The Exchanges

Q can do RAG, and they can have access to everything in your enterprise, but it's not quite the same as the model actually knowing something as a model, and so the models, unless you train custom models inside your company, cannot actually get smarter at what it is that you do, and You know, the 1000th query is not any smarter than the first. Where do you think that goes?

A I think there are two things there. I think, I mean, I think if I understand his thing, it's all about in-context learning or continual learning, right? I mean, that's sort of the ultimate thing, and it sort of speaks to the thing I was saying, which is, if you kind of have the model's cognitive core separated from its knowledge, then you have essentially the continual learning, you know, formula, so to speak, or the algorithm, and then you just unleash it. There, at least there are three things to me that are outside of the model at runtime that I think you kind of have to crack, right? One is memory, and all forms of memory, right? Short, long, um, right? Even like this, these big challenges of, you know, humans are great at long-term credit assignment, right? Which is how does, um, intuitively, like somebody said to me, hey, the day, you know, you know, AI models can both reward and remember that you, you know, how to punish Uh, for some mistake, because they have the ability to do long-term credit assignment, that's when you'll know they have real memory, but in any case, memory is one. The second one is entitlements, right? So, which is, they have to really respect all of the permissioning system at runtime, right? Because this is where, because there are roles, what access do I have? Uh, and so the model needs to meet that, and then the action space, uh, all has to work. …

AI assessment note: “if you kind of have the model's cognitive core separated from its knowledge”

Q I'm sorry, where do we use methanol in the economy?

A Methanol is, it's used for all kinds of things, um, but, uh, but it's a precursor. So, so again, like, this is somewhat subtle point, but, but Terraform's job is to take, you know, sunlight, which is intermittent, but it's free, um, and in the form of solar panels, make electricity that's intermittent, and then do a bunch of, like, very scary, uh, chemical engineering to it to turn it into a tangible, storable, Offloadable commodity, right? So we make hydrogen, but I would never dream of trying to sell that hydrogen as like our major product, because just there's no mechanisms for transporting that to market. No one wants it in bulk. It's a intermediate product. It's not the final product. I make carbon dioxide. Same story. Put them together. Now we have something that you can store in a tank, right? Methanol itself is chemically like in terms of boiling point and so on. It's very similar to water. So you literally store it in a tank. Then if you want to, you can pipe it across the road to a Standard old-school refinery, and they can turn it into gasoline or something. I could do that myself. I could build a refinery, but, but like, what is Terraform good at? Eh, we don't know much about making large-scale chemical plants, and we'd have to go and license the MTG process anyway. Yes. Methanol is a gasoline process. So why not just hand that off to someone who can do that already…

AI assessment note: “they can turn it into gasoline or something... MTG process anyway. Yes. Methanol to gasoline”

Q a team at Shopify that's doing a great job of their product, and there's a team that's lost and has kind of a woolly direction, and actually their dev tools suck, and so it's very hard for them to make forward progress, and they're thrashing if lots of keep the lights on work. Just how do you distinguish between the team that's struggling and the team that's executing really well?

A Yeah, I think, I think that's, I mean, this is why, um, I, the question in my mouth is, I, I should go to, I don't know. Here's what I found works, is, um, have, uh, rituals by which we talk every, um, I mean, at the latest, every eight weeks, right? Um, and so, let the teams, uh, talk about the progress against the goal. It's like the new, like we have, we have an internal system called, uh, uh, GSD getting shit done. Um, it's a, it's a, it's a, it's a central registering, registry clearinghouse, Um, it's our wiki, it's our, like, feeds and stuff like this, but it also has every project in it. And so it's built around, you know, teams updating everyone else, registering, like, here's something we learned. And so, like, that was step number one. Let's get all, like, get the actual state of a business into a legible, um, internal system so that you can reason about it. Where is every project? What are the deadlines? What has changed? And so this is what these reviews are about. So we go, like, this is what we have on the screen. They, they get a little TL draw that they can drag images and mockups and everything into, but that's their area. But like, all the metrics and all the things are like around the, around the edge on the screen, right in front of everyone. And we can talk about what, um, um, uh, that and how many people are on a team and this kind of stuff. And giving eve…

AI assessment note: “get the actual state of a business into a legible, um, internal system”

Q Would you agree with those three? And do you think people can pick and choose? Because we can take some of those things without maybe, you know, the lawsuit department or something.

A Yeah, so I, I think that's all right. I would maybe add one more thing, or kind of distill it out of a bunch of these, which is basically, like, truth-seeking at all costs. Um, at least I, I find this to be the case with him, and I think this is really not, people, people, people who are mad at him really don't understand this. He really, really genuinely wants to know ground truth, and he really genuinely does not want to know anything that's not ground truth. Um, and, and again, it goes back to our thing of, like, how to confront bad news, like, or bad, like, he, he's absolutely ruthless and relentless in making sure that he actually understands what's going on, and I, You would think that that's common, and like, I have not found that to be common at all among people in business. And, and, or you mentioned another, related to another thing, which is, you mentioned the thing where, you know, we're all, like, literally, with Elon, it's, we're all gonna die, you know, if we don't get this, we're all gonna die. Like, every other, typical startup founder, me, when I was doing it, it's always like, you're always trying to come across, you're trying to, you're trying to, you're trying to, you're trying to, you're trying to, you're trying to, you're trying to, you're trying to, you're trying to, you're trying to, you're trying to, you're trying to, you're trying to, you know, Yeah, …

AI assessment note: “Yeah, so I, I think that's all right. I would maybe add one more thing”

Q to, like, step one of founding an internet company was, you know, write a check to Oracle, and then you can do stuff after that, and then the open source databases, um, MySQL and Postgres became competitive in the 2000. Like, you don't like me reasoning by analogy too much here, but like, can you reason by analogy to the database world? Just how does the market structure play out?

A Yeah, no, I think that's right. I think that's right. I think that's a good, that's actually a good comp. Another one is operating systems. So when I, when I, when I was a kid, um, you know, the world's best operating systems were, um, specifically, I mean, Windows is its own trajectory and iOS, but, um, for, uh, for like, what we used to describe as proper computing on real computers, um, like, uh, Unix computers, uh, uh, including supercomputers and workstations and advanced, you know, scientific applications, things like that. Um, you know, the best, the best versions of Unix were proprietary for a very long time. You know, these really big companies like DEC and HP and others that have, yeah, IBM,

AI assessment note: “I think that's right. I think that's a good, that's actually a good comp.”

Q Yeah. We're talking about European air conditioning. You've talked about the EU accelerationism movement. Um, what, uh, what is required for EU accelerationism?

A So, I think it was two years ago, one year ago, um, I saw Beth Jaisels, who's, uh, Guillaume Ferdin, he made eOC, which is like effective accelerationism, which is like, let's accelerate America again. Like it's been a very depressing time, especially with COVID. Let's look at the benefits of technology and AI for the future. And I thought, that's amazing. We need to do the same for Europe. So I did EU slash OC, like, let's do this for Europe because Europe is, it's the same kind of like depressing kind of down vibe about the future. I understand because there's a lot of problems. It's not an easy life for a lot of people. So that got a lot of traction, and, um, I crowdsourced this, like, bug board where everybody could submit their idea how to save Europe, how to fix Europe, and, like, Tens of thousands of people submitted ideas and then they upload each other's ideas and stuff. I think number one is just like remove the regulatory burden for new businesses. Yep. Um, and there's other stuff like remove the cookie banner and stuff. And I hope it will be easier in Europe to run, to start a business, run a business.

AI assessment note: “number one is just like remove the regulatory burden for new businesses.”

Q If you were running Stripe, how would you fix?

A Um, I think it's going really well recently. I think it wasn't going well after COVID. There was a few years where things were getting, the API was getting more and more complicated. We all complained about it. There was every, it felt like it was being run by engineers, not by people who actually use Stripe. Like illogical defaults on the API, you know, like one time I, um, I had this bug and it was from the Stripe API because it changed and it didn't create a customer anymore. Everybody was a guest. I was like, why? Would I want people to be a guest? Like I can't, I cannot even curate them as a customer anymore. Those kind of things. But now it's, now it seems much better, and it seems to be like sanity has returned. But it's scary, and it must be scary for you, because you're running a big company. It's very hard to, to manage a company from small to big, and not, become like corporate, you know. Of course you're always corporate, because you need to do B to B sales, becomes more corporate. But you need to keep this, The original Stripe mentality was simple, right? Stripe.js. It was one line of code you would add to your page and you had a payment button, you know? Keeping that is so important, keeping the original soul of the company, and I think you're doing a great job. Because essentially, I don't want to go into Stripe and do anything. I want to go there to see how much…

AI assessment note: “The original Stripe mentality was simple, right? Stripe.js... Keeping that is so important”

Q Okay, but are there interesting differences beyond, like, correlates like size?

A What I can say is, like, slightly different to your question. The people interacting through voice and the performance we see for, like, how they interact with, with, with, with the business changes just by nature of interacting with voice. A good example, you can contact 11 Labs and register for For your interest, you go through the form, and at the end of that, ah, we have supplemented that, that instead of going through the form process, you can speak with our agent and leave more details. And what happened are two things. One, people were actually much more keen to leave the forms through speaking with the agent, so we would go through the form a lot, a lot easier. But second, they would be a lot more open-ended in terms of what the use case are. So they would start giving us information about the wider set of use cases, the complexity of the use case. So like the writing out was tedious and, and, and, and tricky.

AI assessment note: “What I can say is, like, slightly different to your question.”

Q You guys were long wheels, right? Yes. Right? Yeah. So, that's a good example of like, how do you, because every management team says- Yeah, right? Every management team says we're going to turn this thing around, and, uh, every management team has a projection that, like, looks good, and so, how did you determine that now, finally, they're going to turn it around?

A Okay, so let me start by saying, like, The U.S. and Europe are very different in this respect. I find that, um, if we, let's take a U.S. company that had like a turnaround, the industrial company, like, three M. Three M has been a horrible stock for a very long time. Wasn't well managed, um, New CEO comes in, puts up one or two good quarters, everybody, we owned it, everybody basically understands what's happening, and the stock kind of goes to fair value with the assumption that the margins are going to go to where they should go. Like, the U.S. is pretty quick at, like, seeing change happening, and then pricing in that change. In Europe, I find, um, you know, once a company, like a company like Rolls had underperformed for so long, I guess that European mutual funds, they just kind of got in their head that like,

AI assessment note: “The U.S. and Europe are very different in this respect.”

Q Exactly, all these kind of roles. What do you think Stripe, Silicon Valley companies should learn from Eli Lilly and, you know, companies like Eli Lilly, but that's where you have experience about talent planning and talent development?

A Yeah, fabulous question. I do notice differences At least on the board I'm on and observing other companies, um, some of that might be just the clock speed of the industry and the technology, and some of it might be the newness of companies. Because if you haven't really seen the cycles play out, It's hard to kind of see the value. You got a problem, solve it, you know, work, work on the next thing, but probably even in your company, which has been around long enough now, I'm sure you have people who are, you know, single digit hires, you know, first few people who've really been excellent. And what they're doing now is nothing like what they started doing. And you should examine, like, what are those? Why was it the experience path they took? Was it innate traits they have combination of those things? You know, we've had a 150 years of that, and I'm the 11th CEO of the company. That's, that's one less than Pope's in that period of time. So that's, it was a special honor, actually, and it's not a lifetime appointment. Um, I can be fired any day, but the first four were family members, and then we've had a lot of long running successful and only one external.

AI assessment note: “You should examine, like, what are those? Why was it the experience path they took?”

Q And is it about also kind of finding constructive forms of it that lead to better performance?

A Um, yes, it's a balance. Um, one of the mistakes I've seen a lot of first-time founders make is, is they assume that an hour of coaching is as good as an hour of firing. Because at the end of the day, you can't be some other founder. You can only be yourself. And so you want to build an organization that like essentially is like an extended mind for you and for your hands and arms and the ways you think about a problem, right? And I think Stripe is very much a reflection of you and Patrick. I'm not sure which one the most actually. Um, I have a question for you about that next up. Um, uh, and so to an extent, the parts of Terraform that weren't working with the parts that I was trying to build, not my way against my better judgment. And when I just gave up on that and said, screw it, I'll just do it my way. It suddenly started working because I could understand and intuit how it worked. Yeah, had to become pretty demanding there, and obviously, um, no one's perfect at recruiting, and so the success to that is you've got to give people extremely direct feedback and coaching and mentoring, and ultimately Invite them to build their career elsewhere. Um, it's been really tough. Some of the people I had to let go, people I really like, people I had friends with for many years in some cases. Um, and, uh, and obviously I wish them the best, but again, having read resume after resume a…

AI assessment note: “Um, yes, it's a balance. Um, one of the mistakes I've seen”

Q So it's just literally suburbs is like a great place first?

A Yeah, and then I think certain city centers, you know, depending on the actual area though, right? And so, there's, um, there's just a lot of, uh, those are the two types that we're thinking about, you know, distance, And the types of, you know, one of the, um, coolest things that we're actually building in DOT that's actually, you know, not physical is What we're calling our autonomous, you know, development platform, which is basically, um, uh, you know, the algorithm that talks about which orders goes to which type of vehicles, right? So DOT, for example, obviously, um, can also, uh, travel mixed routes with, you know, humans, right? For example, if it's a longer distance delivery, but, um, but there is a part of that journey where Where DOT is more advantage, say parking.

AI assessment note: “Yeah, and then I think certain city centers, you know, depending on the actual”

Q That's right. How do you measure very online?

A Oh, let me give you an example. Um, I was in a town, a major city not too long ago, and I was asking the chief to tell me kind of what's going on. This is a couple of years ago, right? Like during, during COVID. And she was like, Garrett, These kids are just killing each other. I was like, what? She's like, yeah, they're, they're literally getting in their car and just shooting each other. I'm like, why? I'm like, oh, because this guy posted a picture of him with that guy's girl on Instagram. And I think in a normal situation, you might have called that person and been like, hey, bro, that's my girl. And in other cases, they're getting in a car shooting someone. Like, that's not normal. That's not normal behavior, and it wasn't normal before COVID, and it happened a lot during COVID, and then it's largely gone away. But it was a very specific social phenomena where the, the race to violence was so dramatic. It was so scary.

AI assessment note: “let me give you an example... this guy posted a picture of him with that guy's girl”

Q Yeah, yeah, well, just, I found it interesting how it kind of explains the path dependence of how each country gets to where it is. Speaking of Canada, what's your advice for Mark Carney?

A Yeah, I mean, it's like, Funny, our partnership is, uh, very, very, uh, storied and layered, and somehow, uh, your board members know my prime minister. So, uh, that, that's fun. Look, I love Canada. I live now more in Canada than Germany, uh, for long, for longer. I think it's a, it's a beautiful project. It's a beautiful implementation of different implementation from the United States or from Europe of classical liberalism. Um, it's very much veered into, um, uh, socialism under, under Trudeau. He pushed that very, very hard, um, um, I am very much hoping that, ah, Mark is helping all of Canada to figure out that that's a successful discovery of something that doesn't work, because obviously that kind of push socialism always leads to poverty.

AI assessment note: “hoping that, ah, Mark is helping all of Canada to figure out”

Q Who should try the digital nomad thing? Who does it appeal to?

A I think anyone. It doesn't matter if you're 20 or 70 or whatever. It's so interesting, because you, you change as a person also, but I think digital nomad is so cool because, um, Leaving your own country, which I did the first time, I, I, I did a study exchange program. I went to Korea in like, 2000, uh, 10, I think, and from university, and I was like, this is a completely different world. I never left Europe before. I went to France, you know, like Italy, but like, This is like, there's neon signs, and it's like, it's crazy. This changes you on such a fundamental level, going to the other side of the world. It's amazing, and it's, it's way safer than you think. Like, usually nothing happens, and I've never been robbed except in my hometown. It was a burglary, so in 10 years of this, nothing ever happens. So yeah, I think it, it completely expands your horizon. There's nothing I can recommend as much as people just go travel, and don't be scared, man. Just like, there's always people who help you. If you don't speak the language, you can use your arms and like, how did you make it not be lonely?

AI assessment note: “I think anyone. It doesn't matter if you're 20 or 70”

Q Um, marketplaces slash app stores have been a thing. What comes next? You know, what happens to e-commerce in an agentic marketplace or an agentic commerce? I think these are the interesting things that need to be litigated.

A Well, I want to talk about that and I want to talk about commerce, but actually first, while we're on the topic of, while we're still in the nineties, um, everyone is making comparisons to the dot com bubble right now. It's almost a cliche. And I think it's actually a reasonable comparison, you know, it's cliche for a reason, which is, it is a very CapEx intensive build out for a new paradigm that is in fact a big deal, and yet there's an awful lot of CapEx. You were there, uh, at Microsoft, uh, during the, uh, the thousand dot com bubble, and it really was, uh, you know, Microsoft's share price peaked in the late nineties, early 2000, and then didn't surpass it until 2016, I wanna say. What did it feel like in 1999? In particular, did you know you were in a bubble, or was it like, oh, this is the new, this time it's different?

AI assessment note: “I want to talk about that and I want to talk about commerce, but actually first”

Q a lot of activities in companies that are kind of the same, uh, regardless of the, the, um, the size. That said, there's also probably things that only show up at the, uh, you know, 200,000 person, um, uh, you know, city, state, um, uh, uh, size that, you know, I wouldn't be aware of at the 10,000, uh, person size. What effects only show up when you're that big?

A There are two things I would say, uh, quite honestly, having only worked at Microsoft, it's not that I'm like an expert, but the one thing I would say, taking over for a founder, Steve and Bill built a company. I mean, Paul and Bill started, and Steve and Bill scaled it, and I was sort of the first quote, unquote, non-founder person. The thing I realized quickly, or in fact, I got into the job and I realized that I need a team. And just to have the ability to manage the scope. But then, You know, you know, that AG Lafley thing that we put out there, which I think is a great one. Like, what is this, this, what, and being clear about what does the CEO clearly need to do though in that, right? Which businesses are you in? Which businesses are you out? Synthesizing the outside, having the standards, setting the standards for culture, and then the ability to your point about having that performance culture that you can't say, hey, I'm only about the long term or I'm all over the short term. You've got to deliver both.

AI assessment note: “I realized that I need a team. And just to have the ability to manage the scope.”

Q significant health benefit, and then it leads to them being prescribed more by doctors, because people actually ask their doctors, they do, in fact, do the thing they ask their doctor about them, and it leads to more usage. But yes, people just find the whole thing weird, you know, private for-profit healthcare. And so, do you have a view on where this goes, what we can do about it?

A Yeah. Well, I, not having grown up in a system like Ireland, but I lived in Canada for a while, six years, so I was treated in that. I've seen it. Yeah. I find it weird the other way because that's your conditioning. I think there's good and bad to both. I think actually we were talking about prevention earlier, to some degree primary care. What I experienced in that country was pretty good quality care, very standardized. Which has a confidence boosting thing when it's the same for everybody. That also in specialty care is implemented, but actually to kind of a negative result because take like, um, diagnostics. The US and China have something like 70% of all diagnostic capacity in the world. That's crazy.

AI assessment note: “I think there's good and bad to both.”

Q Like, what's the most ridiculous internally developed piece of software?

A You have like a, a static sort of, ah, internal geosities, like it's like static site hosting, which is like, like, it's the most incredible and, like, um, geosities, yeah, like a, like a nineties web, um, um, with all the pros and cons. Yeah, like it's, it's, I mean, it send us down wrong paths. Like it's, it's very often, I say internal tools culture. This is actually a little bit of a dangerous statement. I shouldn't make it like as such. We have a, Um, very strong, uh, appreciation of good software and, and good in the context of we have to use it. Like some software is really good. Like if you use, like Shopify is really, really, really good software. If you're, um, running like a D to C business or products, some people, uh, try to stretch it to, um, subscription-based grocery stores, something. I, I, maybe that even works, but I clearly didn't build a different mind. And I actually, Although I appreciate all my customers, I'd rather not anyone use the software if it's bad for them. Because that's like, I, I, I built Shopify's to avoid people having to use shitty software, not to add to the part, right? So, so sometimes this means, given our scale and our, our minimum quality bar, uh, that you have to build it ourselves, and in those cases we do that. Uh, sometimes we build it in where we have no business doing it, and that's where things go sometimes wrong.

AI assessment note: “a static sort of, ah, internal geosities, like it's like static site hosting”

Q We had Mackenzie from, uh, Ambrook, uh, on here, which is, um, SaaS for, um, financial infrastructure for farms, but she was saying one of the big trends in agriculture right now is direct-to-consumer stuff, where, you know, you're actually transacting directly with your farmer, and I'm sure a lot of them are running on, on Shopify as well. What have you learned on the Coinbase board?

A I never had another job, really, in my apprenticeship, and, um, um, mostly learned what not to do, uh, as a company, and then, um, sort of had the opportunity to start Shopify, just trying literally the opposite, and that's basically the operations I've been running on. It's been amazingly successful, so thank you Siemens. But, like, it's, it's, it's, it's, it's so hard, especially building it in Canada, like, uh, in Ottawa and Toronto. It's funny, it always had this effect of, it came to value a lot, um, when I, when I had time. And, um, money. You know, I had all these meetings, and then talked with people, and, um, they described to me how real companies do these things, and then I went home and tried to implement it, and only years later did I realize, um, very few other companies describing what they're doing, um, actually did the things that they told me. Everyone gave me sort of a idealized highlight reel, and I compared that to my average, and then I implemented how I think you could do a better job on the stuff that they described, and sometimes ended up with Uh, good things, and I was very, very happy with this. And then, um, now, obviously I'm describing when this goes really right, and then just as often this went really wrong, right? Like, so I just, like, Shopify ended up being this, like, sort of lopsided thing that, like, is, like, um, probably ludicrously over-…

AI assessment note: “I never had another job, really, in my apprenticeship”