and have internal inconsistencies that cannot possibly lead to the outcome they're seeking. so one of the things that the purveyors of AI would like you to believe is there are no constraints. It's a miracle, it's free, it will always be free and it will always be a miracle. But the people who are going to use it successfully are going to erect constraints, boundaries about what they're trying to do with it because it's at the edges that we can make decisions. can't think outside the box. It's very dark out there, but at the edge of the box is where you can make a miracle happen. And so in my personal career arc, but in other areas, I try very hard to say, this is a constraint. Like I haven't had meat in 45 years. I have lots of reasons for that, but mostly because I have a rule and now I don't have to waste time thinking every day about, right? That's a constraint. It's a constraint that there's going to be

Mm-hmm. Mm-hmm.

at least one blog post on my blog every day. And so I don't have to have a debate about whether this is the best post I've ever written, whether it's time for me to write another post. It's a constraint. That gives you a foundation to say, this is everything we can do under the circumstances. If you want to change the circumstances, change the constraints. But first we find a constraint, whether it's ours or somebody else's. And we say, if we adopt this constraint, Is it going to help our customers get to where they want to go?

Yeah, I really liked and connected with that idea that the industry, the AI industry is like operating as if everyone has no constraints and then fear, like, and if you don't drop all your constraints and ignore them, you're gonna die. And they're like. They're like, ah, I don't want to die, but something seems wrong about turning the life support off right now and just trusting I'm going to keep breathing. Like, I'm pretty sure this thing, like, I think I need to slowly acclimate to breathing. maybe we need to like ease into this. It's almost like the starting point is to make a list of like draw the box. Let's draw the box. Let's not draw outside the box. Let's draw what the box is. To your point, find the edges and then we'll expand box by pushing the edges out. But one of the things that I want to flip over to the other side is like, okay incremental change makes sense meet them where they are but then there's there's the fear side which is like how much truth is there to like the whole foundation has been disrupted underneath us and and one of the things I've been thinking a lot about is competition in general and Its purpose right It might just be that competition's prime purpose is to motivate people that are generally lazy. And, and that, that when you look at an AI or an LLM, one of the most difficult things to actually build into the system is getting them to shut if there's not an end of prompt, built into the system, it will just produce text forever. Like its natural state is to never stop burning tokens, right? I haven't heard a lot of people ask, except in the industry, like how does it stop? there's a lot of waste in competition. You know, there's a, if, if competition is like to motivate people, there's also five fricking people showing up at my door to deliver boxes today when one could do all the boxes. Right. And, and I'm like, well, why? Because people are lazy and we need competition. So, so if now systems are doing this.

Okay, so you're conflating a whole bunch of things here, Robb. So let me just interrupt you for a second and let's take them bit by bit. I am not the king of the world. You might be, but I don't think you are either. So we're not going to be able to influence or replace the meta king of systems, the system of systems, which is the idea that there is a long arc of capitalism, industrialism and the

No, no, please.

a notion of a semi-free market. And that is fueled by a sort of competition. There are people who have tried to replace it, but you pretty much need to take over an entire nation to do so. And it's really hard. the purpose of competition is, not, doesn't have a purpose. It is a side effect of creating competitions that lead to ⁓ enviable outcomes for some, but we didn't put competition in because we wanted it. We wanted the outcomes. So the second thing you're talking about is there's something inherent in one-on-one chat. And this was put by accident into LLMs. It's been taken for granted, but it didn't have to be that way. That what happens in a one-on-one chat?

agreed.

is if you type, it will type back forever. That is not the way Excel works. Excel says, here's a spreadsheet, it's finished. ChatGPT says, I'm talking to you. So one of the things we built, I think I'm the first person to do this, but maybe I'm not. It's AI together. It's inside of a discourse, which is ⁓ a open source ⁓ community threaded conversation. And now instead of, the chatbot talking to one person, it's inside a discussion board. So now there's eight people having a conversation with one chatbot. This is a game changer because it keeps us from cycling into mania because other human beings are grounding us. But the second thing, which we had to build in, is it doesn't respond instantly after every person types because that killed the whole thing. It's only allowed in some of the rooms once an hour and some of the rooms once a day. It's the only time it's allowed to talk. And it changes the way the entire community engages. Because it's not sucking all the air out of the room like that guy at that cocktail party. It's simply every once in while shows up and says, hey, Robb, did you see what Josh said last week? You guys should have a conversation about this. And then everyone's off to the races again. And I believe that the future of AI hasn't even been inkled at yet, which is everything on the internet is networked, except for AI. AI is solo. It's the first successful thing on the internet ever that was solo. So it's going to be a network thing and the people who are going to build that haven't built it yet. So this is a huge opportunity and I can think of 500 places you could put it tomorrow.

Yeah. Yeah, I think you just sort of, in my view, demonstrated my point. Like what you built to me is a perfect example of optimal stopping. with humans, we're like, how do we get this thing going? And how do we keep it going? And now you're thinking, how do I stop this thing? so I was thinking of companies that are burning tokens today, building software no one's going to use or generating tokens, like, and they're thinking about how do I get this stuff more productive? So they're counting how many tokens everyone's burning and thinking they're winning. And I'm like, no, you should be thinking about how to stop burning tokens. and it's just a different way of thinking. We're so, we're so in that world of like humans need a push.

They're idiots.

And AI needs to stop.

And false proxies are what's showing up here. ⁓ You shouldn't hire a programmer by how many words per minute they can type. Everyone, I think, understands that innately. And yet we look at someone's GPA, even though it has nothing to do with their output. And we look at how many tokens someone's burning because it lets the clueless boss have a super easy way to manage, right? Because they have a ranking. And IBM discovered this was a mistake. in 1960 something with the IBM 360, right? They would start measuring how many lines of code did you commit every day? And that's a really bad way to build a mainframe.

Yeah, it makes me wonder words too. Like they call them, you know, coders, but yet like 80 % of the time they're not coding. They're designing, they're thinking, they're reviewing other people's code. And you're like, Oh, AI can code. What are you going to do with your job? You mean the 20 or 15 % of my job? And then everyone's like, it's only improved 15%. What's wrong?

Yeah, it's the same with writing too, right? Like, I think we talked with Evan Ratliff, who made the Shell Game podcast. yeah, he was like, you know, the actual writing is such a small fraction of my job. Like so much of my time is spent finding stories, talking to sources, synthesizing thoughts, walking around, pacing in a room, like thinking things through. And then the act of sitting down to write it is, yes, eating lots of chocolate. That's important too.

Yeah, that's absolutely brilliant. Yeah.

Yeah, call them writers and you're like, here we go.

Eating chocolate.

⁓ yeah, maybe I'm, maybe I'm a little hooked on empathy here, but I also really liked how you talked about empathy based compromise. Cause I feel like that is really important, right? Like the, sort of compromise of average you describe is, innately tied to that fear, right? we'll just each give enough so that we can keep the status quo running. But when someone's actually committed to changing something, everything changes. And we had this great conversation with Matthew Barzun who was the ambassador to the UK and Sweden. Yeah, yeah, yeah. He wrote that great book. the power of giving away power he brought up the story of Mary Parker Follett, she hated ⁓ compromise. She thought it was like just ridiculous. Right. And so she I think she focused more on integration. And he told the story about her.

I know Ambassador Barzun, cool guy.

building a out in the Vermont woods somewhere. And she had an electrician come out to wire it up because like this was sort of new technology at the time. And his design was like, here's a light bulb in the center of each room, right? So you can hang yourself with your washing machine on accident. ⁓ And she was like, no, that's that's not going to work. This is what I want. And he's like, well, if I build your house like this, it's going to burn down. And so a compromise would have been half a house that's poorly lit and half a house on fire. But instead and he made it sound like she kind of spent her career like digging to the center of this next interaction. But it was like the two of them sitting down and going through a list of constraints and actually creating new value, like using those edges and creating something new. And I think that gets glossed over when we just think about compromise as an easy answer to solving ⁓ a problem when it's really not.

Mm-hmm. Yeah, beautifully said. So the two kinds of compromise I'm talking about, most of the compromise we make are with people who don't want us to succeed in the first place, who don't want change in the first place, who aren't our customers in the first place, who aren't interested in the first place. Right. So the owner of the steakhouse compromises so they can get some vegans to come and they put a mediocre vegan dish on the menu. And then they're surprised that vegans don't come. Well, you compromise to make the argument go away. You didn't compromise because it matched the purpose of the institution. So we should refuse to make compromises with non-constituents. But the other kind of compromise is the compromise that says, ⁓ I lacked the empathy to see the world the way you see the world. And if I can show up in a way that serves you, given that my purpose is to serve you, if I can do that within the constraints that I need, that compromise is a useful compromise. This is the difference between a great architect and a prima donna. A great architect changes their building so that the people who are going to live and work in that building have a better life because they are useful constituents.

Yeah, and that service mindset, I think is crucial to write like, I'm here to help you. I'm here to help you solve your problems. I'm not here to like solve my own problems. Like I can solve my own problems, but that's not that's not as like gratifying. And it's it's not like creating that sort of ⁓ community that we were talking about before, however big you want that community to be, whether it's your company or your or your country, I guess.

Yeah. Yeah.

But I don't need you, yeah.

Yeah. Yeah, I'm very hopeful that having to codify your values and say them out loud will cause us to really think about our values as companies, as people, I think like improve. You know, like you said, now you got to say it out loud. Maybe we up level.

It's so hard. It's so hard, though. I just I don't know if you know what a volvelle is. I call it a spindex now because it's so hard to say the word volvelle and their old kind of star charts and things, pieces of paper with a rivet. And so I made this I made this two sided volvelle just for for kicks and for loyal readers. And it's got each side has 18 or 20 of these questions that no one wants to answer. Right. And.

yeah. Mmm.

just when I hand it to people, it's very clear they don't want to answer them. But when it's important enough, they do. And that's the key. If you think, you if you look at the history of great seafaring captains, they don't get in the boat and accidentally end up at their destination. The purpose is very clear. So if you're going on a long voyage, you better bring oranges because having your sailors die is one of the constraints you're not willing to deal with, right? They're very clear about the purpose. And if a sailor starts whining and says, really want to stop over there because I need to buy a coconut, you say, no, this boat doesn't go over there. It goes over here. We are very clear about where we're going. And we're willing to say it out.

our friend Anna Lembke calls it radical honesty. It's the dopamine reset that really helps

Mm-hmm. Yeah. AI is actually forcing everyone's hand on this, right? You can't really make it work unless you're willing to find your true purpose. Because if you feed it your fake purpose, the purpose that you say but don't mean, it's almost like Kurt Vonnegut We are who we pretend to be, so be careful who you pretend to be. If you're feeding it misinformation or a false representation of who you are, then... that's what's going to project and it's not going to take you where you want to go most likely.

Brilliant, Josh, just brilliant.

Awesome guys. Man. We could just go on and on.

gosh. Thanks Seth. Yeah, yeah, we really, really appreciate you taking the time with us today. This has been.

This was super fun, the two of you. I appreciate the prep, the insight, the care, everything I was hoping for. Thank you for this.

Thank you.

Thanks for joining us for our conversation with Seth Godin. In our next episode, we're going to expand on some of the ideas we shared here with two of the creators of service design, Ben Reason and LaBruns Lovely. If today's episode was useful to you, like and subscribe so you don't miss what's coming next. And thank you to everyone here at UX Magazine who works behind the scenes to make this podcast what it is. We could not do it without your help. Until next time, OAGI