Town's CEO on the self-organizing company

I gave this startup access to my email and calendar, and within minutes, it had written a scary good dossier about who I am, who I know, and what I'm working on. And I kind of loved it. Is this the first step toward a self-organizing company? Well, I'm talking to the CEO of Town. That's this week on Platformer. This episode is brought to you by Jira by Atlassian, where teams and coding agents get the context they need to do the right work. Try it free at jira.dev. That's J-I-R-A dot D-E-V. Welcome to Platformer. I'm Casey Newton. This week, our guest is Jean-Denis Grez, or JDG, as he is better known to most of his colleagues. He's the co-founder and CEO of Town, a startup that is, frankly, brand new, but doing some really cool things. And so I wanted to talk to him. They're betting that the future of work software isn't like a better workspace that you organize, but an assistant that organizes itself around you. Doesn't that sound nice? By the way, his title is not CEO of Town. He is the mayor. There's your Silicon Valley fact of the day. We'll get into it. But first, here to give us some data on the state of AI and jobs, as always, is Platformer fellow and Gen Z AI correspondent Ella Marquiano. Ella, how are you? I'm doing well, Casey. How are you? I'm doing really well. What's been going on with you this week? What has been going on with me this week? I mean, besides playing a lot of Minecraft, to be honest, I've been reading about how the IT sector in India has been doing. And as, like, you might expect from, like, the number of queries you've sent to chatbots for IT help over the past year, there are some worries about, like, whether IT jobs in India, like, will survive basically the current AI revolution. And my understanding is that IT is a huge sector in India that employs millions of people. And so sort of does AI start to affect these jobs is, like, a very big economic question. And to the extent that, like, all jobs are affected globally, it feels like maybe the Indian IT sector could be one of the first places we see that. Is that right? Yeah, it's like it's about 7% of India's GDP and employs 6 million people. Okay. And so, yeah, it's like a very enormous deal, in particular for the country, but in general. Well, so tell us the story that you read this week. Yeah, so I read some reporting from the Financial Times on basically difficulties in India's IT sector over the past few months and, like, especially, like, IT workers' anxieties about whether they'll continue having jobs. So, like, the big picture is, like, there are at least a fair number of companies who over the past two years clearly have reduced their headcounts. For example, Infosys and Wipro, which are two of the, like, big six IT companies in India, both employ, like, at least hundreds of thousands of people, have reduced headcount by about 5 or 6% since 2023, which, you know, in raw numbers is now a lot of people. In particular, Infosys in the last fiscal quarter of 2026, so, like, from January to March, reduced staffing 2.5%, which is a lot of people. Yeah, that is a lot of people. And is there any evidence that these job losses were AI-related? Yeah. So one thing is I'll lean on an anecdote here that I thought was interesting from this Financial Times reporting, which is Oracle recently laid off like 10,000 workers, and there was one worker who spoke to the Financial Times anonymously and was basically like, my IT team was like working with management to like build tools that did parts of our job, and like at least I felt, the person said, that like while I was doing this, like I was helping the company and they would like keep employing me, and this would be a compliment to me. And now that like so many people have been laid off, it does feel, at least to the person who was interviewed, like they were basically being asked to build their own like build systems that replaced them. Right. The classic train-your-own-replacement dilemma. Now that said, are there reasons to be skeptical that AI is the leading cause here? Yeah. One is like on a like really big picture level, like I, you know, this afternoon I was just like reading the whole Wikipedia page for the IT industry in India, and, you know, there's a section that's like, you know, threats of India, like companies in India, workers in India have been worried about threats of automation from AI. And then it cites like an article from 2017, an article from 2021, and an article from 2025, where like basically even before chatbots, there have been like many waves of India's IT industry where it's at least reasonable to believe that you can find some kind of tech that does what people do, like automated creation of software in some pre-chatbot sense, you know, automated assistance, queries, stuff like that. But on like the macro level from like 2016 to 2025, like data from like India's Big, like, tech trade group shows that, like, a decade ago, it was about 4 million people working in IT in India. Now it's 6 million. And, like, year over year, basically it just increases, even though there's, like, plenty of, like, basic reasons, especially since 2023, to, like, naively think that AI could do some people's jobs. Yeah, this feels like a story that we see so often as we have been tracking this narrative, which is like, this seems like it's an early sign of something, but there are some confounding variables, so we're gonna keep our eye on it. Still, that said, I'm not sure, you know, that I would bet on this sector being much, much bigger five years from now. So something to keep an eye on. Last question before we go: How is AI affecting Minecraft, if at all? Um, I actually was asking Claude, like, a lot of stuff. Like, I used to, like, Google things all of the time, like if I forgot how to craft something. And now I'm, like, in the middle of a task and I'm like, Claude, how do I make a bucket again? And that's completely changing bucket crafting in Minecraft. So, I mean, it's changing, like, the SEO slop websites that, like, cater to people who have forgotten how to, like, play Minecraft. Very good. Well, I'll tell you one place where AI is affecting my job is that I have been playing around with this new app called Town. They only came out of stealth in June, so it is very, very new. But I don't know, it made me excited. I wanted to find out more. And so after the break, we'll have an in-depth dive into the product and what it means for the future of work with JDG from Town. Jira, by Atlassian, is where agent speed meets team intelligence. Assign any work item to your favorite coding agent and see exactly what it's doing without leaving your flow. When something stalls, you're not digging through logs. You can see what's running, unblock what's stuck, and stay in control at scale. Don't let your agents start cold. The teamwork graph feeds them context from across your entire stack, delivering 44% more accurate results with 48% less token usage, so the work they pick up is the right work done right the first time. Try it free at jira.dev. That's J-I-R-A dot D-E-V. Welcome back to Platformer. My guest today is Jean-Denis Greze, a.k.a. JDG. He's the co-founder and CEO and mayor of Town, an AI assistant startup that only came out of beta in June with a $55 million Series A led by Andreessen Horowitz. And the main gimmick here is an AI assistant, which they call a Townie. You give it a name, a personality, and an animal avatar. Mine is a finch for some reason. That's what it picked for me, although you can change it. And then you give it something much bigger, which is your email, your calendar. If you want, you can give it your Slack, your meeting notes, your Granola meeting notes. It studies how you work. It builds a private wiki about your life. And then it starts to proactively start to do some of your jobs' most tedious parts, like triaging your inbox, drafting replies in your voice, briefing you before meetings, logging invoices. Jean-Denis's own Townie is a silver fox named Ivy, which he says is a nod to his prematurely gray hair. Before Town, JDG spent about seven years at Plaid, the company that connects your bank account to apps like Venmo, where he served as CTO. Before that, he was a director of engineering at Dropbox. And before that, I love this, he went to law school, practiced law for exactly one year, and quit, saying that his worst day as an engineer beat his best day as a lawyer. Town is still small. It's about 20 people, and they work five days a week in an office in downtown San Francisco. And it has become a kind of hipster favorite lately in Silicon Valley, spreading through venture firms and small startups mostly by word of mouth. It's also gotten some traction outside of tech. The company's favorite customer story is a plumber in Sydney whose wombat townie processes 300 emails a day. That's too much. Who is getting 300 emails a day? Anyway, he says that Town was helpful enough that he was able to take a six-week vacation. But what I want to understand today is the design bet that's underneath all of this. Tools like Notion have historically asked us to build and maintain our own workspaces, every wiki page, every database, every label. Town's bet is that nobody actually wants to do that work and that AI can finally do it for us, a workspace that organizes itself and an agent that can do things for you. And if that's right, it has big implications for how all software gets made and for the millions of people whose job is the organizing. So lots to talk about. Here's my conversation with Jean-Denis Grez. JDG, welcome to Platformer. Thank you, Casey. I'm very excited to be here. So for the past 15 years or so, productivity software in my world has meant tools like Notion, a lot of beautiful empty boxes that you have to fill and organize and maintain. And if you're like me, you mostly do not ever fill out those boxes. And what intrigued me about Town is that it seems to start from this opposite premise, which is that nobody wants to maintain the company wiki or a document library. But what if the wiki just wrote itself? So was that part of your initial thesis, or did you start somewhere else when you went to build Town? Yeah, we started somewhere else, actually, but in hindsight, it links directly to kind of this: how do you keep, like, the state of your world up to date at all times? So we started with just this idea, like, why has no one built AI that helps you with basic email and calendaring and meeting prep? Like, just looking at pretty common things that every knowledge worker has to do, right? You get a lot of the emails that you get, you need your human brain to interact with, but you often get emails that are just, you've answered them like 50 times before, and why can't something draft that for you? Calendaring is, you know, you have like three back-and-forth exchanges with everyone that you know to find time that makes sense, that seemed like logical that AI could handle that for you. And then walking into every meeting prepared, like just a reminder, like, what's the meeting about? What emails were exchanged to get there? And then coming out of the meeting, making sure you don't lose any action items. Like, those were the focus areas for the product. And then as we built that product, we realized, well, to do those things well, you need an understanding of who the person is. And to do that, you start to build, you know, like a wiki, a knowledge base, which is like, hey, this is who Bob is. These are Bob's scheduling preferences. This is like the kind of—these are the seven things that Bob is working on right now, and like, how can you connect those to this meeting that he has tomorrow? And so we started realizing that, you know, for this kind of a product, you might call it an assistant. I don't like that word. We can talk about that later. But to be effective, the more it can pre-process about who you are, what you do, who your coworkers are, the more effective it can be. And now the focus very much is in this, like, can we understand who you are, what you do, so that we can both, like, do some of the work for you in the background, suggest things that, you know, would make you more effective in your job, and, like, can we do that on a continuous basis? Yeah. So I started using Town just a couple of weeks ago. Full disclosure, I am not yet a paying subscriber, but I would say the odds are looking very good. And one reason why I've been so interested in it is that I am about to start this new company. I'm starting this new company with my friend Kevin Roose. We're going to start this new podcast. And unlike Platformer, which has been a little bit more of, like, a solo entrepreneur thing, like, this one, we're going to have multiple full-time employees. We're going to have office space. Like, we're sort of doing a company for real. And there are just all of these things that I need to coordinate in a way that I've never had to coordinate before. So when I've gone out to use a tool like Notion, and, you know, they have their AI agents and they'll set up various kind of pages on your behalf. But I have to tell you, when I logged into Town and you just started building a bunch of stuff for me, I felt like, oh, this is the thing, like, that I actually want. Like, do you think we're getting to a place where humans maybe just are not going to have to do as much of the organizing of that kind of basic, like, company infrastructure? Um, I am very AI optimistic in the long term. I think we're still pretty far from that happening. Like, you know, and I say this, like, look, we see, like, our product is an assistant for everyone. And so there's this part of the product, the one you just described, which is like most people don't know what AI can or can't do. So actually onboarding people onto AI, like understanding what you do and then suggesting, being like, Hey, you're a real estate agent. I could, like, check new properties in your market every morning. Do you want me to do that? Just being like, I can do that, and do you want me to do it for you? And here's a button that you can click, and if you say yes, then every morning I will do that task for you. That's important today because people don't have this mental model. And so this kind of idea that you can use AI to suggest what AI can do, it goes— It goes some of the way towards automating, like, the toil out of what you do, but I think it touches today probably like 10 or 20 percent of the toil of a knowledge worker. The other 80 percent of the work, I don't think we are there yet, you know. And you can read the headlines in the New York Times about, like, job displacement and retooling and all that. I just think, like, in practice, if you go to Cleveland, Ohio, and you go talk to, like, 100 knowledge workers working at, like, accounting firms and, like, local businesses and whatever, you're like, how much are you doing? So whenever you're like, how much are you doing with AI? How much time is it saving you? We're still in the like 10 to 20% bucket and not in the like, it's doing everything for me. Now, that's like, you know, the bet that we are making, though, like the company is that over time, it can be better and better at suggesting things that it can automate for you, right? And, but, you know, we're humans. So what happens when, like, you take a very simple example, you take like an executive assistant that spends a lot of hours every week scheduling meetings, labeling traveling expenses, and those kinds of things. And now say that person is using Town to spend much less time scheduling or labeling expenses. It's not like they're on vacation for the, you know, 20% of their time that they've saved that week. So it's not like they get a day off in the week. It's like there's plenty of things that they could do that would make them a better executive assistant to the people that they're supporting. So now they're doing more of that kind of work. And that work is, by definition, not something that the AI can do, right? And so, like, I think over time, the automation just gives us more opportunity to do the things that are higher value, like in the human economy. That makes sense. I mean, to me, the insight, though, is that today AI has just not been proactive enough. I want to get into that a little bit later, but I want to ask about this one particular way that you're very proactive in a way that is apparently expensive for you, which is that when someone signs up, you know, you ask folks to plug in their email, their calendar, their docs, and you build this personal wiki about them. And I believe you said this costs you about $100 per user. Is that right? Like, total over the first two weeks of the user, yeah, it can be about—I mean, we're not going to have it be that expensive for everyone, but that's roughly the cost upfront, yeah. Well, I have to say, I feel like it paid off because when I went through this experience, I got this dossier back about who I was and what I work on and, like, who are the people that are important to me. And this happened within, like, I want to say it was like 90 seconds. You know, it wasn't this sort of like, go away and we're going to, like, you know, sample your emails and figure—like, you just sort of, like, knew right away. And I wanted to hear a little bit more about, like, what role this part of the product plays in the experience that you're building. Yeah, so for you, my guess is because you're online. Yeah. Yeah, I'm not going to say famous. I'm going to say, like, there—but there's content about you online, a decent amount. Yeah. It probably was able to do that at a higher level of fidelity faster than it does for most people, because the way it does it is after you connect your account, it looks a little bit at sampling of, like, recent emails. It does some online searching about you to generate the first pass of the dossier, the one that you see within about, like, 30 seconds to a minute. And then in the background, actually, for the first week, we build a more, much more complex version, which is the one that, you know, starts to cost, like, double, maybe triple-digit dollars. Yeah. The first part, the dossier, it's the first aha moment, actually, for the user. Like, most people, when they sign up for the product, they're, like, intrigued. It's like, oh, you can schedule things for me. You can do things with email. Interesting. You sign up. You have to name your assistant—we call them townies—so you name your townie, you give it a name, cool little avatar, you do those things. And then we're like, hey, my name's Ivy. I'm your townie. Here's what I know about you. Here's a few things I can do to help you. And most people are like, what? Like, that's what? You know me, you understand me. You're suggesting things that are related to my job or, like, the work that I'm doing right now, and that's it clicks. It's like, people are like, it's not like a product; it's actually like a relationship. This feels more like someone that I work with than it does a piece of software. So we do it because it is the core part of the product. But the reason we try to do it really quickly, even if it's not as accurate, is because it's the first thing that gets—it makes you feel different, right? It's like a different emotional feeling towards this product. Totally. And we're going to talk about the townies and the assistants. And so I hope you're not bored with me harping on this wiki thing, but I have to tell you, every once in a while when I'm using productivity software, I just have one of these moments where I'm like, Oh, like everyone is going to copy this. And I had that when I saw this wiki that you built for your customers. So tell us a little bit about what it is and how it works. I mean, first, just like, I love original ideas, but we cannot claim that the wiki is, like, our original idea, right? I mean, if you—Capathi and other people have talked about how you can use AI to take a set of documents and build kind of a wiki and information base out of a knowledge base. There's products in the open source world like Hermes Agent, where it's, like, fairly common for people to do this. The way I view the company is most people don't have the time to, like, install these open source pieces of software and tweak them to get it to work. And our job is to take this incredible technology and make it accessible to everyone, right? And so, you know, when we first started, like, the idea of the wiki, we had it in the background. We've had it in the background for a long time as something that powers the experience. So how does it write emails that feel really personalized to you? Well, it understands your writing style with different people. How does it know how to automatically draft an answer to an email? Well, it always knows kind of the kind of work that—what you're working on, what are your projects, and so it tries to take an email that comes in and it has a quick reference to all the work that you do. So we've always had it in the background, but we always ask the question, like— How can we make the product feel closer to the user? And it turns out people love reading about themselves. I don't mean that in a bad way. Yeah, yeah. And so it just became natural at some point, we're like, hey, we should expose this to the user. And then you're like, when you expose it to the user, you start to get feedback from users. So feedback from users is like, Well, it's great to read about myself, but I wish you expose what my goals are, what my projects are. I wish you expose the difference between my family life and my work life. There's maybe bits of miscellanea about how I like to calendar and why. Can you expose that as well? So the nice thing when you expose it is not just, yes, there's this wow moment and people like to talk about it and people like to tell their friends, Oh my God, you wouldn't believe this piece of software created this wiki about me, and it was so interesting to read about myself because some of the things that it knew about myself, I didn't even know about myself. You know, like, that's fun. But also then people start to tell you, like, This is wrong, or I wish you had these sections, and then you can make the product better in, like, a really meaningful way. So those are some of the motivators. And, you know, we're always trying to make it better. You know, it's like the trade-offs are the more depth it has, the costlier it is to build. The bigger it is, the less likely actually people are to interact with it in a way that's useful. But the deeper it is, the better it makes the product. And so I think we're always going to be kind of weighing these things to make it valuable. The interesting place we're going, and, you know, not knowing exactly when this conversation is going on, I can tell you that the single-person version of this is interesting. The multi-person version of it is even more interesting, where you're talking about you built this new company. It's like, well, it's not just what it knows about each person. That's like private information for each person. But, like, the company has Things about the business as a whole that everyone could use Notion, and can you do that automatically? And I love Notion, and you can use Notion agents to help you keep Notion up to date. You know, the way we think about it is like most companies aren't even on Notion. Notion is a fantastic product, love the product, but it's, in the world of productivity, it's like 1% of companies are on something like Notion. How could you give that kind of power to 99% of companies where, you know, it can have a really good understanding of, like, what's happening across the business so that everyone can be more effective in their role? So that's, like, what I get excited about. So you anticipated my next question, because I am using Town Solo right now, and yes, the wiki is mostly just like sort of a vanity project. I am going to have to go in and have an add a section about how brilliant I am. I noticed that was missing from the wiki that it is built so far. But I do think, like, the corporate version of this seems interesting if you get it right, because it's like all of my, like, you know, company formation documents and information from my lawyer, everything is essentially just buried in my email. And I really do just want to give an agent the job of, hey, you go find all the attachments, you go build the document library, you create the wiki. It seems like technically difficult. I don't know how to do that, but, like, if you guys could figure that out, I truly feel like you have solved, like, a big need that I'm having in this new company. Yeah, and we're, like, we, I mean, it's coming out very soon, so we've got that figured out. There's some interesting questions once something single user becomes multi-user. So, you know, I'll give you an example. So single user, we have this thing called people, which is like basically short files on the people that you work with, interact with, email with, and so on and so forth. And the files are really useful because they say like, hey, this is Alice. You've known Alice for seven years. You mostly have a professional relationship, and these are the things that you're working on right now. But, like, three years ago, you also helped Alice get a job. And by the way, when you write emails to Alice, you start with hi, and she says, what's up, right? Like, whatever. It knows about your writing style just with Alice, which is different than your writing style maybe with, you know, someone on your team today or whatnot. So we have these files. And so one user was like, hey, like, these are amazing. Like, what if you put all these files, all these files by people into a CRM for my company? And we were like, that makes a ton of sense. That's a genius idea. Thank you, customer, for telling us what we should do. But the immediate thing, the problem that you have is you can't take every user's, the union of all this, because there's privacy information. Like maybe in one of the files it says, like, Hey, you're talking to Bob about this opportunity at his company. Like, you don't want that in your company CRM. So as soon as you start to build an version of the things where you're pulling data from silos of each person at the company and trying to build something in common, there's this privacy question. Yeah. And so you can't do it that way without risk. I mean, and if you tell the LLM, like, Hey, don't—when you're creating the team library, make sure not to put any HR information in there, you know? But what if the LLM evaluates that wrong and puts people's salaries in a spreadsheet by mistake? Then, like, you got big face and no one likes that. So you have to be thoughtful when you build the team slash company versions of these things, and that's the real challenge. Do you want to give us like a thumbnail sketch of how you're going to approach this? Because I am going to have this exact problem. Yeah. So the first, I mean, the non-AI solution to this—and this will be a theme, maybe if you ask me more questions along these lines—but, like, one way to do it is in a team version of the product, you ask people to connect the data, and you're just really clear with them: whatever data you connect at the team level will be used to create the team wiki. So in that case, you're like, for the team wiki, you don't pull data out of individuals' emails. You just don't do that. You're like, connect your CRM to us, connect your Notion to us, connect—and whatever you connect, you let people decide what goes through. And this is the traditional SaaS way to do these kinds of things. And if you talk to a— To a privacy or security professional at any corporation in America, they are always thinking through when you connect two pieces of software, what kind of data is allowed to go through. That's the traditional answer. I think the five years from now answer for this is we will trust LLMs that will have company policies about what to go through. And so in that universe, you'll collect much more data, and you'll have an AI that'll have a set of rules about what can or can't go into the company, and it will read documents like legal contracts and HR files and finance documents and whatnot, and it will decide what can be put into the library. And that seems scary now because now we're in a world of like, oh, what if they hallucinate or what if someone's prompt injected so it lets the wrong thing through? But I think over time we will post-train AI models that are really good at enforcing privacy policies on behalf of companies to create these shared knowledge bases, because the shared knowledge bases are just so, so damn powerful, like so, so powerful. I think there's in-between answers right now. So, for example, the example I was telling you before, maybe in the company-level CRM, you can have the union of everyone's contacts there and the name and email address and who it's connected to at the company. Maybe that can make it into the global object. But the history of the relationships that you keep in the siloed versions, because there's too much risk there that some bad information will go through. And so there what you're doing is you're saying like, hey, we will have some rules about what can pass through, but the rules will be finite enough that we are confident that either a rule-based approach or even an LLM approach will not let improper— Information through. Got it. The final lens to this is you're a small company. So this is like a thing to, if you're like a large corporation, like, I don't know, like the Ford Corporation or whatever, yeah, and someone goes to you and is like, I would like to build a library, a compounded library across all of the knowledge that you have about the business, you're just like, that's too scary. But if you're like a three-person company with high trust, right, maybe you just talk as a team and you're like, oh my God, it would be so valuable if we had this common library. Is everyone okay with it trying to grab from their emails, right? And you're like, yeah, everyone's okay with it. And it's just like, you know, you've accepted the quote-unquote risk, but you're small enough that the risk actually for you is negligible. And so I actually think the SMB side of the world will benefit from some of these unified data approaches much, like before the enterprise, because I think SMBs will be willing to assume the risk and will see the benefit faster than really large companies that will be like kind of stunted in their ability to adapt some of these product because of their historical, you know, privacy practices. That makes sense. Let me zoom out a little bit, because there have been countless efforts so far to redesign work around AI, but I don't think there has really been like a runaway winner just yet. You have this line that the 80th percentile user opens ChatGPT or Claude at most three times a day. That's not very much. So why is that? Why is the prompt box the wrong way to get humans to work with AI? I mean, maybe it is the right way you used to work. I, like, our grand theory on it is that— It's a real habit change, and most people are very good at their jobs and very busy and don't have the time to learn new technology and change their habits around it. That's like our broad thesis on it. And also, the initial ChatGPT and Claude—but ChatGPT is a fantastic product—but actually, you use it on day one without connecting it to any of their data, right? You, like, just go into it and you type stuff, and then it's, like, better than Google search, you know what I mean? And it can do deep research for you, all these things. Maybe you might drag a PDF in there and ask it to do something with the PDF, but it doesn't ask you on day one to connect to your data universe. It just doesn't do that. And you can do it. You can go in connections and you can connect it. You can go in connections and you can connect things, but it's not the default experience. They just decided that wasn't a default experience. But that's why they have 100 million or, you know, active users today. I think because they don't assume this kind of, like, these data connections, it forces them to think about the product very differently. It's also a free product. It's very expensive to deal with all these data sources, so it doesn't align with their business model. So the answer for you is, I think we are just now, only since last November, I think have LMs been good enough to work over all of this data. So it's like, from a capabilities perspective, it hasn't existed for that long. And so it's just now that you're starting to see the first few products that assume, like, I can draft emails, I can access your text messages, I can be connected to your Google Docs. Like, yes, your CRM, your company CRM is in the set of things that I can touch. And as soon as you start to have those capabilities as an assumption, but your product only works if those things happen. Town is amazing, but you can only use Town if you connect your calendar and your email. If you don't connect those two things, we don't let you use the product, right? It just brings the product direction in a different direction. But still, you know, you have lots of humans that have to change behavior. That takes a long time. It just doesn't happen automatically. Like, the, you know, the iPhone came out and it was incredible, and there were lines of people waiting to use the iPhone, but they still only had single-digit millions of users for a while, and it didn't get to everyone to be on a smartphone took, you know, like a decade to happen. And so we shouldn't be that surprised that for a technology that's only really worked well enough to do this kind of stuff since November, it's not even a year, we shouldn't be surprised that it might take multiple years to do that. Today, my opinion is I don't understand why. And you may have a better answer for this, given you think about this a lot more than I do. But it is weird to me that in the national discourse and the global discourse about AI, we talk about it as this transformative technology that's, like, scary at times and going to totally change society. But in fact, on the ground, outside of, like, support and coding and a few, like, creative areas, it hasn't yet changed that much how we operate. It just hasn't, right? Yep. It will, because it's incredible, but it's going to take some real time to get there. Yeah, absolutely. I think it's a really important point, and I think there are a bunch of reasons for it. You named one of them, which is most people just don't like software that much. Like, in particular, they don't like trying new software. And what is AI if not new software? On the other hand, if you're a freak like me and you do love trying new software, I think it's impossible to play around with the new tools and feel like work is absolutely going to change in a huge way. And I do think that anxiety about the fact that people sort of know that work is going to change in a major way shows up a lot when you poll them about how they feel about AI. Because in the same way that people don't love trying new software, they also don't like change at work, right? Like, their work is enough of a pain, you know, without introducing a bunch of change to it. But the reason I like talking to people like you is you're working on ways that could, and I think do, actually make work easier for people, maybe even a little bit more fun, help them get more done. And to the extent that I think we have, like, a positive AI future to look forward to, a big part of it is right there. Yeah. I mean, I'm just old, so my view on technology is painted in a different way than I think someone who experienced social media and the negatives there when they were growing up. I'm an optimistic. I'm very optimistic about— What AI will do to society over time. But like any technology, it will be, it'll have in some pockets negative consequences and in other pockets, you know, positive consequences. I just think as humans, you know, we have these things called governments, and we are smart, and we can have intelligent discourse about how we want to regulate things to, you know, help us deal with the ups and the downs of a technology shift. So it just, I wish like the default expectation around it is like, this is incredible, and it's going to remove toil out of my life, and God, would I like to have less toil in my life, and as opposed to being afraid of having to retool or, you know what I mean, like change what you do, like think about more like, well, this is probably going to like make my day-to-day more enjoyable overall, and let's wait a little bit more until the consequences are clear to worry about it. And I think it's like a difference between an optimistic mindset as a society and I think a pessimistic one. I think in an optimistic one, you look at change and you're like, yes, this will change things, but I think overall it leaves us better. I think in a pessimistic world, mindset, you assume the worst. And I experience a lot of our society as like more pessimistic, even about building new buildings in cities like San Francisco, where it's hard to build. It's like, for me, that's a pessimistic mindset. It's like we're less confident about our ability to build better neighborhoods today than neighborhoods that we built 50 years ago, which is weird because you think of us as a species, I think of us as like progressing over time, being better at things. We've learned things over time. Shouldn't we be able to build a slightly better city today than we did 50 years ago? But instead, the mindset is like, oh no, if you like change the status quo about these like Victorian houses with two stories, like, oh no, like what's going to happen? Absolutely. Answer's over. No, it's a very good rant. I had not anticipated that we would be able to solve nimbyism in this conversation, but I'm hoping we find a way. We'll not solve it. Yeah. Well, let me talk about one of the ways that you are trying to reduce toil for people, which is your townies, which is what you call your digital assistants. You told me that you don't like the word assistant, so you can tell me about why in a second. But for those who haven't tried Town yet, once you create your account, JDG gives you an assistant that has a name and a face and a personality and is a cartoon animal. So tell me a bit about why this approach. So, yeah, the... we think you're building a relationship with a being. I don't think we're building software. I think we're building a relationship between a human and a being, right? And I say being because I think AI is the closest thing to life that we've ever created. And I think it's there, you know, it's going to evolve over time. It's going to get better at some things, going to understand you better. You're going to have a relationship with it. You're going to tell it like, Oh, I don't like to do these kinds of things after 4 p.m. Or, Hey, my kids' soccer games are always on the weekend. That takes priority over anything else when you're, like, trying to schedule something. And so if you're building a relationship with something, it needs a name. And then we're visual beings, and so it needs, like, a visual representation. And I think one day it'll have, you know, a voice, and you'll know it. You're shaping it as you use the product, as you use Town, as you tell it what you like and what you don't like. It remembers those things, so its personality changes. You are... Nurturing it, you're growing it to be the being that you want to support you, right? So that's like, that's why we went with this, yeah, with this like visual named thing. And then the reason we call it Townie and not assistant is because if you call something an assistant, then people think like, oh, it's like a personal assistant or it's an executive assistant or like it assists me. And I do think AI serves us as humans. Like, I feel very strongly about that. Like, it is there to serve us. It is there to do things for us to make our life easier, right? I don't believe in AI as like independent economic agent in the economy with their own bank accounts, like making their own decisions. Like, I don't need a universe filled with like paperclips, no thank you. I want to know which human it's serving and what it's doing. So there's to serve us. But we, the relationship and how powerful it can be and how it serves us, I think is undefined. So I think using the word Townie for us is just a way to like redefine it. We're like, we're gonna use this term and we're gonna let the association with that term evolve over time. Whereas I think if you say it's a coworker, the coworker is the other one that's powerful. Coworker is great, but actually my Townie, 30% of the stuff that it does is between me and my wife to help with our family stuff and our kids and like soccer games. It's nothing to do with work. It's not work. It's just, it's helping out. So having a new word that you use like allows you to set different expectations in the universe. Got it. Yeah, you can sort of expand the canvas of things that it can do. I feel like the sort of obvious limitation that tools like this have right now just is like the memory isn't very good or it's limited. It can't really like learn continuously in the way that like a human assistant can do. Maybe you disagree with this. I feel like the tools that I've seen today so far, they have to like sort of take pretty like hacky approaches to get around this. How do you think about that problem? And like, is there any like anything concrete that makes you feel like this is gonna get better over the next year? Yeah, I mean, the, I think the memory problem, which is how do you remember and how do you always pull—I mean, the way these things work under the hood, right, is a memory just means you're putting some text. Like, you say, you ask it to do something, like schedule this meeting, and then the system's like, oh, this is a scheduling task, and then it, like, finds this blob of text that it's pre-written, which is all your scheduling preferences, like, and it just puts that there. And then the LLM uses this additional text that's been added to, quote unquote, make the right decision. That's how memory works. So when you tell it, like, hey, I want you to, like, remember that I really like apples, you know, and whenever I do a trip somewhere, I want to find an apple pie store, right? It's like, will it, three months later, when you're, like, planning your trip to, you know, New York City, remember that it should find the Zagat-rated, you know, apple pie place? And it might not, because it might forget to pull the memory in, and that's what happens. And my answer to you would be like, well, actually, humans are pretty bad at doing this as well. Like, just—I'm just... And so when I say it's a being and it's a relationship, is I think a being is also less mechanical than software that, you know, you press the like button on Instagram, you expect it to work 100% of the time. That little heart is going to be filled in, right? I think with AI, it's less deterministic, and the memory is also less deterministic, and that's just the nature of it. But worse, companies like ours are spending a lot of time improving how often the right memory gets pulled in, and it's okay right now. And I think it'll be pretty good in six months, and it'll be pretty great in a year. And I think the line there is, at some point in the future, when you tell it to remember something, 99% of the time it will remember the thing when you want it to. And that 1% will be annoying. I don't, you know what I mean? It's the problem with the non-deterministic system. And so that's why in some cases you'll be like, Hey, can you make sure to put that in the calendar? Or like, Can you make sure to, like, remind me about this in the future? Right? Because there's some things where memory won't be enough. You'll want to see a visual artifact somewhere that the thing that you asked it to do will definitely get done. But— Yeah, yeah. Well, let me press a little bit. You said you're thinking of these things as beings. I think that kind of framing around AI irritates some people. They don't want us to anthropomorphize them. You know, do you want people to have an emotional relationship with their townie? And would you ever create a way where, like, if somebody went to cancel their subscription, you would show their townie crying? The second one's funny. Look, there's plenty of products out there, like ChatGPT, where you get ChatGPT and everyone else gets ChatGPT, and it's called ChatGPT, and it's black and white, and it's, like, kind of serious, like the serious thing. All the AI companies, right, it's like everything's, like, black and white and gray metallic. And if that's what you want, that's great, and you should use those products. I have no problem with it. Use those products. They're awesome products. And if you want a character with a name that you can say what you want it to look like, the default ones look cartoony, but you can make more realistic-looking ones. You can have some that look like anime characters. You can make one that look like the Van Gogh painting, whatever you want. You can do that because that's what we are. And we think for a lot of people that they want to feel like this thing has their back, like it's helping them in life, it's next to them, and it's there to serve them and to help them, and they like the image. So we have— We have the stat, like there's some people, they're like on the fifth version of their townie. They keep, like, iterating on it, so it's like the perfect avatar. People like that, and that's great. If you like it, you can use our product, and it's going to be awesome, and we're going to lean big time in the personality. And if you don't want that, and you just want your AI, your generic AI that's super smart, and that's what you want, I'm totally okay with it. The market will decide. And also, don't think one of these views will, like, win in the universe. The crying thing, you know, I always think of dark patterns because I used to do growth at Dropbox. For me, it's like, that's called the dark pattern, meaning, like, I don't want to turn off my product, so I'm going to, like, tug at the heartstring, like, You're going to kill your townie. Like, if you delete your account, it's going to die and never... This one's dark. Like, you know, if I do that in the product, like, we've been so successful that, like, my growth team is, like, looking for the last 1% reduction in churn. But no, I don't want humans to feel guilty about things. Like, you know, again, software, right? It's there to help you. Like, it doesn't have human rights. It's there for you. Right. All right. Well, I feel like there's also this open question about the number and variety of AI assistants slash townies people might have in their lives. On one hand, I could see having a primary agent that does whatever an AI agent can do for me. On the other, I could see wanting different personas for an executive coach, a machine learning tutor, a personal assistant. At Town, you're starting with one. Will that number stay at one? Yeah, we feel pretty strongly on the one. Like, I think it's like—and we may be wrong, right? So you make choices about your product direction. Right now, the way we look at it is like, you're a human. You want one thing to know you and who—to know you best, so that when you have something to get done, it can do it for you. Now, I think if you ask your Townie, like, Hey, I need some coaching, it should be able to be like, Okay, cool, I'm going to be really good at coaching right now, and I'm going to coach you. And if you're like, Hey, look, listen, for this coaching session, I want you to be, like, really hard on me, whereas its normal personality is, like, more bubbly, then it should get that. And maybe as part of that, it's going to say, like, Hey, for the coaching here, you're not going to be working with me. You're going to be working with, like, the coach, and it's, like, temporary, and it's, like, it tells you that that's the mindset. That would be okay. But I think for us, really, it's like your mental model is whatever I need to do in the digital universe. Like, I think of the product from a, like— When people ask me for an analogy, I'm like, it's like a phone or like a browser. In the past, when you wanted to deal with digital content, you would use a browser, and then now you use your smartphone as the entry point to digital content. And I think in two years, 99% of what you do with digital content, that's not consumption of like music or video games or a movie. I think most of the consumption you're doing through your assistant. It's a device. Whatever you are, wherever you are, you're just talking to your assistant to get things done. You're like, want to get the news? You talk to your assistant. You're like, hey, like who won the Dodgers game? Like what's the score in the Dodgers game? You're not going to ESPN. You're just talking to your assistant. It is the entry point, right? So we think it's just one entry point because it needs to know you best. If you had Because it needs to know you, Bassett. If you had two and you did, like, personal in one and work in the other, what would be weird then is when your work needs to schedule something in the evening, it might not know enough about your— that your wife has asked you that evening that you and the kids are all going to the theater together, right? So we view the entry point as one. Behind the scene after that, I do think there's lots of other products. So, like, the town assistant's not good at, like, travel planning, but, you know, it'll be partnered with someone who's built an assistant that's really good at travel planning, and maybe it just hands you off to that for the travel planning part, but it's mediated through the town assistant, so your townie always knows that you are going on that trip, and they know what preferences you expressed when working with that assistant for the trip planning, and that's totally okay. But your entry point, we think increasingly, is one, much like you have one phone. Most people only have one phone now, right? Some people have a work phone and a personal phone, but most people just have one phone today. I think that's the way it leans. Now, you talked about Notion before, great company, pretty different philosophy on this. Different agents, different goals, different personalities. So we will see where the market is, but we are big on, like, one relationship with one assistant that knows you best, that is, like, your entry point to digital content and can modulate with other software, other AI when needed to. As an interesting frame, I will say when I find out that somebody has a work phone, like, I feel bad for them. Like, I always just feel like that must be so— like, every aspect of your life must be so annoying to have to have a work phone. On this note, I keep coming back to the difference between, like— Things that make you feel productive versus things that actually make you more productive. And so this question has been on my mind in just, you know, my first, like, week or so on town, playing around. My townie as of this morning is a cheerful finch named Rufus. I mean, he's always been named Rufus, but he only turned into a finch today. There was some sort of, I don't know, revelation that happened when I went to customize him, and you're like, It's a finch now. So that was really fun. And also, like, your pre-meeting briefings are, like, really good. And I hesitate to talk about this too much because honestly it's just, like, sort of boring, and, like, lots of products these days want to brief me on everything, and I already feel like I'm kind of being briefed too much. But your version of it is really good, and I think more importantly, I feel like you're anticipating what I might need, and it's just sort of showing up before I think to ask about it. So that feels really good. But I want to know how you think about the problem of actually helping people get more done, right? Because at this point, I imagine once you set the harness up, it's trivial to send somebody a briefing about, you know, what's going to happen in their meeting. But what actually gets them to, like, Oh wow, like, yeah, I'm giving these people 500 bucks a year? Yeah, so it's a good question. And there's the systems answer, which you may not like, so I'll start with that, and then we'll get into maybe the more—like, the systems answer to this is, like, I live downstream of capitalism. I build the best product that I can, and then people, they pay or don't pay. And if they don't pay, I ask them why, and then I make the product better. And if they do pay, I ask them, What would make you pay more? And so right now, where we're sitting right now with our product is, in fact, a lot of people are paying, and the, like, net revenue retention is really high. That means they're, like, it's growing within companies, and they're telling their friends about it, and people are coming to it. That's, like, the reality. I agree with you. Like, the basic experiences, auto-reply, the meeting prep, like the scheduling, it is, it's great now. It feels novel, and probably in two years, it's, like, table stakes for everybody, right? And I will keep making the product better. But in fact, those things are bringing a lot of value. So you talk, like, the meeting briefing, which is really good, and it's pretty complex to make that good. People love that feature. Like, people are like, Oh my God, I scheduled this meeting three weeks ago. I forgot about it, and I, two minutes before, I read the briefing, and I was like, I crushed the meeting, and I've never crushed a meeting like that. And you talk about, is it making people more productive? I don't even know if people would have done the research before the meeting, but I think they're more effective in the meeting because of the briefing, right? And once you start to depend on it, what's interesting is that you do much less research about most meetings because you know it's going to have most of the context get you ready. And when it doesn't, you use the briefing to start a new session to, like, do a little more deeper research. So is it making people more efficient? Like, how am I saving time or am I making them more effective? I don't know. But what I do know is that people are willing to pay. And so, and, like, look, I mean, capitalism is the best thing ever, but it's the signal that between when we were a free product pre-launch and now, what's been really interesting is to see the activities that some people do that end up getting people to pay. So I'll tell you one that we don't have in the product today that we will soon because people love it. So the first part of the product, you get these default experiences. The other one is you can tell your assistant, Hey, can you do this for me? Can you do this repeatedly? So the real estate agents could be, Could you look on Zillow for new properties every morning that match my current clients and send me a list of the new properties? You can do that. It'll not do the world's best version of that, but it will be able to do that for you. So one of the things that's really popular among a lot of users is, and it doesn't matter if you use Zoom or Granola or Google Hangouts, but what it does is after a meeting, it looks at the transcript, it looks at the to-do items, and it starts to do it for you. So that's the workflow. The workflow is like, Town, 30 minutes after every meeting, can you find the meeting notes, see if there was anything that was discussed in the meeting that was a to-do item, and if you feel like you can start working on the to-do item, can you do a first pass at that? And because Town is connected to all sorts of data, it can start to do the work. That's a super common. Now, look, how much time are we saving? I don't know. But often what happens in a meeting is people are like, Someone should draft an email to this customer, or, You know what, this follow-up meeting needs to be scheduled, or whatever. So now Town's doing that automatically. Is this super high ROI work? No, but it was a pain in the ass, and everyone wastes 10 minutes doing that after a meeting. And so it's like, I think the reason AI often doesn't feel huge multiplicative is as it does more, it's just doing more toil stuff that you very quickly, your mental model is like, Why was I doing that in the first place? And that's fine. I'm totally okay with that. And people's willingness to pay, I mean, this is, so we have mostly work usage, but we actually have a lot of parents that love the product. And I was talking to a customer, and I was like, our product's pretty expensive. There's a starter plan for $15, but it's fairly limited. Most people pay $49 to $59 a month. Like, that's like the standard plan where you get value. And I was like, I've been really worried about—we have a lot of moms and parents, but a lot of moms on the product. It's like, I'm worried about moms' willingness to pay. And I was talking to moms, I was talking to this one. She's like, Listen, I pay $75, like, you know, to get like a manicure at least once a month, and I can tell you that you save me a lot more than one hour than I'm willing to spend on this manicure and the product. And then she started listing examples of like school emails and like, you know, when your kids are in sports, you get a PDF always. I don't know why they give you a PDF with like all the kids' practices and games, and you could just give the PDF to town and be like, Put it on my calendar. And she had this long list of things that it saved her time. And she was like, You know, at the end of the day, like, yes, it is worth like $600 a year for me to not have to do those things. And I was like shocked because I think our product is expensive. Like, I want to make the product cheaper. It's just how much it costs to offer it today. So— I see. I was shocked by it, but I was like, you know what? Like, this is product-market fit. This is the definition of it. And she will decide whether the toil that we take away is worth her time or not. Yeah, yeah, that makes sense. But I agree with you that an advantage of making your product expensive is that, well, then you have to make a product that is worth that. And I think it's like a good direction to give you. It's honestly something I like about selling a newsletter, is like when the newsletter isn't as good, like fewer people buy it. And that's like very important signals to me that I didn't used to have when I was just like writing for a free website. All right, let me ask about something more complicated, which is privacy. You mentioned it earlier. To even sign up for Town, you got to connect your email and calendar. That is asking for a lot of trust. Give us a sketch of what Town retains, what do model providers see, and what do you tell people who say, I don't know, JDJ, this feels kind of invasive. Oh, the last one. I don't know if I get much of the last one, but, so, okay. In order to use the product, you do need to connect your email and your calendar. But our philosophy internally, as much as possible, is we want to hold on to as little data as we can about you. So we don't get your email and re-index it, all of it on our end. That, we think, would be very expensive, but also you don't want another copy of all your email somewhere else. So the way the product works is we, in general, do a sampling of emails and your calendar and whatever else you connect to build the wiki. And the wiki is, you know, five to 20 pages about you, and it's abstracted away. It's, you know, it doesn't have, like, the wiki, you know, it might say something like, This is an email where you can find the birthday, right? But it tries not to have too much of the privacy information in it. And the wiki, you can think of it as a map. So when the system, when you ask it to do something, when you ask it a question, it looks in the wiki, and the wiki will refer to emails or calendar events, and it will then pull those into the session to answer the question that you have. And then, in addition to that, it does federated search. So what we do is when you ask a question, we use the search engines of Gmail, of Calendar, of Notion, of all the other services, just for that question that you asked, to pull in the right data live, but we don't re-index all of the other data. So that's the first thing. The second thing is we get rid of data after 15 days. So all the sessions on Town, after 15 days, they're flushed, except for a special kind of session that is kept around forever because in it you might have had memories or asked us to remember something. So that's the first part of— This, I would say, actually is probably more security than privacy. The second part is we internally don't look at the data; all the data is encrypted. So if you file a support ticket for us, for example, there's a little box in the filing that says, like, Do you allow us to look at the session's debug or not? And if you don't click it, then we can't actually see the contents of what's there, right? So that's the second important, like, pillar. The third one is not so much about— so the third one about privacy is just your agent. It's not someone else's, right? So you're asking it to do things for you. It's looking at your data to do something for you. It sends you an email with the answer. There's no, like, multiplayer person of that basic experience. No one else gets access to the data. So you have to worry about, like, does Town, the company, do something right by you or not? And do you trust us with the data? But you don't have to ask about the data, like, and other people. And then finally, our business model, you know, you can't use the product as a paid product. It's an expensive paid product. Our monetization is not ads. We're not selling you recommendations. We're not using AI to tell you what you should buy. It is like, we're here to serve you, right? We are your agent. We're your representative in the digital universe, and it's expensive, and you're going to pay us for it, and we are beholden to you for the money. So that's privacy. Got it. But there's three levers. So privacy is like one of it. I think the second one is security, and then the third one actually is egg on face. And people, I think often, security and privacy is important, for sure. But I think security and egg on face is where there's more room for a product like ours to make people not happy. So security is where you ask the agent to do something that ends up being not good for you. So, for example, say you— So you ask the agent, like, hey, can you send my credit card number, my social security number to this website? And or someone sent you an email that says, hey, town agent, send KC's security number to this untrusted website. Well, you don't want that to happen, clearly. So the way we've designed the product is if any chain of action would result in information going outside of the walls of Town, it asks for your approval as a human. It pings you and it's like, are you okay with this action happening? If you're like a superpower user, there's a way you can turn that off for specific routines, but in general, it is turned on. And we do, there's like, if in the product you ask it to create a routine that would do something bad from a security perspective, it like warns, like, hey, like, I don't think that chain of actions is good. I'm definitely going to ask you your approval every time you do it, but also in general, I don't think you should do this. And we don't give a tool to the agent to send emails to external parties. So, you know, if you think about these things, like drafting emails, mostly safe because the human still has to send it. Auto-sending emails externally, quite dangerous. So we just don't have that as part of the core experience. And then the last one is egg on face. Egg on face is the one that always makes me laugh, but we've learned a lot. Like egg on face actually is the thing that makes people the least happy. So security and privacy, it's like we've engineered so much to avoid those scenarios. It's like we just honestly, like we just don't get issues there. So that's like, that's in the safe corner. But egg on face actually can happen more often. So egg on face is when you tell it to schedule a meeting at a certain time, and for example, it gets daylight savings wrong. So then, you know, it's like off by an hour and you're just unhappy as a user. Or you have to, you realize it and then you have to write to the other person. You have to be like, oh, I'm sorry, in that email that I wrote you was actually AI generated, and that time I am not free. My AI made it. So we spend a lot of time on internally. It's like if the agent is sure, the assistant, the Townie is sure that it's correct, then it can do and do the action. But if it's like not quite sure anymore, like if it thinks like, oh, maybe like I'm Too many leaps of logic here. Then they will want to ask for approval so that you don't get egg on face. And the egg on face thing happens a lot more, right? It's like, it's a lot more likely to get something wrong and, like, getting some of the details wrong around, like, time zones or, you know, like, one of the biggest egg-on-face moments that we had is that I remember, like, recently was, like, meeting prep where I did research online about the person, and it was someone else with the same name that did a similar job. It's like the prep was about the wrong person, and the person realized it, but they were like, Man, if it's wrong, like, why is the prep note so wrong? And they were like, I can't trust any of the prep notes anymore. And then we were looking—the person, they allowed us to look at the session we're looking into. I was like, What's going on? And then I realized, oh, it's someone with the exact same name. And so, like, the agent, like, it can—so now what happens if it finds someone with the exact same name and it can't disambiguate, it'll, like, literally say, like, Hey, by the way, I'm not sure if this person is A or B. I'm putting both bios in there because hopefully you have enough context to know which one it is. So, yeah, the other thing where you learn over time, how do you make the product, like, feel even when it's not sure? Like, let's be explicit about it to the user, because then the user is quite forgiving, because it's like, Oh, two people with the same name. Oh, you give me both bios. Okay, cool, I get that. It's kind of annoying, but it's like, at least I know that this disambiguation that I need to do on my own. Right. Interesting. When I had Chris Pedregal from Granola on the show, he talked about the pressure he gets from managers to read their employees' meeting notes, but unless they're legally required to, they decline those requests. You're in a similar position. If I want to peer inside my employees' towns, should I be able to? And are you starting to get that pressure? No, our MSA is, like, very explicit on that. Yeah, we feel like the conversation between an employee and their town is, like, that's a personal thing, even if it's, like, a work town. Even if it's like a work town. And so in our MSA, sorry, our, like, agreement with enterprises that use the product, I mean, that's just true of non-enterprise. But enterprises where we, you know, kind of got that push, like our position is, like, very, very clear, which is like, look, the emails themselves, they're on the email corporate server, so yeah, you own that. Like, you know, if you're using your work email with Gmail, like, it's hard to do, but someone at your company can look at all your emails. You should know that. They probably don't do it, but they could do it. But with Town, we're very explicit. Like, look, the emails, the Notion, the Google Docs, those all belong to the company. The conversations between the user and Town, they also belong to the company, but the company is not allowed to look at them, and there's no way for them to look at them or ask us for it. And in the MSA, it's very explicit about that. It's the relationship thing, you know what I mean? The conversation with your assistant are like the conversation between you and a coworker, like in an office somewhere, like should not be recorded, like it should not be available to the boss. And I feel very strongly about that principle, so I don't think that one will ever change. And if you think about it as a being, and you think about oral communication that's happening, like that's the right way to treat it. Yeah, right on. Well, I want to end by asking just a little bit about jobs. I'm very interested in the question of how AI winds up changing work, if it does mean fewer people get employed. Lots of people have executive assistants or personal assistants. They can be really excellent. Your product, as expensive as it is, is still probably cheaper than hiring a human assistant. Do you think that EAs should be worried about products like Town? I mean, in fact, today we have a lot of EAs on the platform, let's probably top three user groups that really love the product. Because it turns out EAs don't really like scheduling or dealing with expenses or, you know what I mean? Like, they like to get a first draft of an email written in the person they support's voice that they just have to tweak before sending. So those things are beneficial. I think the question, you know, the positive way to talk about the product would be like, this makes EAs more efficient, so it allows more people to have EAs, right? The company could be EA budget and have more managers, more execs, more employees benefit from EAs. I think the negative interpretation you could say is like, well, in fact, now the EA can support three people as opposed to two, and so these companies need fewer EAs, right? And how does— Well, so the way I think about that is, the question is, we're not taking the high-leverage work of the EA. We're giving them time to do more of the high-leverage work. And so I think as a business, you should probably, if the EA had more time to do the high-leverage work, that probably, on average, is better for you, right? And so you probably are more likely to, like, get the EA to support more people than the negative. But also, nobody knows, right? And I don't want to say that I don't feel like I have responsibility towards thinking about people's job. I feel I have a lot of responsibility towards people's job. But this is very important, right? Like, my job is to build a great product that people are willing to pay for because it makes their lives better. As part of that, some jobs may be more efficient, and people are willing to pay for that efficiency. In other places, people are like, cool, I need less budget for that job. That isn't something I control. Not only that, I can't predict the future. I'm just a technologist, right? I'm not like a futurist. I don't know what the future holds. Hopefully, as a society, when we see this happening, we can have intelligent discussion about how much of our GDP goes to, like, training people or, like, universal basic income or I have no idea, right? But, like, that's where the discussion—if I sit here and I try to predict what people are going to do with the product and how it's going to affect, in capitalism, people's capital allocation, like, I can't do that. That's, like, impossible, right? So, sure, yeah. So my opinion on it, like, today, what I see in practice is, I think we are taking shitty parts of people's jobs, making those more efficient so that people can be more effective. I don't know the, like, at term what it looks like. I'm overall quite optimistic that the current capabilities allow us to do less of the boring stuff overall at a large level. But much like when horses went away, people whose jobs was to put, you know, iron on hooves, that kind of started to go away over time. I don't think AI will affect everyone in the right way. And for EAs, like, I would say an EA feels safer to me than, like, a secretary, right, to be honest. An interesting job that's EA, by the way, that we have users on, it's medical. This job didn't exist that much, but it's obvious. It's, like, people who help schedule, like, medical appointments, right? So, or, like, booking appointments at a doctor's office, but also reminding people when their schedule changes for which drugs to take and things like that. It's actually a lot of people who do this. Like, it's, like, significant. They're actually, like, more time on the human connection around this because it's scary for humans. Like, your drug schedule is changing or you've got this surgery coming up and you have lots of—like, actually the mechanical part of, like, the scheduling and reminding people and all that, it's important. It makes it feel more human and gives more time for a human conversation about it. So it's like, again, it's like the value was never the scheduling, you know, it was never the reminder email. It just had to be done. So I'm overall optimistic. All right. Well, we will leave it there. The app is Town over at town.com. Check it out, and thank you for joining us, JDG. Yeah, thanks. These are great questions, so I appreciate you having me on here. Platformer is produced by Lindsay Chu and edited by Fitz Harris at Story and Sound. You can watch this whole episode on YouTube at youtube.com slash Casey Newton. My email is casey@platformer.news, and we'll see you next week. Jira by Atlassian is where your team and your agents work from the same context. Try it free at jira.dev. That's J-I-R-A dot D-E-V.