How Notion Reimagined Productivity Tools | Ivan Zhao

29:28Oh, it's fine. Yeah. Did you not consider staying at Inkling? Like, did you not think, okay, now that I've seen Inkling, there's all these other companies that are interesting. I'm meeting all these cool people that are doing other things. Like, it strikes me as a little bit strong-willed to move from China, not speaking English, learn our culture through movies and music basically and then your friends that you met there and then basically put your flag in the ground that you're going to start this company

30:11and then stick to your word that you're going to use inkling basically as a way to get your visa and then see that through. Yeah. You weren't tempted to deviate from the plan? I mean, how often does somebody have a plan and they just, it turns into reality? Well, the intention is never to join a company or start a company. The intention is to build a thing that doesn't quite exist. That's the intent. And the only way to do it, to start your company to build it. And what were you using in, I guess, college that made you feel like there needs to be more? The one that really inspired me

31:02was discover the early ideas of computing. Like where this industry came from. Most people don't know. Most people working in tech, they don't know where modern tech come from. Modern tech come from this group of people, the hippie generation 60s, 70s, not far from here, on the West Coast, San Francisco, Stanford area. Those are the people who took drugs, assets, and that same generation that apply the same thinking to this device we call computing. So back in the context, computers, computings are room-sized calculators. There are people who don't use much beyond calculating taxes or even before the ballistic missile projectories.

31:57That's what it's used for. And this hippie generation say, wait a second, if you just put a monitor in front of those room-sized computers, which back then there were people only printing out papers right back and forth, There's no displace. There's no interactive. You can create a new medium out of this. You can make this thing be interactive. And that's why this mouse was invented in that generation. And that created this computing industry, this like multi-tree, tens of trillion dollar industry that drives the world today. It wasn't like that. Right? And so this is our industry. Most people don't know the history of it.

32:43And even fewer people understand the intent of that, the original generation's intent. Their intent was computing are just like reading and writing. It's a medium that we can pick up. We can shape it. Just like you can write out your English sentences, you can phrase it however you like. some really good people can be a poet, some people can be a novelist, or you can use for business work. It's a medium that you can shape. Computing could be a medium that everyone can shape too.

33:19And, but today, by and large, we're in the AI era, so the story changed a little bit in the past few years. But until recently, computing is by and large the medium that only the programmer class can shape. Almost programmers are kind of like the scribes of the modern era. Only few people can read and write. So that's not the original intent. The intent of the 60s, 70s generation is computing software is a medium that everybody can shape and create whatever tools they need to do solve whatever problem they want. Just like you can use English too for all kinds of purposes. um when i learned about that idea i remember reading those early papers by people like alan k who was invented one of the inventor of the first personal computers or douglas angle bar who was more famous for inventing the mouse but his intent is actually the purest um realized this is the most one of the most powerful things yet to be realized and i want to build a product company whatever that is to work on this idea that got me to want to move to Silicon Valley to start Notion.

34:38It's kind of like Steve Jobs is part of that generation. And you know, when his daughter first writes Lisa on the Mac and he goes, he gets delighted because she's expressing what he wants the computer to do, which is that a little girl can write her name, I guess, with a mouse. Probably, yeah. From the movie or something? Yeah. Yeah. The Macintosh and the Lisa before that was one of the first graphic user interface-based computers for mass market. And the idea come from Steve Jobs almost 10 years ago before that, visit Xerox PARC. A lot of the people I describe, Anglebar, Alan Kay, they are, those people work at Park at that time.

35:28Not Anglebar, but Alan Kay's people. And when Steve Jobs saw it, he said, wow, computer should be interactive. Computer should be graphic-based. So he took a lot of this idea and wanted to create Macintosh or create Lisa with. But even he admitted he did not copy the whole thing. So the famous story is they show him three things. versus graphic user interface. He thinks that's the best thing ever invented. He also, he saw that and took that. They show him networking computer, park computer, network together. Maybe there's like dozens or low hundreds of them together. He didn't see that. So Macintosh was standalone floppy drive.

36:17The most, I would say nuanced one is object or an environment. He didn't see that too. That's the intent I was expressing earlier, is allow computers to be like Lego-like, object-like. People can take it apart, put it back together without being programmers. He didn't see that one. So in a sense, Steve Jobs and Bill Gates' generation popularized computers by putting a PC on every desk at home. but they lock computing, lock software into this application present. That's easier for people to understand in some ways, but a lot more rigid, kept the ceiling of how far you can go with your software.

36:59So that created a dual class of application software makers, which are programmers, and application users, which are the rest of us. And do you think, you mentioned three years ago with the AI stuff, things have changed. Do you think now we're going back to the original essence of what the hippies would have wanted, which is, you know, like I am in whatever replet, lovable building stuff in natural language? It's hard to say that's the 60, 70 generation original intent. Well, do you think we've gone back to the accessibility that the industry was intended to, that it was striving for?

37:58Technically, yes, it opened up more. Like more people can vibe coding. and I think it's more important to think, I think it's more profound than that. I think, and I don't think anybody sort of knows the right answer. What I mean by that is we all have an interesting relationship with the tools and medium we use. So right now we're in this transition from one major medium of the classic computing to this new medium of this more squishy, more language model based more AI based computing we're right in the middle of this and when you do this medium transition you sort of, people tend to use the new medium like the old medium let me give you a couple examples like when TV just happened people are sitting on the television do radio speech treated like the old medium radio.

39:01When telephone just happened, which replaced telegram, people initially only use telephone as a very parsimonious short speech note, right? Just transmitting information. Like they don't do gossip. They don't chit-chat with their friends. They put a telephone in a very formal place in their house. If someone rang the telephone at nighttime, someone died, right? It's very formal, right? When car just invented, people don't know what to do with to put a flag, someone with a flag in front of a card. Right? It's a very, we tend to use the old medium like the old medium, tend to use the new medium like the old medium in the beginning.

39:37I think we're in the same phase with AI. Like if you put together next side-by-side ChatGPT's UI next to a Google UI, it's kind of similar. Search box and give you some answers, right? But there's way more you can do with the language model with AI. We just haven't fully figured out how yet. And the best way we can do that is through the lens of the old medium, which is generating the answer, we're generating the code, your, you know, code gen examples. Maybe take us half a decade, a decade to figure out how to use this. But we're in that interesting transition point. And I think this language model started AI medium, it's probably as revolutionary as the computing medium.

40:32So my guess is a different question and answers that the original computing pioneers were thinking about with the computing medium. What that is, I don't know, but I feel they're close to the same order and magnitude and they're different. I remember when the public cloud first came about, that was the industry that I was in my whole career until I joined Kleiner. And in the beginning, it was the same feeling of, oh my God, like, look at all the advantages that we can get as a business, right? You're not racking and stacking servers and data centers. but everybody was porting their pre-existing applications with all of those dependencies and trying to shove them into AWS.

41:27And, you know, like it took years, but eventually folks kind of realized, oh, we need to rewrite these things from the ground up. And then there's an entire ecosystem built around that, that needs to be born. And, you know, I think if you believe that in this analogy, and maybe that pull of technology is even stronger in this case, I also think that's probably why, on the one hand, there is like more entrepreneurship and excitement about starting new companies. And on the other hand, there is more fear that you're going to be one of those old applications that's about to be rewritten. Because fundamentally, all of your dependencies are a native or a foreign language to what's happening moving forward.

42:22Yeah, the entire chessboard is redrawing. The ingredients we use to build tools, there's a brand new alien type of ingredient, type of wood show up on our doorstep. everybody trying to figure out what to do with it, right? That's the exciting part. Is it exciting for you or is it nerve-wracking? Meaning if the world continued to exist three years ago in the same way that it did, there wouldn't be an existential risk for notion of somebody building the entire stack from the ground up.

43:03now i think the question for you like any incumbent which is weird to think of you as an incumbent but nonetheless an incumbent in this era can you refactor yourself do you need to if you did what would that look like the existing ux paradigms that you've been building on do those need to change. I think a lot of things will change. Probably most things will change. And the question, different companies will answer this question differently. I would say Notion, classic SaaS, vertical SaaS company will answer this question very differently from, let's say, like us, Notion. Because our starting position or intent is very different from just vertical cloud-based SaaS companies.

43:49So maybe think about what's our strategy, right? Because we were inspired by computing pioneers. We want to make software malleable and tinkerable for end-users. In the past five, six years, we've been consolidating different vertical SaaS into one tools. So start with documenting, knowledge, project management. Most of those Lego blocks for knowledge work, we're taking them out one by one and put into Notion. So did most recently calendar and mail, right? So by and large, we have most software building blocks for SaaS. So our SaaS era strategy is consolidating those tools into one platform. And it turns out this is really good for AI because AI needs the contacts and tools to be together.

44:40We have been spending five plus years consolidating it onto one place. Now allow us to build automated, no longer just providing tools for our customer, but automate the work for our customer. Providing those building blocks in the form of knowledge work agents. Actually do the work for them. Right? So if you think about this, like, agent is a buzzword. Agent happens essentially software that do work for you. Right? Agent happened first in customer support vertical. And most recently in last year or so happened in coding vertical. agent really hasn't happened in knowledge work yet. Why is that?

45:23Because the tools and contacts for knowledge work are really spread out. Tools and contacts for coding is actually, if you think about it, fairly self-contained. Most, for coding agents, all the contacts are in the GitHub repos in one place, plain text file, easy to read and write. And the tools it needs to use are reading and writing those files. So it's very self-contained, So easier to create coding agents, extremely difficult to create knowledge work agent because knowledge work today are those a dozen different tasks that you and me do across a dozen different apps. Notion is one of the few companies in the past five plus years actually been consolidating those contacts and tools.

46:07And so you no longer need to go to different places to do different things, all the Lego blocks in one place. This allows us to build knowledge work agents. and this is what we're working on right now and give us quite a good head start. But then they have to be bought in to the Notion ecosystem end-to-end. Otherwise, the agent will probably struggle to hop out of the documents that it can't see. Not necessarily. We work with external contacts too, which last month we just launched our enterprise search product. So right now Notion integrated with 10 plus different external contacts from the Google stack to Microsoft stack, Atlassian stacks, right?

46:51Work across all of them. You can search across them and you can aggregate and do knowledge work across all of them. And when you think about the organization of your company relative to the product that you want to deliver, has your thinking evolved over the last few years around the organization of the company? I think many organizations need to, like how to build AI products is fundamentally different from building classic software. The analogy I like to use, it's more like, it's more akin to brewing beer than building bridges. The classic software, you sort of can engineer anything you can imagine.

47:38So like how do people build software in the past 20 years? you either do the Y combinator style, follow your customers, or you can do the Steve Jobs style, you have a vision. You don't listen to customer because you know better. The bold start with some kind of end. Then you spec out what the design is and engineer build it. Language model actually worked differently. You cannot build everything you can imagine. Oftentimes take you 70, 80 % there and never close the gap of remaining 20%. We actually learned this the hard lesson. And we tried to build a knowledge work agent a year and a half ago.

48:16All the ingredients are there on the Notion Lego side, but the language model wasn't fully ready. You can't really force it, right? So it's more like brewing beer that you have to create a writing environment, just channel the East to let them do their work. And this changes the sequence of how the software development process, in a sense. So you have to experiment a lot. You put design and engineer side by side together and with the data, with the right context, and just keep trying things and see what sticks. Versus the old way is more like waterfall-ish from your vision where customer need work backwards.

48:56And I think it took us a while to realize this. I think many other companies are also realizing this. Even larger companies, like it's really hard for them to, once the company gets large, it's hard to shift this way of building software. And with the beer analogy, I actually really like that, the beer versus the bridge. If you're brewing beer, what you're basically saying is, look, there's some ingredients that go into this. They're not ever exactly the same. And then, or maybe they are. And then what comes out is a slightly tasting, different tasting beer. Is that what you mean by the beer analogy?

49:36Beer analogy, that means it's organic, therefore you don't fully control it. You cannot tell the yeast, hey yeast, please taste that way. Right? You cannot force it. You have to channel the model. The wisdom and capabilities inside this model. Sometimes it does, can build the thing in your head, sometimes it doesn't. But the best thing you can do is just try and see if you can channel towards it. You cannot force it. Versus classic with building software, you can pretty much build anything in your Figma file. Maybe it takes three months, might take six months, but you can't get there. For beer, you might never get there.

50:16That's why if you go on Twitter, there's so many cool demos, but they're not harder to find real product. Because to be a real product, you need to close the gap of remaining 20-30%. That's the biggest difference. Is there a world where, you're right, we've done some of the lower hanging fruit use cases today, coding, customer success. Is there a world where we're overestimating in the near term, moving into some of these other periphery agentic-like systems that we would like to automate? It's just going to take longer and is going to be harder than what we thought? Probably.

51:02Like who did, like Andre Kaparthi talk about, he saw self-driving car 10 years ago, right? Took 10 years to go through all the coordinate cases and system bureaucracy, human bureaucracy to get there. Likely will be in the same, not likely, but very likely will be similar with knowledge work agents. But when you play with this language model, the underlying capability, you know, it's just so powerful. It's just so different. There's a human in there. It can do more and more of the things than we do. And then it's trying to figure out the tooling around it, the stack around it, how to unleash this capability.

51:53and then there's the organizational side of thing, go-to-market side of thing. How do you help your customers to understand this? They can adopt it. They can evolve their organization for it. That probably even slower, right? But you know that the nucleus of this, the capability is there. And the capability, the funny thing is the capability gets better every three-ish month. Doesn't see no trends to stop. Pretty interesting. What do you mean there's a human in there? I think this is the largest shift for all software companies, like a SaaS cloud software company. It's transitioning from selling tools to facilitate work to providing the work itself.

52:36The tools is providing the tools for other person to use to do some business needs. But now language model is having more and more of a human capability to reason, to think, to understand, right? Through largely through text. So software company, what they can do is to not only providing the tools, but providing the human sitting behind the tool as a package. You sort of see this in customer support, right? This no longer just make your support agents more efficient. It's actually doing the support. You see this in the coding agents, right? It's no longer just like IDE to help your developer tap completion, but actually to provide a final output.

53:25We haven't seen this for knowledge work yet. We haven't seen for many other sectors of knowledge work yet, but we're going to see more and more of them. Like we're building product for this, right? Like last, you can see early shape of this. Last month, we launched three AI product. Enterprise search is one of the things I mentioned, right? It's pretty baby hop, pretty straightforward. You don't have to read 20 articles to find answers, generate answers for you. Research, so you don't have to synthesize and write a report, you draft a report based on all your internal content and data. I used to something can take you hours.

54:04Now you can down in five minutes. That's another baby hop of knowledge work. And the third one is a meeting notes. You don't have to be typed out in the media notes, not piece together the notion like also a calendar and documents. And with some AI Lego of transcription and summary, you can have the notes better than any human can write. So those are some of the most fundamental knowledge work chunks that used to take people a lot of effort. Now AI can do this, not only providing the tool, but providing the work that humans used to do. And those three products, how much easier did it feel for you to build those three products in the last year than it would previously have just to build the product?

54:53I would say it's definitely a huge advantage going back to your question of how is Notion positioned in this, right? Because using the meeting notes as an example, we build a product in two months because we already have the building blocks for this. We have a collaborative document surface area. That's one of the best building blocks. We have a calendar product. So we know your meetings, when you should start meeting, push notification on that. That's another building blocks we have, right? Now we can bring in the building block of AI transcription and AI summary. Those four building blocks together allow us to ship one of the best-in-class A Media Notes product under two months because we already have this and we have your contacts.

55:39So all your knowledge, the rest of your company knowledge also lives in Notion. And this and work with Enterprise Search, after you finish the meeting or during your meetings, you can ask questions for, hey, what's the someone so-and-so talk about in previous meetings? You can do that. Again, compounding, right? We can consolidate tools. So in the SaaS area, we consolidate tools. so their company could be more productive. Like customers, like one, a good example recently, Ramp fully moved on to Notion. We reduced their tooling cost by 70%. But most importantly, it's not just the cost of the tools.

56:14The company can ship things much faster. Imagine each employee at Ramp has those three, four AI agents out of the box. All work seamlessly with the rest of Notion and the rest of the Ramp employee. That's a lot of productivity gain, right? that's still a SaaS error value proposition, reducing tool costs and increased productivity. AI even, in my opinion, dialed this up even more because language model wants the context and tools to be together. It's no longer just human business buyer prefer this. The model itself, we'll keep hearing the term context window, context window, because language model wants the context to be together so they can start thinking and problem solving.

56:57for you together, right? So if the context is fragmented, like the SaaS era, it's like, it sort of works because human is the glue, right? You go to your Slack, read some snippet of messages, go to your Notion, go to your Google Docs, go to your Jira, and go to your email back and forth, all the context is stored in your brain. It's really hard for AI to do that. Human barely get it to work. So if anything, AI prefers the context to be together, tools to be together, and we have the context and tools together. If one of the few platforms out there can do this, that's why I'm saying we're quite well positioned to build the knowledge work agent that yet to exist.

57:40As you think about moving into this really next era of Notion, would you agree it feels like a step function, moment in time for the company to move through? like in the early days of the company, correct me if I'm wrong, but you were searching for product market fit for years. Four or five years. Four or five years. You hired a team, then you had to fire the team, then you and your co-founder moved to Japan, which I guess everything comes full circle as you go back to Japan with an eyepatch, is then you go hole up for a year and a part, a year? Is it a year? rebuilding this? Yeah. Maybe two years.

58:23Two years. You two, three of you, two or three of you, are holed up in an apartment in Japan for two years. Two. We holed up for two years, but we've been traveling. But we spent the two summers in Japan. We went back to SF. And that was to get from one version of the product, which was maybe crappy, not good enough, wasn't getting adoption, to what then became Notion, right? I don't know, like, do you, when you went back to Japan, did you feel like, do I need to do that again? Like, do I need to, how do I find clarity, especially now when, you know, like, you raised it, whatever, $10 billion, like, you have, how many, like, thousands of employees?

59:18No, it's close to a thousand. A thousand employees. It's like, there's more to lose now. How do you, I don't know. Do you think about like what it takes to maybe re-underwrite all of your mental models? It's funny because I just got back from Japan. And one of the Nintendo trip was, we had an event with the mayor of Kyoto. Cool. Yeah, Kyoto was the city that Simon and I spent two summers in. And Kyoto mayor heard about our story. So they set up a tech conference, then we did a fireside chat with the mayor. And we'll talk about how much is a full story, come back to Kyoto. If you ask me one day, I'll be meeting the mayor and talk about that story.

1:00:10I would say, wow, that would be crazy. But here we are. Right? In a sense, the same question for the entire software industry. Like the chessboard is different. The material is really different. The SaaS era is in some sense very boring, very steady. You figure out your product market fit. You figure out your go-to-market, hire a sales team. Done. AI era, there's a lot of fog of war. And that made the game a lot more fun, a lot more exciting. You can go all the dimensions. Nobody knows what the right answer is. Um, I think we're in that position and we have one of the most unique kind of cars because what we did in the SAS era by consolidating and that allow us to do something really unique, like we just talked about.

1:00:58And, uh, it's revigorating for myself in a sense to think about like what you can build with this, all the Legos in the SAS era and this new type of Legos that's AI.

1:01:13it's almost like how long have you been building ocean for 10 plus years like uh you probably can't admit this to me but like you know pre-ai you're eight years in it's a slog like eight years of doing anything basically your entire adult life you're probably tired like to your point like it was kind of boring and you've been doing the same thing in a less dynamic game for a really long time.

1:01:48I wonder for a lot of founders that have been doing what you've been doing for the, in the SaaS era for as long as they have, for some, they'll probably pack up and go home. But for others, maybe it's like a, you know, I wonder if it breathes new life into you. Like it, it, it gives you. So Notion is one of the, we built one of the first, you know, application for AI product. And we got this conviction pretty interestingly because Simon and I were friends with OpenAI folks and we got early access to GPT-4. And that was late 2022. And when we saw it, it's like the world would be very different.

1:02:32There's a human package inside of this. and we actually lock ourselves into a hotel room in Mexico and for a week and a half build the first version of AI product and we actually launched the month before ChatGPT happened. So this is like when you build things, when the ingredient changes, there's just so much excitement. Like what new thing can you build with this? There's a lot of things become really hard. What's really hard and impossible before when you never thought about that before. Now you can't think about that way. That's a lot of fulfillment for people who build tools. So you did do your redo of Kyoto.

1:03:10We did. Yeah. We did the Kyoto thing in Cancun this time. In Cancun. You two in the hotel room. Yeah, us two in the hotel room. Different rooms now. Yeah. There's some funny photos, because like in Cancun, in Mexico, you cannot drink tap water. Right? So we're just being coding for a week and a half during a company retreat, actually. So the whole company was there for the week to do a retreat, and we skipped all the events except the opening keynote. Are you serious? Yeah, because we thought everybody else got early access to GPT-4 too. It's like, holy shit, the race is on. Let's just get out as fast as possible.

1:03:49It turns out, no, we were one of the first ones to get access. So the company is doing dinners, lunches, probably like awards and celebrations. and you and your co-founder are nowhere to be found? People know we're hacking on things. Are you serious? Yeah. I did a keynote, of course. I need to give a speech. Opening keynote. Welcome. I did an opening keynote. And I did a closing dinners toast. But besides that, we're just coding for the entire week. We actually stay longer because we want to finish the prototype. People know we're working on things. Are you kidding? No, it's understandable. You would like come out.

1:04:28I assume you would come out of the hotel and check in with the team and then come back? No, we're just in the hotel rooms. Like you didn't leave the room? Of course, we need to leave for dinner or food or something. But most of the time, we're just hacking on things. Yeah. There's an entire table packed water bottle at the end of it. Then we took a photo because it's so amusing. And so you would go out for dinner on the hotel property or something, and then you'd come back, and then you'd work into the night. and then you wake up do it again while everybody else is whatever like I assume there are other partners there too so everyone's partners are there oh no there wasn't okay it wasn't partners so all the employees are there they're all having fun I mean it's part work part fun part work part fun yeah and you had no interest yeah because it's like the once you see language to be the four level it's just the world will be different right and that's the excitement that's incredible it's fun to build things with this it is fun when you move from Vancouver went to Inkling left McInnes McInnes is going to get a good laugh out of this left McInnes actually he's going to be just fine he's the first investor he'll be fine so it's more than fine he was actually check number one check number one yeah he's fine Yeah, it's fine.

1:05:55Then you go to move to San Francisco for, with you, you, you meet your co-founder. Uh, I think I started to start by myself. Yeah. Then, um, there's another person, Toby. Uh, my mom actually helped with initial funding to get the visa. Cause he needs some, like put on some cash to get the visa. Right. Uh, she helped with it. Then Toby joined, then Simon joined, and Toby left after the year. Then just me and Simon. We went up to five rich people, went back to me and Simon for a while. Then you went to Tokyo. Then two of us went to Kyoto. Kyoto, yep. And we did that. We traveled to Kyoto, Sicily, Hawaii.

1:06:44Simon loves to swim, so we usually go to a place that he can swim, can't come included. We did that for like a year and a half, two years, and finished rebuilding Notion. And you were equally as convicted during those years that this had to exist? I think earlier your question is like, how does it feel working on something like a same thing for over and over again for so long? First of all, it has to exist. So if it doesn't exists, it's, there's a joy and focus to help you just working in that thing. And second, if you're just building this thing, if there's like, um, it's fun building things.

1:07:27Are you getting, you get lost in the flow state. You just, it's somewhat repetitive, but it's also really fun. Right. And that can just can keep going as long as there is. Um, like the funny is like, if you look at Spotify annual report. Towards the end of year, Spotify gives you how many hours of music you listened that year. Sometimes it shows how many hours you listened previous year. So if I look at mine, the years that I was coding and designing in the weeds with Simon during those Kyoto, Japan years, I listened to so much music because I would just be coding and listening to music. nowadays when you run a company, you have to be in meetings all the time.

1:08:12So that bar is actually a lot lower. Uh, different type of flow states. Do you find yourself craving the flow state of what it was like? Listening to music? Oh, definitely do. Yeah. Like I haven't properly co-worked designed for many years now. I think it's easier to get in a flow state when you're coding. but I'm better at design. Yeah. But it's like, I don't have kids. People say, when you have kids, it might be interruptive to do a flow state, but it's very gratifying when you're zooming out a little bit. Right? So building a company is probably like having a kid. There's a lot of things coming at you, but in aggregate, the thing's really fulfilling.

1:09:04Do you want kids? Oh, yeah. You do? Yeah, we're going to have kids. And do you worry about the trade-off of what that means for the company, for your first kid? I have a really capable wife. We'll figure stuff out. Yeah. It will definitely, people say you get more efficient, more focused. Right? Do you have kids? Not yet. I will. So, uh, we'll find a balance. When you started it on your own, why didn't you have, why didn't you start it with somebody? I had an idea already. I can code and I can design. I can write a, you know, a business plan to get my visa, so I can do most things. I don't need other people.

1:09:51And sometimes Simon and I, two of us, can do most things a company needs. So that allows us the notion to stay really small, and we sort of hire people in that shape. So our designer can code because we want people to make holistic trade-off in their head. Both the design context and engineering context, that also helps the company to stay lean and stay small. So we're not thousands of people. We're a high hundred, a thousand people. I cannot wait to see what you do. Can I ask you a pretty different question? I'm just fascinated.

1:10:33if notion went public and you had the whatever if notion got bought it doesn't matter if you walked away from notion what would you be doing i'll get bored but what do you think you would do to fill your time i might do architecture like i would love to build buildings creating physical products would be fun? Probably architecture. Like, my wife and I love interior designs for our offices or for our home. Architecture is like a more adult version of that. I read somewhere that you get support tickets sent to your phone. Does that still happen? Yeah, it's on my phone. So if you just like, right here, send us tickets.

1:11:22Yeah. Oh, it like actually pops up. Mm-hmm. It's, uh... How many tickets are there right now? I don't know. That looked like a lot, like, meaning, not that, oh, there's a bunch of people filing bugs, but every time, if you're away from your phone for an hour... There'll be a ticket in my lock screen. It wouldn't send me bus push notifications. I know, but when you open it, when you open your phone... There'll be one-ish thing on the very top, so I get a rough impression of what's going on. It's, like, ambient. Like, you get... You care about your office, you care about your customer, you let the ambience into come at you.

1:11:57It wouldn't buzz me, it wouldn't buzz my watch, but we're just passing by, you get something. You strike me as someone that might have unique management styles. I wouldn't say I have a unique management style. So you, do you have the patience for a one-on-one? Oh, I do. Yeah, I do one-on-ones for, uh, like my staff team, but they are all very excellent, mature leaders. They're all much better manager leaders experienced than I have been. So I don't need to manage them. When you interview, how do you interview? I do spend a lot of effort with interviewing. That's something I care and consider one of my crafts.

1:12:48First of all, having a normal conversation. So when usually people come to interview, we call a business jacket that's on. You need to take off their virtual business jacket. You make sure the other person understands you are a human, and you see them as a human, and you are here to find a match. It's not here to judge. It's not here to critique or exam.

1:13:17And because at the end of the day, it's like fundamentally it's a match, right? you have your value, you have your skill, your craft. We have, company has a value and company has to craft a skill for that role. You figure out a match. The best thing is to be very honest with each other. Both sides take out the business jacket and spend an hour to find a match. And so then a normal conversation, then we do a lot of references and a lot of back channels. Because it's one hour where you have a panel of four or five people that's only four or five hours. It's very like blind person touching elephant.

1:13:56You can't get the full picture. But if you have back channels, those people spend years with the other candidate, you have a better picture. Then they find a match. And what are you, are you steering the conversation in a certain direction? What are you, where are you exploring with a candidate? Depends on if you, if I'm interviewing for their craft and or their values. Like if it's a product design intro, I usually go deep on the craft a little bit. Just get a sense, like from talking about craft, you get a sense their value too, right? But if it's a value, then you will have a few values in the company.

1:14:42I want to probe around and see like, how does the person match with our values? There's not, it's not a procedural. It's usually just a bag of different questions or tricks, and you just follow your notes after that. Otherwise, it's too checklisty. You don't get the real out of it. You don't get a realness out of it. Yeah. That's how I feel about doing this podcast. People walk in with a business jacket, and the goal is to take that off as quickly as possible, but actually have a normal conversation. I like to start just by don't telling me we're starting this, we'll just go right into it. Right.

1:15:17It's very good. Do you have other crafts? Do you have other things that you think of as a craft, personal or professional, whether it's in your job or in your home life or whatever it is that you also treat like a craft? I like to think about going back to why the AI is so fascinating. like philosophy it's philosophy or history the overlapping ground of both it's something my wife and I spend time reading we had a philosophy tutor during COVID every weekend if we're not working we're usually reading books in history or philosophy she's actually spent so much time in philosophy nowadays it's almost like taking a master or PhD on her own Right?

1:16:12So, um, I think that going back to the AI topic, it's like, like we're in, you almost practice this. Right? Like we're in this bridging ground from one type of medium, which is classic software, classic computing to this new AI power computing. What do you do with that? Or you can read about how the past mediums, like I talk about the telephone telegram of the world transition. and now by building Notion, I have a front row seat to experiment, to see what happens and to see what sticks, right? So I find that really satisfying. Going back to your like, hey, does AI make you be a different type builder?

1:16:56Yes, because finally we have this like shift from old medium to new medium and you can build with a new medium and my interests were learning in the history and philosophy of the past allow you to apply that a little bit. Not fully, otherwise be too much like armchair philosophy, but you can at least guide your direction of company a little bit. And maybe just one or two more, like on the personal side, I'm very fascinated with like, how do you run a day? Like, how are you organizing your days do those days have a specific shape around a time an end a start physical activity food like how do those things all yeah it's not too fancy i'm more on burrito than side than the michelin star side um like we have staff meeting uh what time are you awake 7, 6.30, 7-ish.

1:18:00What time are you in bed? Last night, I was jet-lagged, but typically 11.30. Okay. I sleep seven hours. I usually try to do some exercise in the morning. I was running this morning. Like this morning, for example, I woke up. I try not to let my Slack notification, I turn off notification on my Slack and emails. I go to them, don't let them come to me. really important thing will come to my text messages. And this morning I was trying to do some writing. I flushed out some of these ideas we just talked about. And I did a couple hours writing, did not finish it. Then going back to Slack and emails and did a run for two and a half miles and took a shower and got here.

1:18:51You wrote first. You woke up, wrote, then you ran. I woke up, make a tea for myself, make a coffee for my wife. Then I tried to write. And are you like, do you need to turn your brain off at night to go to sleep? Like, do you have to do something to get away from work? Usually I'm on my phone all the time. I try to read before I go to bed. You read some philosophy or history? That's on weekends. I'm trying to re... Like last night we were learning about this Japanese woodcut master from early 1900s. So how woodcut blockers are made. And pretty complex procedural. We're watching a little bit of YouTube videos and we're looking at the books.

1:19:44And yeah, we talk about it, then we fall asleep. I think there's more con than... more pro than con. Because fundamentally, we're living in Silicon Valley. South Park is one of the center of this place. It's a human endeavor. And a lot of the craft of trade, like the practice you learn from other people. How do you work with language model? Nobody knows until we're starting to figure out how to work with this new thing. You learn from talking with other founders, talking with people who are building with this, how to build a company. we sort of know this a little bit more through the SaaS era now.

1:20:22Now it's a completely different game with the AI era, right? You learn from other people. They happen to all live in San Francisco. There's no others to live in the city, so you have to hang out with them too, right? Isn't it crazy? Like, I don't know if you experienced this when you were in Japan, but I was in the East Coast for the 4th of July, and I was with friends, and, you know, one afternoon, some question came up, and I opened up ChatGPT, and I just started talking to it. And everyone's like, what are you doing? And I'm like, just hold on a second. And then, you know, starts talking back.

1:20:56And they're like, what is that? And I'm like, oh, it's OpenAI. And they're like, yeah, but like, you talk to it? And I'm like, yeah, all the time. And right then and there, I realized like, oh, we live in this crazy, like, they're not from like outer space this is like i'm on the east coast of the u.s you know and they were like and somebody was like so i can just talk to i'm like yeah and by the way the more you talk to it and the more information you give it the more thoughtful it'll respond to you and i just had a almost like an outer body experience no pun intended because sometimes you forget what kind of world we live in in this seven miles of the city.

1:21:50Yeah, or probably plus a little bit more Berkeley and it's different. That's true, yeah. Well, technologies never distribute evenly. They dissipate, right? And the epicenter is here. And we are, people self-select to be living in this region. So they tend to pick up new things quickly. to my earlier point, like we're in this very interesting shift of new tools and old tools and new tools. And we're in the new camp because we're the fact that we're here, we're living, working and living in tech. Uh, yeah, the world will be very, I think it would be very, very different. It's just sort of can, like a knowledge work.

1:22:38This is the thing I'm reading and thinking a lot. So we'll be very different. It might take some time for the technology to truly dissipate to get into organization for them to talk. But like, if you think about, you can sort of see this in coding agents a little bit, right? Like Simon, my co-founder, in the past few months, he still codes every, he's still one of the most productive programmers in the company. But he doesn't manage, he just likes to build things. But in the past few months, he hasn't coded for a few months now. So if you walk by his desk, he's kind of like managing five, six different windows of coding agents do the work for him.

1:23:23Right? And it's like our relationship with information, with the work is changing, evolving towards that. Like you no longer do the work, but you're managing agents to do the work. but I think that's one part I think the bigger part is the scale that information and work can go through a person and company will also increase it's almost like it's almost like you go into a small town or small you know go into a very packed European city like Florence where everything's packed in that small part that you can walk from one end to another in 20-30 minutes versus a larger city like Dallas or Houston.

1:24:07It's made for cars. Much further, you cannot walk but scale and speed is much grander. And sometimes our relationship with information is moving towards that. Right? Someone like Simon can go in through managing output way more things than just if he hand-coded, hand-crafted everything by himself. I think where this region is probably the first to realize that and coding is probably one of the first to realize that too and will go to your friends on the east coast working in different industries different sectors how long that take I don't think anybody knows but it's the it's the shape of things to come this is why I think timing matters so much and so often things can be right but mistimed that come from Silicon Valley.

1:25:01Because, you know, let's take the example that you shared of getting the early version of the GPT model. You thought, oh my God, I'm late. So you spent a week and a half in Cancun, hold up. You realize not only are you late, nobody even has access to this thing yet, you know? And so it is a good reminder that, boy are we early we're still so so early and so many of these tools like whatever the expression is that redwood falls in the forest and nobody hears it better make sure that when there's tools people are ready when the tools are ready well if people are not ready sometimes bureaucracy is a good thing because it can slow down things, allow the rest of the world to digest, to take time, right?

1:25:59Silicon Valley, people are so open-minded. Hippie starts here. Tech boom starts here. We're just taking the new things, like no tomorrow. But that might not be the best thing. It's like it's... By the time you are 20-some, 25 years old, you sort of don't want to learn new things. And that could be good, could be bad, right? but all organizations, all personal things need to adopt new tools, going through these things, to have a personal or system bureaucracy against new things. I think with this AI wave, we don't know, but it will be, it could slow things down a little bit, but it might be necessary.

1:26:41Thank you for doing this. Are you hiring? Yeah, always hiring. Anywhere? Where? Kyoto? We have a Tokyo office. We don't have a Kyoto office yet, but Tokyo office for sales. So if you're proud of engineering design, we have office in SF and New York and Hyderabad, India. So we're hiring those locations for proud engineering design. Other offices, we have Dublin, Ireland for sales. Sydney, which has opened an office recently. Tokyo, Seoul. So we're opening Paris, Munich, and London this year, too, for sales. You're going to be busy. Well, our customer base is 80 % international. So that's why we have to travel a lot to meet them.

1:27:29Yeah. When you hear the word grit, what do you think of? Grit. I really like the word grit. Why? It's generally a positive human characteristic. like you go through things doesn't matter the up and downs you allow whatever you believe to compound usually takes many years a long time to compound and create something unique out of the other end right um grid is the characteristic to allow that happen to allow compounding and compounding creates something beautiful usually useful I love that answer I love this I love this conversation. Thank you. Oh, I'm glad. This is delightful. I appreciate you doing this.

1:28:17That's it for now. If you liked the episode, please leave us a review or go back into the archives where we've done more than 200 episodes with some fantastic folks. This podcast is a Kleiner Perkins production, and I'm Juven. Thanks for listening.