Sign in Subscribe

No More Hustleporn: Bret Taylor on the origins of Google Maps

By Yiren Lu —

We pulled out the highlights from Shopify founder Tobi Lutke's interview with Bret Taylor. Transcription and light editing by Anthropic's Claude, curation by Yiren Lu :-)

Highlights

Bret Taylor: I do think as companies grow implicitly, organizational structures punish generalists. It's the idea that you start off with an engineering team, and when you were selling snowboards, you made the whole stack right from top to bottom. And then over time, you have an infrastructure team, and you have maybe a front-end team. And the engineer that masqueraded as designer, it's no longer socially acceptable to do the design anymore. That's someone else's job and you end up kind of going into swim lanes.
It makes sense because you're creating a system that scales and so when you do that you tend to think of all these functions as sort of literally that functions that take inputs and produce outputs and to scale things, you tend to have specialties. And I think it's for people who either are generalists or would naturally be generalists, the most natural thing in those environments is to specialize because that's how you further your career. But I do think if I think of all those great products, I'll just take Google Maps, one of the ones I worked on. It was a really good user experience. I think at the time, dynamically rendering images was really expensive, so most maps looked really ugly. But we really wanted them to look like A to Z maps from the UK with road names written inside the roads, even if it was curved. We wanted really good antialiasing and we worked all this technical infrastructure make that work. We wanted the maps to be draggable. We were crashing the browser left and right because no one had really done anything interactive. And it was this really vertically integrated set of innovation. Like you had to have a vision for the user experience. You had to understand pretty deep technical things both in the browser and even at sort of the kind of networking layer to make it work with sort of the low bandwidth connections that were typical back then. And I just don't think you would have had it if everyone was in their swim lane and didn't have the whole vision.
So I think it's a really big challenge for innovation in larger organizations is just how you scale and how you do that through specialization, but you don't extinguish the natural kind of vision and innovation that comes from people who can put on many hats.
Bret Taylor:  There's a paper by a professor named Coase called "The Theory of the Firm." It was basically why do companies exist? In theory, you shouldn't have large conglomerates, let alone companies that just grow in arbitrary size and complexity. Again, my Cliffs Notes version of this is the reason they exist is transaction costs. If you and I want to do something together, we probably need a contract to do it. Whereas inside of a company, things like allocating new servers or even having a designer work on something or producing a new presentation for a customer, there's no transaction costs if you set up a firm correctly.
So in theory, if firms can scale output and reduce transaction costs, which give them an advantage over just a free market equivalent of that, in practice, I think companies often don't think about transaction costs. They introduce themselves. The classic thing is that every time something goes wrong, you make a process so that never happens again. And you fast forward ten years, and all of a sudden you have cascading layers of processes, and unless you're an anthropologist of Shopify, you have no idea why they exist anymore.
And I think that when you think about building an organization that produces innovation, I think it requires you've always been very intentional about how you grow Shopify. But that intentionality where you ask why to all the whys like my youngest - it can be annoying, but I think it's also without it, the only safeguard against the natural kind of friction companies introduce.
Bret Taylor: I had a really impactful sort of early part of my career where I was reminded of not just the power of entrepreneurialism, but actually how much, especially in the technology industry, we're not entitled to our future success. So when I first started at Google, it was a couple hundred employees, give or take, maybe a little more, but a small company, one building in Mountain View. And when we grew up enough that we moved to a few neighboring buildings that eventually needed a corporate campus. And rather than building one, we bought Silicon Graphics campus. It was right next door and Silicon Graphics was going out of business. It was really interesting because when you move into another company's corporate campus, when you first move in, it sort of has all of their branding and paraphernalia still there. So you're walking through sort of the carcass of a company that was once great, that was once great enough to have that campus that you just moved into.
And then later in my career, when I joined Facebook, they acquired a social network I made. We were in one building and eventually grew up, and we needed a campus, and instead of building one, we moved into Sun Microsystems old campus. And it was just really interesting because both Silicon Graphics and Sun Microsystems at one point were like the top of the stock market, the shining star of the technology industry. And that was like when I was graduated from university, not that much earlier than these moments. And it had a real impact on me that you're sort of walking through the halls of a company that built this amazing campus in the heart of Silicon Valley, and both of them ended up selling it for pennies on the dollar to the next company that was sort of graduating there. And it really, to me, it just had this impact, which is like, no matter what success we've had for the previous decade, we are not entitled to success in the future decade. And just the half life of technology companies is actually quite short. There's really relatively few exceptions, I guess, IBM and Know that have been around for a reasonably long period of time, IBM in particular, but there's not that many. And I think it's always good to remind ourselves of that because that sense of urgency that all of us have in the early days of our company, if you end up losing that, it's almost a near certainty that you'll suffer the same fate, which is kind of crazy. But also I find it really energizing because I love the kind of creative destruction and I love the sense of urgency that that drives. But it had a huge impact on me.
Bret Taylor: Yeah, it was interesting ingredients that led to it. So the story of Google Maps was at Google, I was a product manager for a product called Google Local, which essentially integrated yellow pages listings into Google Search so you could find local businesses in addition to web pages. It was an okay product. It kind of had a me too. There was a product called Yahoo Yellow Pages. It was somewhat ubiquitous at the time, and I wouldn't say the product was particularly differentiated.
So we launched the first version of it, and we spent a lot of time just talking about how do we produce something way more powerful than the competitors? Because we kind of made a first version that looked a lot like a lot of our competitors and really thinking like, what are directions we can take this that are significantly more powerful?
We started anchoring on the map we had licensed from a company called MapQuest, which you have to be our age to even remember MapQuest, but we put a little MapQuest tile next to the search results. It was really ugly. And we you know, just imagine if you really anchored the whole experience on mapping.
By coincidence, there was these two Danish engineers named Lars and Jens Rasmussen who were building a Windows app called Expedition that was a mapping application. Even in 2003 or four, whenever this was, no one was funding a company building a Windows apps. They had trouble getting anyone to return their calls. They were referred to us by one of our investors, I think, and we ended up meeting them and sort of doing an acquire of that company.
And it was actually turned out to be the biggest blessing in the world that what they came in with was a Windows app because it could do all of these things that you couldn't do in a web browser. So the reason we acquired of them is we love the look of the maps. Large and Yens just were mapping obsession. Like, they were just obsessed with maps. But all of a sudden, the watermark of what we wanted to achieve in the browser was so much higher of a standard than I think we would have set naturally.
So if it weren't for the fact that Jens felt so strongly about doing this in Windows, I can't say with a straight face that we would have had the ambition to go as far as we did in stretching the browser as far as we did. And then there was a lot of other ingredients at Google, Gmail was in development or had come out kind of around then. And there's a lot of people in the very early days of pushing web pages to become web applications, to make them truly interactive, to make them so you didn't need to refresh the page to interact with them.

Full Transcript

Tobi Lutke: Welcome everyone. This is like a super fun conversation. I've been looking forward to this for a long time. I want to introduce you to Bret Taylor, who I have the pleasure of welcoming to the Shopify board of directors. So that makes you basically my boss. I hope this is going to go well. Welcome, Bret.

Bret Taylor: I'm really excited to join the Shopify board. We've known each other for a long time.  You joke about the technical topics, but I think the first time we met each other,  I don't want to put words in your mouth, I think we both surprised each other with how geeky both of us were. And anyway, I've always loved working with you and to have the opportunity to work with Shopify is amazing.

Tobi Lutke: So really excited. I mean, you and I are definitely technologists by background. I think that became pretty clear. But in the context, I think where we met, this was one of those places where people actually unironically wear suits at these events. I have to say, I have to confess, I always feel like free squirrels in the trench.

Tobi Lutke: So when you have that conversation and it goes to the next level, next layer, the next layer, next layer. That is delightful and unexpected. And I think we instantly bonded over that. I don't know what we talked about, but might have been any and I.

Bret Taylor: Also thought because it was so unexpected, it made it all the better. So I loved it.

Tobi Lutke: I guess where I would love to start is like you're in Silicon Valley. You didn't grow up there. But you have been at the most exciting times of these companies' histories with Google, with Facebook. With Salesforce in one of the most exciting times imaginable in not sure in a good way at Twitter and of like co CEO of Salesforce, but also like a series of fantastic startups. It's just like how did you line that up? How did you set up that story? Can you sort of connect the dots a little bit and take us through this?

Bret Taylor: Sure. I'll give you the Cliff Notes of the history and I'll describe I'm not sure it was quite so intentional, so I'll connect the dots. But there's some maybe life philosophy that has guided it. So I was a computer scientist at Stanford, almost went into the PhD program, left with a master's degree at the time, it was right after the .com bubble had burst and so academia felt slightly appealing. But a teaching assistant of mine had gone to work for Google named Marissa Mayer, who ended up running sort of the consumer product group there, but also became the CEO of Yahoo eventually. But she could not stop raving about how special this company Google was.

And I ended up going there not as an engineer, but as a product manager. The amusing part and no joke, I interviewed for the role. And even by the end of the interview process, I had no idea what a product manager was because everyone you asked gave a different answer. I joked with someone recently that you ask a product manager, they perceive themselves of in charge of anything you ask an engineer and they perceive perhaps the opposite. And it was just really funny. But I ended up taking a bit of a leap of faith. It was the only role that I was interviewing for that wasn't an engineering role and just followed Marissa, actually, just because I admired her and I loved the product that was Google. I loved the space that it was in.

I don't know if this is technically Sheryl Sandberg's quote, but I think of her associated with it, which is if someone offers you a seat on a rocket ship, don't ask what seat? And I sort of subconsciously followed that advice at that moment, just thinking, I love Google, the company. I think Marissa is someone I looked up to. I'm just going to go and hopefully figure out what this product management thing actually is.

I followed similar I don't want to say it definitely wasn't impulsive through my career, but I'm very open-minded. I actually came to salesforce by the acquisition of my last startup, Quip, and I'm friends with a lot of entrepreneurs, as you are, and Shopify is a company that serves entrepreneurs, so being an entrepreneur is a big part of my identity. And I think everyone who knew me just presumed I would be there for a couple of years and leave, because that's what entrepreneurs do. But I was more open-minded than I think a lot of people thought I was. And I grew a lot of affection for the company and the mission and in a very different way, it serves an ecosystem that's really broad in ways I grew to really deeply admire and I ended up staying for almost seven years and becoming co-CEO.

I think if there's something that characterizes, it is. I very comfortable wearing different hats and also very open-minded to experiences and I am really driven by impact and I think impact can take different shapes at different times in my life, but I'm just a big believer in not having rigid thinking and seeking out truth. And the thing that I've really worked at is having not being so rigid in my own personal identity because I think that can be really limiting. If you have a conception of yourself, that is one thing I think you implicitly close yourself off to new experiences, new ideas, and I've kind of gone between technical role and non technical roles. I've unironically worn suits at times over the past seven years and so it's been great. There's a bunch of stuff in between I'm sure we could talk about, but that's kind of the Cliff Notes version. I think I'm pretty open minded to who I am and to new opportunities.

Tobi Lutke: But like it's actually hard to track down the people who are up for this kind of thing and it goes against a little bit against society's narrative, doesn't it? Right. I certainly have seen this in a sort of group maybe clustered a little bit around my age. I wonder if there's something like in sort of like early 80s, end of 70s time period that caused people to see so much change during in the computing and so on that maybe everyone kind of did everything and really proved out this sort of full stack kind of people call it. I don't have a good mental framework for what I'm talking about here, but I'm just really deeply curious about what makes some people choose to interview for literally the opposite of what we just kind of study for and then some people just trying to sort of join the see your career as sort of a game that other people define, and then you join and you try to sort of excel in it. Do you have any thoughts about that?

Bret Taylor: I you know, it's interesting. There's a few friends I have that are study innovation. Like Patrick Collison has researched this a lot. And one of the I'm not an expert in it, but one of the things that I've heard and it might be an anecdote, but I think there's some research behind it is that research is so increasingly specialized now. That to publish papers and to get tenure, you end up so deep narrowly that it really closes the door to a lot of scientific breakthroughs that could happen through sort of the cross pollination of different disciplines that used to be more commonplace.

I do think as companies grow implicitly, organizational structures punish generalists. It's the idea that you start off with an engineering team, and when you were selling snowboards, you made the whole stack right from top to bottom. And then over time, you have an infrastructure team, and you have maybe a front-end team. And the engineer that masqueraded as designer, it's no longer socially acceptable to do the design anymore. That's someone else's job and you end up kind of going into swim lanes even same is true on go-to-market teams and other things.
It makes sense because you're creating a system that scales and so when you do that you tend to think of all these functions as sort of literally that functions that take inputs and produce outputs and to scale things, you tend to have specialties. And I think it's for people who either are generalists or would naturally be generalists, the most natural thing in those environments is to specialize because that's how you further your career. But I do think if I think of all those great products, I'll just take Google Maps, one of the ones I worked on. It was a really good user experience. I think at the time, dynamically rendering images was really expensive, so most maps looked really ugly. But we really wanted them to look like A to Z maps from the UK with road names written inside the roads, even if it was curved. We wanted really good antialiasing and we worked all this technical infrastructure make that work. We wanted the maps to be draggable. We were crashing the browser left and right because no one had really done anything interactive. And it was this really vertically integrated set of innovation. Like you had to have a vision for the user experience. You had to understand pretty deep technical things both in the browser and even at sort of the kind of networking layer to make it work with sort of the low bandwidth connections that were typical back then. And I just don't think you would have had it if everyone was in their swim lane and didn't have the whole vision.
So I think it's a really big challenge for innovation in larger organizations is just how you scale and how you do that through specialization, but you don't extinguish the natural kind of vision and innovation that comes from people who can put on many hats. And I think that's going to require for a lot of innovation. And it's just interesting you see this in society as well. That's why when I read that sort of Cliff Notes version of that paper on innovation and specialization, it resonated with me because I saw whispers of that or echoes of that in a lot of the organizations that I've scaled through my career.

Tobi Lutke: Yeah, it's a fascinating thing. I'm curious if any companies have really cracked the code on it. I wonder if you are aware of any kind of organizational structures or places that have sort of leaned into the generalism thing, anything that we can learn from them? Because I certainly have seen really world class teams be a mix of generalists and specialists, but also have a mutual appreciation for each other. Because one thing all generalists underestimate is how deep the expertise of specialists actually is. But I think all specialists also underestimate how wide the range of generalists goes. Right, because the Pareto principle works in your favor. You can get to 80% of any field in actually probably less time than 20% of the time. So when there's a mutual appreciation and a mix of it, I think I've seen magic happen. It's funny because it seems like there's very little organizational theory around how to tap into that.

Bret Taylor: There are a couple of thoughts there. I've thought so much about this because fundamentally, when you're scaling a tech company, your job is to consistently produce innovation. Because the reason Shopify exists is because you had a vision for an integrated commerce stack that could really run a business end to end. And you saw that through your personal experience when you were selling snowboards and you built the thing to scratch that itch, and it was true innovation. And how do you produce a company that outputs that?
There's a paper by a professor named Coase called "The Theory of the Firm." It was basically why do companies exist? In theory, you shouldn't have large conglomerates, let alone companies that just grow in arbitrary size and complexity. Again, my Cliffs Notes version of this is the reason they exist is transaction costs. If you and I want to do something together, we probably need a contract to do it. Whereas inside of a company, things like allocating new servers or even having a designer work on something or producing a new presentation for a customer, there's no transaction costs if you set up a firm correctly.
So in theory, if firms can scale output and reduce transaction costs, which give them an advantage over just a free market equivalent of that, in practice, I think companies often don't think about transaction costs. They introduce themselves. The classic thing is that every time something goes wrong, you make a process so that never happens again. And you fast forward ten years, and all of a sudden you have cascading layers of processes, and unless you're an anthropologist of Shopify, you have no idea why they exist anymore.
And I think that when you think about building an organization that produces innovation, I think it requires you've always been very intentional about how you grow Shopify. But that intentionality where you ask why to all the whys like my youngest - it can be annoying, but I think it's also without it, the only safeguard against the natural kind of friction companies introduce.

I think the economic rationale is simply that if you let companies accrue friction as they are prone to through inertia, you're essentially reducing the reason why the company exists in the first place, and we should all just go be entrepreneurs and set up services.

The other thing I read that was really interesting, and I don't know what is folklore and real, is when Jeff Bezos sent out a memo in the early 2000s at Amazon saying everything is an API. And if you look at it through the lens of Coase's theory of the firm, it's essentially mandating reduced friction between teams.  And I think if you do that as a sort of theoretical case, and the practice is always harder than the theory, you can end up in a world where smaller teams, in Amazon terms, two pizza box teams, can have the autonomy and agency with the value that you get from a larger firm.

And it's always easier said than done. And despite having all these thoughts, it's like I've failed at this many times, and it's really hard. Which is why there are not many companies of the scale and stature of Shopify. But I think it's always really important to think about constantly be very self-critical. I think about these things and not presume that the processes you have in place are correct, and more often than not they're not.

Tobi Lutke: Yeah, trust me, Shopify  is definitely one of those companies which when we get 98% of the tests, everyone asks what will happen to the last 2%? But we kind of have to right. It's so interesting because we are talking about the thing we're circling around is we have seen this magic,  we've seen this work and we are fully enabled and fully in charge and have been in many parts of our careers. And we kind of can't bottle it  fully up and roll it out and it just goes back to I'm pretty sure we are all going to be terribly embarrassed about the companies we ran in the early two thousand and twenty s at the end of our careers. And I think the best if there's sort of a Platonic ideal of the perfect firm where everyone trusts each other and internal transaction cost is absolutely minimal or nonexistent, and everyone's organized in such ways that autonomy and autonomy is almost total and accountability is completely mechanized and automatic, then if that's a ten out of ten, then I think the best company in the world right now is like a four out of ten. Maybe there's I was going to say.

Bret Taylor: Three but yeah, I think we're you're.

Tobi Lutke: Even a tougher judge would say it's going to work great. I think some companies are competing maybe for entering the next kind of integer here, but we're all very low on the scale. It might as well be on a logarithmic scale at the end of the day. I think this is why my stated optimization target for building and running Shopify is that I would like to be slightly less embarrassed at the end of my career than all my contemporaries.

Bret Taylor: Ironically, Shopify is probably more than any other firm in the world reducing the transaction costs for businesses.When I started my first company, we built our own servers, bought the parts, constructed them, drove them down to the data center, plugged them in. When my next company, Amazon Web Services, existed, it was great. Not only are there things like serverless computing now, but there are platforms like Shopify on top that we would have had to build ourselves. There are also things like Rippling for HR and WeWork. WeWork provides office space.

The innovation, particularly over the past ten or 15 years, has just made it so much easier to start a company and transact and fulfill. Shopify is contributing to that. I think it's actually really good for the economy. As much as software executives like us worry about our own companies, making it possible for small businesses to succeed is such an important thing for the world. If the world is dominated by large firms, it's probably not a world any of us want to live in.

It's sort of interesting that we're having this conversation about competition. To some degree, the Shopify mission is creating competition for these firms, which is great because this competition is good.

Tobi Lutke: The wonderful thing is that we get to the one technology that we don't have inside of companies - which is like a market, right? Capitalism has its detractors and has rest days and is certainly not perfect around the margins, but it's the best system we've ever come up with. And the thing it's brilliant at is the allocation of capital on in bad ideas, right. Unfortunately, this is always the first thing governments tend to backstop and this is when we get into real problems. But it's a high velocity and especially small scale. It's brilliant. It's the reason why our restaurants are getting so good. It's the reason why this is why I love our space so deeply, because we have millions of millions of entrepreneurs like Port England says out there and building always because for basically irrational choices, right? Like this is the age old and wonderful, you're going to spend 100 hours working for yourself so you don't have to work for someone else for 40 kind of thing. It's a deeply irrational kind of idea, but some people must and I think that's important and they make things that then are delightful to everyone else.

Bret Taylor: I had a really impactful sort of early part of my career where I was reminded of not just the power of entrepreneurialism, but actually how much, especially in the technology industry, we're not entitled to our future success. So when I first started at Google, it was a couple hundred employees, give or take, maybe a little more, but a small company, one building in Mountain View. And when we grew up enough that we moved to a few neighboring buildings that eventually needed a corporate campus. And rather than building one, we bought Silicon Graphics campus. It was right next door and Silicon Graphics was going out of business. It was really interesting because when you move into another company's corporate campus, when you first move in, it sort of has all of their branding and paraphernalia still there. So you're walking through sort of the carcass of a company that was once great, that was once great enough to have that campus that you just moved into.
And then later in my career, when I joined Facebook, they acquired a social network I made. We were in one building and eventually grew up, and we needed a campus, and instead of building one, we moved into Sun Microsystems old campus. And it was just really interesting because both Silicon Graphics and Sun Microsystems at one point were like the top of the stock market, the shining star of the technology industry. And that was like when I was graduated from university, not that much earlier than these moments. And it had a real impact on me that you're sort of walking through the halls of a company that built this amazing campus in the heart of Silicon Valley, and both of them ended up selling it for pennies on the dollar to the next company that was sort of graduating there. And it really, to me, it just had this impact, which is like, no matter what success we've had for the previous decade, we are not entitled to success in the future decade. And just the half life of technology companies is actually quite short. There's really relatively few exceptions, I guess, IBM and Know that have been around for a reasonably long period of time, IBM in particular, but there's not that many. And I think it's always good to remind ourselves of that because that sense of urgency that all of us have in the early days of our company, if you end up losing that, it's almost a near certainty that you'll suffer the same fate, which is kind of crazy. But also I find it really energizing because I love the kind of creative destruction and I love the sense of urgency that that drives. But it had a huge impact on me.

Tobi Lutke: Yeah, it must have felt a little bit like I sometimes think about what people in the Middle Ages fought when they sort of came across Roman ruins. Right. We don't even have the technology to create these things anymore. And I don't know what it inspired, but it sounds pretty inspiring to try to get back to former glories. I do know the cities we built today. Maybe I shouldn't say this, I got to spend a little bit of time in Rome with my family, and maybe we actually haven't improved on the practice of building cities. This is just a remarkable place.

But this is also like we can talk about company building for absolutely ever. There's no end of interesting topics related to it. Right. And I think partly because it is so organic, right. Like and you know, like a company in Silicon Valley is exposed to a completely different fitness functions than, you know, a company outside of it. And a company in retail is completely different from a restaurant.  And there's a thing about what's wonderful is when companies work, there's always luck involved in timing and all these kind of things because without this, this doesn't work. But some ideas, they're important to the success and some ideas are not. And it's funny because I'm not sure it's clear which ones are which. It seems like after you get after through the initial success, you then spend the next decade kind of trying to sort them into two different buckets. And I think you can only be like, hopefully you get the majority.

Right. But I think back to your point from earlier. Being extremely good at putting process in place after something goes wrong is probably not part of the core expertise of companies that worked. In fact, I have a funny hypothesis that there is no good process in the world. And this always terribly upsets my coworkers, but it turns out like whenever process is good, it actually gets renamed. I don't know if you've ever noticed this, it's always been the expense process, but to submit code is like a pull request also process, it just got renamed. Right.

Bret Taylor: 25% of the engineer developers use for their own tools. On other parts of humanity, we might fix all the world's problems.

Tobi Lutke: Well, good. You mentioned Google Maps earlier. I should know this, but you were an engineer on that team, weren't you?

Bret Taylor: Generalist or it's a complex footnote. I'll tell you the short history.

Tobi Lutke: Because sorry. And the reason is, man, that was an important product in Internet history. You were like, yes, let's make this really good and make the maps drag around. But there was like a before and after. This was like, okay, the web browsers can do things that we don't fully appreciate yet. Moment.

Bret Taylor: Yeah, it was interesting ingredients that led to it. So the story of Google Maps was at Google, I was a product manager for a product called Google Local, which essentially integrated yellow pages listings into Google Search so you could find local businesses in addition to web pages. It was an okay product. It kind of had a me too. There was a product called Yahoo Yellow Pages. It was somewhat ubiquitous at the time, and I wouldn't say the product was particularly differentiated.
So we launched the first version of it, and we spent a lot of time just talking about how do we produce something way more powerful than the competitors? Because we kind of made a first version that looked a lot like a lot of our competitors and really thinking like, what are directions we can take this that are significantly more powerful?
We started anchoring on the map we had licensed from a company called MapQuest, which you have to be our age to even remember MapQuest, but we put a little MapQuest tile next to the search results. It was really ugly. And we you know, just imagine if you really anchored the whole experience on mapping.
By coincidence, there was these two Danish engineers named Lars and Jens Rasmussen who were building a Windows app called Expedition that was a mapping application. Even in 2003 or four, whenever this was, no one was funding a company building a Windows apps. They had trouble getting anyone to return their calls. They were referred to us by one of our investors, I think, and we ended up meeting them and sort of doing an acquire of that company.
And it was actually turned out to be the biggest blessing in the world that what they came in with was a Windows app because it could do all of these things that you couldn't do in a web browser. So the reason we acquired of them is we love the look of the maps. Large and Yens just were mapping obsession. Like, they were just obsessed with maps. But all of a sudden, the watermark of what we wanted to achieve in the browser was so much higher of a standard than I think we would have set naturally.
So if it weren't for the fact that Jens felt so strongly about doing this in Windows, I can't say with a straight face that we would have had the ambition to go as far as we did in stretching the browser as far as we did. And then there was a lot of other ingredients at Google, Gmail was in development or had come out kind of around then. And there's a lot of people in the very early days of pushing web pages to become web applications, to make them truly interactive, to make them so you didn't need to refresh the page to interact with them.

So there was a crew of four or five of us who were just like, really deep on this stuff. And I was technically a product manager, but no one else knew JavaScript. So I just moonlighted and wrote all of the JavaScript for it. And it was just so incredibly fun because we'd crash the browser and then go talk to the members of the Firefox team. We had a lot of them working at Google at the time, and they'd open up the browser and the debugger, and we'd figure out what we were doing, and we were just testing parts of the browser code base that had literally never been exercised before. And it was just an incredibly thrilling thing.

And then, as you said, one of the achievements I'm most proud of is just changed web applications. The term Ajax, which I don't know if it's used anymore, came into vogue. Even sending data over as JavaScript, which we did, is now known as JSON, which was not named at the time. It was just very efficient to call Eval on a blob of. JavaScript while we did it, it ended up having sort of an outsized impact.

I love Google Maps, obviously, to this day, but I think it created a lot of the sort of Web 2.0 companies that invested interactivity. Many of them tell me they were sort of inspired by some combination of Google Maps and Gmail and all that, but it was a lot of different ingredients. You had a company that had a lot of people tinkering with the technology, which was important because it just meant you had colleagues to walk over to and say, how'd you approach this problem? How did you approach this problem?

And then the thing I love know Lars and Yen's silly Windows app ended up making our standards so impossibly high that we hit them. And I just wonder, had we not I know we wouldn't have obviously built it without them, but I also wonder, even if they had started in the browser, I think our standards would have ended up too low and we might not have done what we did.

Tobi Lutke: That's fascinating, right? It's like because man, I don't know. Is that story really well told outside.

Bret Taylor: I'm not sure it is. Lars and Jens and I are still friends, and I think Lars has done some interviews about it. I should tell it more.

Tobi Lutke: It's a great story. It's a great story partly because every time you're doing something, like, the first decision you have to make is actually are you going for are you just making another instantiation of what everyone's doing in the overtone window, or are you trying to go do better? If you decide to do better, you have like three choices ahead of you, right? You can try to make something perfect that's hard. You can try something that's like attempting to isn't totally approximate perfection, or you try something very quickly that just needs to be better than what other people are doing. Just doing a lot of work in the sentence. It's like the engineering culture, hacker culture, corporate culture is very good at the third. I think it's very obvious when people go for the second, which is sort of more like the Apple steve is the best example of that, although he might even be somewhere between one and two. And then one you never hear about because that never ships, so it doesn't matter. Lots of people go the funnel is like a wall. Lots of people go in, no one comes out.
Bret Taylor: It's really helpful to have a North Star. I think engineering can end up very iterative, and you can end up on local maxima for what you're trying to achieve for your customers. It's remarkable how often, given a really concrete goal, we're going to go to the moon before the Soviet Union, or in a much smaller capacity, we're going to make this draggable maps thing work in a web browser. When the goal is so unbelievably clear and you have single-minded engineers running towards that, it's remarkable. I think it's why I really believe in all developers spending some time with customers. What I find is that it makes the goals of the product more clear to the people building them, which often leads to much more obvious and acute solutions than the typical translation between sales, product management, engineering. If you cut out a lot of the middlemen and you have people spend time with often, you can end up with much stronger, much more clear-minded goals and much more elegant technical solutions to many problems.

Tobi Lutke: Yeah, partly because if you communicate through a channel, especially about product requirements, the only thing that will end up making it through there is like the highly quantifiable bits or the things that can be sloganized, but often the things worth solving are kind of unquantifiables. In fact, often the solutions that people suggest make it through, but it's actually the problem that needs solving.

That's interesting. And the problem, I think, for Google Maps was clearly that people just kind of got lost. Their maps were just not good or not available in the right form factors or like couldn't route or like GPS and all these kind of things.

And so, you know, I it's, it's, it's a fascinating story in general than when the you know, it makes you think like, just all all the limits are really matchined, like they're made up. There is no speed limit on all. There is no quality limit. It's just kind of safer to go like one step past what people are doing right now. There's probably less competition for just knocking it out of a park because no one else is trying so low likelihood of success. But if you can do it, it's what you really believe. Everything changes.

Bret Taylor:  It really does. And actually it took us almost 47 minutes to get here, but it is a good AI segue just because we.

Tobi Lutke: We're trying to be really good about this. I spotted like five possible jumping points and I just let them pass by because I knew there was no problem with getting there at some point.

Bret Taylor: Well, you talked about just like, it's the opportunity to knock things out of the park. I do think that we're at a significant inflection point with technology and just like when before and after the web browser or before and after the mobile phone building products that are native to that platform is very different than adding some features and functions. I think we're definitely at that point now with large language models and AI. And I think the opportunity to as you sort of describe the different ways of approaching product innovation, I think there's an opportunity to really reimagine some product experiences in ways that customers and consumers just absolutely love that are just completely different form factors than we had before.

Because the way I think about it is large language models and these foundation models essentially open the door to natural language understanding and conversational interfaces that actually work. And I think that's a meaningful milestone just because it's the only interface to computers you don't need to teach anyone about. It opens the door obviously, to generating text and images dynamically, which I think is incredibly powerful. It also just opens the door to basic human-level reasoning with computers. And computers to date have been conditional statements. If this then that. The idea that you can give a computer a goal and some guardrails and have it achieve that goal is I still haven't completely wrapped my head around it, to be perfectly honest with you. Anyway, I've never been more excited to be in our industry.

Tobi Lutke: As a consequence, you have seen again, like, all these transitions again, the before Netscape times, and then we had a web browser and sort of like it had Marquee and images. Images were controversial in the beginning of internet, right? And then we got Web 2.0, which really Google Maps and Gmail really showed us a way. And then I was on the other side trying to make frameworks to make it easier for people to partake. So we all contributed. And then of course, the mobile phone on the Internet, you can maybe even say SaaS might actually be a little bit like one of us technology shifts because that was like we understood the economic engine by which software is delivered in a way that leads to better software and happier customers. And so these are all shifts, they're all huge. And then comes AI. Which one of those does this compare the most to?
Bret Taylor: I've thought a lot about this, but if I had to pick one, I'd probably pick the advent of the browser. The move to mobile was disruptive, but the value proposition of applications was often similar; new applications did emerge, but many existing ones adapted. The browser required a completely new way to deliver software. AI has the ability to integrate many services, synthesize information, and act on your behalf in ways that were science fiction just recently.
The analogy isn't perfect, though. Technology waves have built on each other: the Internet enabled the browser, which enabled mobile and smartphones. Now 6 billion smartphones have connected much of the world, and AI can spread at almost no cost. Early Internet users had to search out information; AI and systems like ChatGPT are immediately available to all. The pace of adoption could be phenomenally fast.
Though there are historical analogies, AI feels different and overwhelming. In the early Internet, you had to be "in the know." Now anyone can access ChatGPT and start using AI. We haven't fully grasped how quickly new technologies can spread.

Tobi Lutke: Yeah, I guess I don't have this thought through terribly rigorously, but you raised this really interesting point there.  These transitions we talked about, almost all of those are a delivery innovation, right? Like at the end of the day, all of them, like mobile web browsers, web two kind of making things a little bit better are mechanisms by which it's now more convenient and easier to access. The thing that is running on a computer somewhere that's doing something really valuable.

Bret Taylor: Yeah, durable storage with some logic on top.

Tobi Lutke: Exactly. I don't know when Shopify first launched in beta, this was around the time the website TechCrunch launched too, and comments were all like, wow, but that's just a Crud database. And that stung a little bit, but it wasn't incorrect. Wouldn't you have imagined that Crud databases are kind of pretty valuable for the users, but I guess what AI is now a completely new thing running on machines. So all the Surfaces are now benefiting from this. And at the end of the day, we can utilize something that some people would describe as intelligence now in software, which I don't think we could before. This is why we had things like Mechanical Turk and so on, so that we could access intelligence.

Bret Taylor: It is really interesting, and especially, I think even the short term is very interesting because one is the cost to serve of artificial intelligence. Modern large language models is nontrivially high. So I think that's really interesting because traditionally if you think about the cost of rendering a web page or serving a GraphQL query to a mobile app is relatively free in the grand scheme of things. I think the cost of executing particularly large models is a nontrivial expense, which changes the dynamic a bit.

Similarly, they're probabilistic, not deterministic. Traditionally so much of the applications we've built have been rules and you file a bug if something goes wrong and you fix the bug. And to some degree, like inside of a software organization, your job is to make it as deterministic as possible. Even the most powerful models and the most simple questions will be wrong some percentage of the time. And that's okay because most humans are wrong some percentage of the time, and most things AI will displace or replace or augment is certainly not perfect by any means, but it's a very uncomfortable thing.

And so both the way you interact with these I actually was talking to a researcher at OpenAI who's a friend of mine that we were having a beer and he said today half the code I wrote was English, because he was basically prompt engineering. And he said I just could not have imagined this even a handful of years ago. The way we engage with this technology is different, the way our customers will engage with it, and just the naturally probabilistic nature I e errors will happen. And that's just part of the technology.

I'm just excited because I've never seen a technology that was simultaneously more powerful, yet so hard to reason about with the frameworks. We had to reason about software before, which is why I was so excited to become an entrepreneur again, just because I always have sort of a fly philosophy. If there's a really hard interesting problem, it's a lot more interesting and fun to be a part of it and to experience the nuances of it than to be on the sidelines. And so that's why I jump back in.

Tobi Lutke: I couldn't agree more. You invoked that it was important. I think at every one of those transition periods it was important that there's places of tinkering basically for lack of a better term. I tried to make that sound better than it is but it's actually just like kind of people doing the blue-collar part of innovation. When we didn't have browsers, we used telnet.

Bret Taylor: Actually, going on outside in the office right now, one of our researchers is doing a paper review going over an interesting research paper. But similarly, for every researcher publishing a paper on parameter efficient tuning or something really clever and advanced, there's a hacker on GitHub running an open-source model and fully automating their life with agents. And it's just as interesting. It's because, you know, it's moving so fast. And I love that expression, like the blue-collar part of it. It's so remarkable to me that in the morning you could read a paper from, you know, a Princeton PhD student and then in the afternoon be reading a GitHub repository of someone who automated their life with an open-source model. And they might have an equal amount of insight into the future of artificial intelligence.

Tobi Lutke: I love this. I really do think you need the cathedral and a bazaar we almost didn't have that we should all be really thankful for. At some point this came down to Zuck and making a choice that was probably against inconvenient for him and probably he needed to accept significant personal risk there because Facebook, if you think as a technology company you're somewhat quote there's no company as observed and watched and potentially maligned as matter. Yet they have given us basically everything for free that is used to do modern AI with PyTorch and these kind of things and they deserve a lot more credit. Shout out to Mark.

But now we have Llama models and with that I'm going to tell you I get up, I get on my phone, I get on the GitHub application which is like crazy and I'm going through the pull requests of five projects on GitHub. That is where the action is today, now, next week that might be different or maybe in a month this is different again. But the degree of innovation is incredible and the papers are coming my way throughout the day through Twitter or messages of people who know that I'm into this kind of stuff and it's just like I just find this absolutely fantastic. I love everything about it. I did not think there's going to be another golden age of computing but we are definitely in the middle of a golden age right now. And it's actually humbling. What's our role in this going to be. And it's really fun to, in the evenings, tinker on these kind of things. We have such an amazing opportunity here around our mission and it's super exciting.

Bret Taylor: It is. I couldn't agree with you more. I feel it's a golden age of technology and I love how unpredictable it is. I love that it's hard for me to predict the experiences I'll be using like a year from now. And it's fun to be in the vantage point, I mean, just segue a little bit back to Shopify. I love the AI-augmented chat you put out in the Shop app. But I also just think that as a platform provider, you're in a unique position to give this technology as a building block to all your merchants and all of their consumers in ways that you can't just do as a product company. And so it's actually doubly exciting when you have a platform as broad as the one that you have. Because I think you'll end up being the vehicle by which hundreds of thousands, if not millions of companies first bring AI into their business. And that's a cool opportunity.

Tobi Lutke: That's a cool opportunity. And it's also like, I mean, the mission of Shopify is we want to make entrepreneurship easier. And this is why I'm so excited about this. I think the mission is somewhat we got a lot of good bits out of a previous technology paradigm. We built a very good piece of software. I think it's constantly getting better and certainly I'm aware of all its flaws, but we were working on them. But there is a ceiling and that ceiling has opened because what people actually want is like, my God, how much of entrepreneurship is not governed, not prevented by regulations or lack of funding or it's like what people like, man. We can give everyone an infinitely patient, deeply competent, knowledgeable, permanently hardworking sidekick to the hero's journey right there, which can answer the questions, take over toil some tasks, and if everyone gets to start out with that, that is something that right now only very few privileged people have access to. And again, I love what entrepreneurship does. It's important everywhere. It's more important in the sort of less economically fortunate places. And one thing that the previous paradigm has done is it's provided incredible value to everyone on planet Earth. But there's sort of like a fortunate billion and then the less fortunate rest. And I think in general, what AI can do is, I think, act as a know equalizer is maybe not the right word, but like the value is just going to be accrued to everyone exactly in the same way because it's like it meets everyone where they are because everyone can communicate with their thumbs at this point. We've literally trained planet Earth for this moment.

Bret Taylor: It's interesting to bring that up. I don't know if you had this experience. But my grandparents, they never got a PC, but they did get iPads. And it was interesting just that it wasn't just Apple's product design, but just the form factor of a touch interface was so much more intuitive that that was the first computer that I think they just felt comfortable using. And it was great because before they passed away, I got to send emails and they were really connected to me digitally in a way that they weren't before. Now, as you said, talk about a great equalizer. Everyone can talk, or most people can talk and most people can. Some combination of talk type communicate in natural language. And I think that's incredible because you've essentially removed all the gatekeeping aspects of technology. You don't need to learn how to use a mouse and keyboard. You don't even need to learn the rules of the system, which are often opaque. And so much of using a computer, even easy-to-use ones, is understanding all these idiosyncratic things that have accumulated over the years. And I think we'll take a while to get there, but I think we'll very quickly get to the point where using the interface we use to communicate with each other, we can communicate with supercomputers in the cloud that are doing powerful things on our behalf. And I just find that incredible. I think people who have been on the sidelines of digital for a long time can take advantage of it. It'll be less intimidating. As you said, infinitely patient, not judgmental. I was talking to an entrepreneur who has a language translation application and they were reflecting that people prefer the automated version because it's not critical. It won't judge you for your poor accent or not knowing the vocabulary. And I was just thinking just the psychology of this is really fascinating. It's infinitely patient, not judgmental. I just think it could be incredible to onboard huge amounts of society onto all the technology platforms we've benefited from.

Tobi Lutke: I think this is one thing that is often missed. All of us in technology face criticism of AI and quasi-religious thought experiments about where this will lead. I find those interesting over beers but disruptive if they leak too far because they're just thought experiments, in my opinion. When you talk to people in rural areas, I'm currently in rural Ontario, and I ask locals about GPT, they say, "I love it. You guys finally made computers do what you always claimed they would." That's lost when everyone says they're not "deterministic" or "hallucinate." How did "hallucinate" stick? It's the funniest term. It's a funny world, but maybe AI is underhyped still.

Bret Taylor: I feel the same looking back at tech waves.  Technology, user experience, and business models will change. Salesforce delivered CRM via the web and charged subscriptions, not one-time licenses. Both were important to its success and SaaS taking off. We haven't figured out new business models that will emerge. When tech transitions bring new experiences, capabilities and business models, it may be underhyped. That's a lot of change. It opens doors for new companies, entrepreneurs, your customers, and lifelong software engineers like me. I've never seen a moment like this.

Tobi Lütke: That's awesome, Brett, thank you very much. This is a super fun conversation. Can't wait to have many more of those. And it's a privilege to have you. On behalf of all Shopify, welcome. Thank you for throwing in with us. I know you certainly could have picked anyone. So thanks for getting involved with this crazy company here and I think that's just wonderful. And yeah, thanks for trusting us.

Bret Taylor: Well, thank you for having me. And I'm incredibly proud to be associated with the Shopify brand and excited to work with you and to meet everyone else at the company, too.

Tobi Lutke: That's awesome. Thank you very much.