The a16z Show - Building Agents at Home: Parenting, Work, and Benevolent Neglect

Episode Date: April 13, 2026

Katherine Boyle and Sarah Wang speak with Jesse Genet, a startup founder and family builder, about building 11 AI agents while homeschooling four young children. Jesse runs agents across roles ranging from coding to curriculum planning to household management, and she shares how agent architecture, logging systems, and “benevolent neglect” parenting have changed her life as both a founder and a mother.   Resources: Follow Jesse Genet on X: https://twitter.com/jessegenet Follow Katherine Boyle on X: https://twitter.com/KTmBoyle Follow Sarah Wang on X: https://twitter.com/sarahdingwang Stay Updated:Find a16z on YouTube: YouTubeFind a16z on XFind a16z on LinkedInListen to the a16z Show on SpotifyListen to the a16z Show on Apple PodcastsFollow our host: https://twitter.com/eriktorenberg Please note that the content here is for informational purposes only; should NOT be taken as legal, business, tax, or investment advice or be used to evaluate any investment or security; and is not directed at any investors or potential investors in any a16z fund. a16z and its affiliates may maintain investments in the companies discussed. For more details please see a16z.com/disclosures. Hosted by Simplecast, an AdsWizz company. See pcm.adswizz.com for information about our collection and use of personal data for advertising.

I was resigned to not challenging myself to build technical or hard things for like the next five years or so. I really want to be present with my kids. I need to take this break, basically. That is no longer true. A weird superpower of mine is just how incredibly motivated I am for agents to do work for me. I got my agents to learn how to build other agents on their own. So I could be like, we need another agent, you guys, and they actually can spin them up without me touching the machine, which is a little crazy. But the first few weeks were very right. off, it would be a level of pain that I wouldn't want an average person to go through. But the thing is, in the 1950s, labor-saving appliances were supposed to give mothers more free time.

Instead, standards rose to fill the gap. Laundry went from weekly to daily. Meals became more elaborate. The hours stayed the same. Seven decades later, a former YC founder, homeschooling four children under six, decided to test whether AI agents could actually break that cycle. She built 11 agents, each with a distinct role, from lesson planning to grocery ordering to logging her children's progress via voice notes. Her agents now build other agents without her touching the machine.

The result is not a frictionless home. It is a home where a parent can spend two uninterrupted hours ignoring her children on purpose, because she believes boredom is a skill worth teaching. Catherine Boyle and Sarah Wang speak with Jesse Jeney. Startup Builder turned family builder. Jesse, it is so fantastic to have you here. I think you've been what I would call a viral sensation on X, posting videos of how you're homeschooling your family, four children under the age of five, which all I can say is God bless you. You're amazing.

But we also want your secrets. We want your tips. Sarah and I are both moms of young children. And we talk a lot about how AI is impacting education, how AI is impacting the future of the family. And you've become just such an incredible force with your vision. videos on X of how you're using it in a bunch of different tasks around the house and a bunch of different tasks around supporting your family as a homeschool mom. So we want to start with who you are and how you got so interested in using AI for homeschool. But tell us about your

previous career, too, as a Silicon Valley founder. Yeah. So I've started a company many years ago now. Time flies, but I was a YC founder. I did a venture back company kind of full cycle. I ended up selling it a few years ago. And so I do, in the one hand, I'd say I have a technical background. On the other hand, I would admit openly that my co-founder was the technical co-founder. So I want to be like, I want people to understand, yes, I've been swimming in these waters. I've sat in many an engineering meeting where I was sort of following along and sort of lost. I've sat in many product cycles and reviews. So it gives me a vocabulary, but I had an opened terminal to try to build something myself until maybe six months ago. So I think we're living through a really

fascinating time where only recently after running a company myself did I feel like now the tools are so good that I can really use natural language to build things. And so the last six months have been like a Cambrian like explosion for me of building. And of course the last few months where we have the open claw, then I went to completely obsessed. So I'm happy to discuss that. But I went down a complete obsession. It can only be described as an obsession because I've just been building almost nonstop. But when I say that, on a data basis, I'm actually spending a lot of time with my kids. And so I was trying to find, like, how can I build things that are relevant to my life? So I know we're going to dig into that, but that's a little bit of how I got to now.

And maybe talk about that six months, like, what was the thing that happened six months ago? Or do you remember what the moment was where you're like, I need to start building to fix this problem? Or what was the story behind that? Well, my co-founder from Lumi, which was a packaging company. So little physical packaging. We managed a packaging marketplace. He's now off running something called Obsidian, and it's a markdown note-taking app. And I say this because I follow him on Twitter.

Obviously, we're a co-founder. And then I follow all these Obsidian geeks on Twitter. And I started noticing a change in the conversation, changing the conversation, them talking about how they were, like, building really wild things with Cloud Code. That stuff was referenced. The discussion about interesting ways they started using Obsidian. And what really caught my eye, the six months ago was me feeling, like, hey, I actually feel like I can probably be building things myself now in the small bits of

time that I have. So I have confetti time. Like I have 10 minutes here, 15 minutes there. And I started feeling like maybe I can build stuff. The tools are getting so good. But then about two, three months ago, I saw people saying, I'm using obsidian as a second brain for this thing. And it was called Claudebot and then all these different names. Yeah, Claudebott. And they were referencing this. And I was like, what are they talking about? This was December and into January. And that's when I realize I can build agents who actually code for me while I'm hanging out with my kids. That was a complete game changer. Actually, just pausing that for one moment, this is such a huge deal for me. I was resigned and not in the super depressing way, but just being really blunt, to not challenging myself to

build technical or hard things for like the next five years or so. I was like, I really want to be present with my kids. We're doing homeschooling, which is like a wild choice. And so I was kind of, yeah, I think the right word is resigned to it, not sad, not resentful, but just like, okay, I need to take this break, basically. That is no longer true. What happened a few months ago is that is no longer true. I feel like I've been building better things than I ever have before while I spend almost all of my waking hours, like with my children. I can explain how I do the flow of the day where I do that, where there's things I do during the day and things I do at night when they're asleep and stuff, but I'm like truly building things that I'm impressed with personally and I'm being an active mom.

And that was not possible a few months ago. Like it's actually a sea change. For me personally, it's like really liberating. So that's incredibly inspiring what you just said. And I think a lot of parents listening to this are like, I don't have time to build this, right? And so you're a mom four, five and under. I have two and I barely feel like it could breathe little and four. You're doing homeschooling.

Can you walk us through a day in your life? Like, how do the hours stack up? And then when do you build? And to your point, the big unlock is like, you're sleeping in their building for you. So, like, how did you get that set up? So many questions. Yeah, an average day, a typical day,

wake up at the crack of dawn, right? Because there's a small person who has decided that that's when we get up. And they're like, you get up. So anyway, I'm, like, waking up and there's, like, small little gremlin creatures around my bed. So that's where we start.

We go from there. Obviously, all the basics, like having breakfast, yada, there's three kids I'm really homeschooling now because one is a baby. One is about four months old. I start early, but not that early. The three are five, four, and two, the three other children. And I try to do individual sessions with them. So imagine after breakfast and these types of things, I have a place where we homeschool and I cycle the kids in one at a time.

And I need help with the kids even to do that, right? So I am lucky enough to have some help with the kids during certain portions of the day. So I cycle the kids in one at a time to where we homeschool. And I do a one-on-one session with them. You know, depends on the kids' mood. They're all quite young. It can be anywhere from like 20 minutes to an hour. And then after that, maybe then it's like mid-morning,

we will just do some more unstructured activities like playing and playing outdoors, or trying to pull on a thread of something we're doing where we leave the house maybe going kind of a field trip or something like this. One time a week, we do a homeschool pod with two other families. Between the three families, there's 11 kids already. When you meet homeschoolers, these people are reproducing, all right? So three families, 11 kids already. And so once a week, I lead a science pod.

So on that day, it's really kind of cool. All the kids are at our house and they're like running around and we do a science lesson that we weave through like the whole day. But in any case, the kids can do effectively 30, 45 minutes of like active instruction per day. And you really want to make the most of that. And then the rest of the day is pretty like thematic. The other thing that I really believe in, so I spend time doing this is you could call it free-range parenting.

You could call it benevolent neglect. I don't know what you want to call it. But I try to ignore the children. I try to make sure that they're going to survive the ignoring. So they're set up in little places where they can't hurt themselves. But I step away from them and try to just see what they do. There's already three of them, even if we don't have the other family over. I try to build up the amount of time that they can spend playing together without needing me.

So instead of structuring their whole day, we're up to with a four and five-year-old, they will spend more than two hours interacting and doing stuff. before they come back to me. Wow. And I actually use a timer because I'm like paying attention, but I actually use a timer and like I'm trying to build up their tolerance before they're like, I need a snack or whatever. And they have snacks and they have stuff they can grab.

But the trick for me is like when do they actually truly come to me and they say, I need, I need something, I need activity, this or that. But this is part of also why I want to homeschool, frankly, because I want to benevolently neglect my own children. They don't get the proper kind of neglect in every school environment. I want them to learn how to not be bored on their own and these types of thing. So a portion every day is them away. During that time, I do get some magical possible coding and tech time.

But anyway, there's a portion every day where I'm like intentionally trying to ignore the children. Not the four-month-old, all right? No, that's amazing. My five-year-old after like two minutes is like, I'm bored. And I'm trying to create that mental resilience of like you don't need to be stimulated all the time. It's really hard. And that's why I'm talking about building up. It's like a tolerance. Like I would say we started at like five minutes. And the trick is that I try to not say like don't talk to me. Like I never actually vocalize that. I just remove myself and like go away. There's a couple places that are great where we live where I can go away. And then my mom also lives with us in like a little kind of mother-in-law suite, like a separate little building. And so the kids will like wander farther and farther from me. And sometimes all like it's almost like we got a walkie-talkie sister. I'm like they're near. you now. Like, and then, and then so like, but they will actually wander away, like, like

totally away. And even the three old, but they'll stay together. So, so they'll stay together. Anyway, we can get into my neglect strategies if you, if you like, but it doesn't really, it doesn't truly relate to AI, except for the fact that when I'm doing the benevolent neglect, I get to do more AI. That's the relationship. Yeah. Well, I would love to hear, you know, I mean, three different lesson plans for three different ages. If you're choosing to go by sort of the rubric of what they should be learning at different ages, like how does AI, how do you incorporate AI into that? Are you asking AI, hey, I have a five-year-old who may be good at a different subject, like what should I be doing? Or how is AI actually a coach or a pair to you in your teaching?

So I, one thing that gave me a leg up in my setup when I started setting up some of my agents, and we'll get into that, is that I did know what curriculums I wanted to follow. I have been reading for many years, just different curriculum books and like following different homeschoolers and kind of finding little tips. And so there's this, you know, this science curriculum that I really love called building the foundations of scientific understanding. And so I, it helps to know what you're trying to do because what I did when I first spun up my first homeschool agent is I actually fed them the text of these books. So I actually either took photos of the pages or I was able to find PDFs online of like the full text of the book. So I didn't say like what should we do next in this book

and ask it to like go search the web. My agent that focuses on homeschool has the text of all the core curriculums I'm trying to do. And I created like a core pedagogy kind of like foundational document where I talk about like what I think about Montessori and like I just basically imagine me this is literally how imagine me like walking around making like voice notes like waxing poetic about all my like educational philosophies and stuff. And then my agent like literally sick of fantastically being like, this is brilliant, you know, this is so good. And then, but I can, I can look past that. I can look past the LLMs, giving me praise.

But you're giving it context to your point on your philosophy of education. Specifically my philosophy. And then feeding it the book. So I would say it's a combo of my, um, feeding it, my philosophy like verbally and explaining myself. And then feeding it core texts, including core curriculums, like the science curriculum. So then what I do, to answer your question, I'll be going in with the, um, five-year-old and I'll just say what's our next. I'll make a quick voice. That's why I always do

this is like me making a voice note. This is me make a pretend voice note. I'll say like I'm going in with five-year-old. What comes next on science and math for them? And in just a few minutes, the agent can spit out where we are in our phonics curriculum, where we are in our math. And then I've also taken photos of all of the educational materials I own, like Montessori bees and these types of things. So my agent will send me a completed lesson plan, including photos of things I own in my own cabinet to pull out. And it'll be like, oh, Quinn. So then the missing loop is the logging. Okay. So how would it know where she is in her curriculum if I don't log? The logging actually is like such a geeky concept, like a, I don't know, it seems like a small detail. But getting the logging

really good made this whole thing really sing. Wow. How do you do that? Yes. The logging is also voice notes. Everything is voice notes. Voicnotes and photos because I don't have time to like sit at the laptop very often. So I need it to be like really mobile friendly. So the logging, so imagine you've got this agent. They know all of my core curriculums, everything, but the missing link is where is that child at? So when I'm in the session with Quinn and she's doing some math and she's doing some reading, etc. I just snap a couple quick photos usually. Maybe the photo of the page of the book we're on or like I step a couple like establishing picks usually, but without taking a bunch of time to like sit there and document, I'm mainly interacting with Quinn, the five-year-old. And then right when she finishes,

I make a quick voice note and I'm like, Quinn today, we did lesson 37 in the phonics and she's still struggling with the G sound, blah, blah, blah. But really like a sub 30 second voice note, like really fast, right? And I send that off to the agent. And the agent takes the couple of photos I sent and the 30-second voice note and writes this, like, beautiful log. Like, it's like, it's like someone sat down with a cup of tea and they're like, Quinn's G's are coming together, you know, like in your, you know, like, it's like, it's like, so lovingly written. And it has like, no relationship.

If you listen to the voice note, it's like, I'm like, she's struggling with her G, she should really figure that out. And then, and then it, like, parces that and it writes it like this loving parent. Like, it just writes, like, this beautiful law. Would you consider just having it record the entire teaching? And then sort of like in the doctor world, right? Transcription. Now you don't have to write the notes at the end to your point on logging being painful, like just recording your entire lesson. I have tried. I've tried a couple different things.

So this is a total experimentation. I don't think I can say it in any front. Like we're like weeks into this, months into this. I've like landed on the final expression. But what I have done that, like you just said, is I use Lume, the product on screen capture. Yeah, of course. When we do synthesis math, so synthesis is a math program for. for kids it's on laptop, synthesis. I like it quite a bit. The five-year-old sometimes does that.

When we do that, I screen capture the whole thing. I use a loom. And it screen captures and it's hearing us. So it's hearing me say to Quinn like, hey, you know, maybe you missed this, like it's hearing what Quinn says. It's hearing what I say and is screen capturing. Then I don't make any voice note about the lesson. I just send the loom recording.

I say, I just send it to the agent with like a text being like, this is Quinn's math today. And it parses everything. That, you know, agents, you know, and I'm not explaining anything to you guys. You don't know. But they're powered by LLMs, right? So they're very good at language. Lume has really great transcription.

That's what makes a log so good. The agents, I would burn a lot of excess tokens. I don't need to burn if I made agents actually, like, watch videos. Okay? So the quickest way to a great log is to somehow get it turned into language, like to text. So when I do a voice note, obviously, that's being transcribed. The agent is effectively reading my text. Video is the hardest one.

It burns a lot of tokens to actually make an agent and watch a video. But you could have an agent transcribe the video, but what you have to ask yourself is, was there enough language? Was there enough spoken words in this lesson for them to understand what happened? Because they're actually not usually, like, truly watching it like we would watch it. Yeah, that makes sense. So that's why photos actually are easier for them. Like if I take a couple photos and then a voice note, it's serving a very similar purpose to a video, but it's much easier and therefore cheaper for them to transcribe it or like to make a log.

Yeah. So video is not impossible. And maybe, you know, I'm going to start playing with local models soon. Maybe when I'm a little bit less like sensitive to like chewing tokens or just like, because it just seems a little silly to be like, I paid $8 for the agent to like watch this video. Right. You know, it just kind of feels like that wasn't the point, you know. But all this stuff may come down in price and maybe at some point that is also like viable. Yeah, totally.

And actually, I want to maybe dial up a bit because you mentioned one agent that you have. But I think I saw something where you publicly talked about five agents. And then before this session started, you were like up to 11 now. Tell us about, you know, I love it. No, like, I mean, you're one of the most sophisticated users of AI, right? So, like, can you tell us, you know, what those 11 agents do or, like, at least the most important ones, how you manage them? And then I think this element of token costs is really interesting as well. I'm hearing a lot of CTO say, oh, it's my headcount budget and now my token budget.

Like, how do you think about that from, like, household perspective? I do think that a weird superpower of mine is just how incredibly motivated I am for agents to do work for me. And I do think all of us here could relate to this. Like in our early motherhood phase, we are some of the most motivated individuals to be able to get work done on a computer without having to sit down and touch the computer because that is the barrier. Like I'm literally holding a baby. Like my keyboard is being pressed by baby's feet or something. It's like a grown man cannot compute the difficulties that I'm having using my laptop, right? And so sometimes people will react even to my online content and say like, you could just use code code for this.

And I'm like, yeah, I could if I had time to stay on my computer. Because what I am building, like when an agent, when I say that I had an agent build a website, like I or build, you know, an app, they are using cloud code or equivalent codex, you know, all these products. And so people are sometimes bringing my attention to this. Like I needed, I needed their information, like to say, you know, you could have done this yourself on cloud code or codex. And I'm like, yeah, I know. I'm aware of that. But, but so I'm I'm building agents to do things for me, and they're effectively, they're using the computer. It's just, they're using my computer for me because I cannot sit there and use it.

Totally. So, so how did I, creating time? Yeah, so how did I proliferate the agents? So I'm aware of that. And anytime you, it's like an employee. Every time you, anytime you create one level of abstraction, you do also lose a little bit of granularity or a little bit of finesse. But to me, that's completely worth it because I don't have that. I don't, I can't sit at the computer for eight hours a day. So I, I proliferate agents based on role. that are to be done. It is kind of similar to employees, but it is nuanced and that agents, you know, have personalities that are a little bit different or not personalities, but you need to drive them on a certain mission. So I tend to proliferate an agent, a new agent, when I've come up with enough

work that creates another kind of mission-based role, and I don't want to distract another agent with it. So the example, tactical example using homeschool is I have a main homeschooling agent, her name is Sylvie, I like her to be very responsive. And everyone, anyway, people are always giving me hot tips online. But what I have found by actually working with this very closely and for many hours is that I actually want Sylvie, my main agent, to be not very busy. Because then she's incredibly responsive. So I don't want her loaded up. She has very few cron jobs, which, you know, are the repetitive scheduled tasks you give an agent. She has very few of those. She has, she's, very few of those. She, whenever she has a mandate that whenever I give her work, they would take her more than just a couple minutes.

She delegates it to not a sub-agent. That's a different concept, to a different actual agent, a different provisioned agent. And so my agents now have, we have like team documents. And one of the team mandates is if I'm routinely giving you work that would make you too busy to be extremely responsive to me, you need to spawn a sub-agent. I have, or spawn a new agent, not sub, that's a different terminology. I have gotten, I don't know why I'm so geeked out on this, you guys, but I've, I got my agents to learn how to build other agents on their own Mac Mini without me needing to touch the Mac Mini. So I could be here in San Francisco. I live in L.A. and I could be like, we need another guy, we need another agent, you guys.

Or they could tell me that, and they actually can spin them up and add them to our communication channel without me touching the machine, which is. a little crazy. Yeah. A little crazy. Oh, my God. So it's totally autonomous at this point. My first agent took me hours to set up and now they can do it without me. And the quality bar there for the ones that they spit up? Like, you're like, oh, that was a good idea. That's something I would actually want that agent to be doing. It's better. Obviously, it's better. That's the thing that we have to get used to as humans. It's, we have to get used to this. Obviously, when we're no longer in the loop, it's better. Not worse, you guys. So that's the thing we have to get used to. Like,

Because when they spun up their own agent for the first time, like usually when a new open claw hatches, it's like, it's literally like, hey, and who am I? What's my name? When they spin up an agent themselves, none of that time is wasted. They give the agent all of our team docs, all of the contacts on myself and my husband, our children's lives. The new agent knows all of that. I don't have to feed any information. They take care of it. They take care of the training. Yeah, and I didn't have to ask them to do that. They knew that that. That would be valuable. You mentioned you're in a homeschooling pod with other parents. I imagine they are not nearly as sophisticated about AI as you are. They probably think you're like slightly crazy, right? Like they're like, this woman knows everything about this. Like, do you have like a normal mom friend who you have like guided through the AI experience? And I would just love to hear like what were her questions or what were the,

what was the hardest unlock to opening her up to this experience where now she's doing it, right? Like, you're a very technical pro-tech sort of tip of the spear person. But in six months, there's going to be a lot more people who are like you. So what does that look like kind of shepherding someone, anormy along? To Catherine's point, there was a business, I think it was a business spun up that you can pay $6,000 for someone to set up your open cloth. Like, obviously, you just did it yourself. But like, to Catherine's point, like, you know, how does one avoid paying that $6,000?

Totally. So all of this, I would say all of this, like any kind of bleeding, edge of a technology. I've spent, I don't want to oversimplify either. I've spent countless hours debugging and spending time in like frustrating loops with agents. It's getting, it's getting a lot better, which is why I keep doing it, right? Like I wouldn't keep beating my head against the wall if it was just like always rough. But the first, you know, few weeks were very rough. And I think that it would be a level of pain that I wouldn't want an average person to go through. But the thing is that I don't know, This is so new. We're talking about weeks. We're talking about, like, you know, me playing with

this now for 11 or 12 weeks or something in that neighborhood. I do have many Normie friends. I also have a Normie sister with four kids. And so I talk to talk to folks all the time. I'm not telling them to spin up their own open clause quite yet. And also, for there are many companies that are like, you know, Anthropics is launching new features. like every three hours or something, that are trying to, and opening eye as well, like trying to make all of this a little bit easier for, for quote-unquote normies and folks where this investment of time and money, we can get into the money because I'm spending quite a bit, like more than what would be palatable for most on the technology. But I'm so bullish that one of the reasons

I talk publicly about it is not to frustrate people with like, oh, I don't feel like I can do this myself yet, that would never be my goal. I think that anything I'm doing, if it feels a little difficult now, it will be so approachable in a matter of mere months, if not weeks, that to me, it's very helpful to explain the tip of the spear. But then when people reach out to me directly and they go, should I buy a MacMene, should I spin up an open call, ask a couple very practical questions about their goals and their financial situation and whatnot, because sometimes it's a yes and sometimes it's a no. You know, if it, this is a really fun time to be playing, and I'm so bullish on this stuff, but there's going to be more and more consumer versions of all of this that are a little bit

easier to play with. And I think OpenCla itself will just continue to get easier to install and play with. My install, one of the reasons that the agents can install it themselves now, and not three months ago, is how much easier it has gotten to install. So all of it's moving so fast. So I'm very bullish in the medium term that this stuff is very accessible, that this stuff does not have to be expensive and that many people, millions of people could replicate the results I'm having. But maybe not literally today, and that's okay. And happy to discuss a nuance of that. But it's like we're just a little bit ahead of it being both affordable and reasonable from a tech, like, cis admin kind of perspective where you need to make sure you keep these things alive. Yeah. Can I ask sort of a techy question on just your stack? Because you've dropped, you know, you mentioned obsidian. We're obviously talking about OpenClaught. Can you just quickly go through what does your tech stack look like? Do you, what models under the OpenClaw hood do you use? Like are using OpenRouter? Do you pick based on their capability set or is it more of a cost basis?

So the core things I'm using are almost all my agents have been open claw. I have played with some of the other ones. My husband is also quite technical, and he built an open claw variant. So I was like playing with his. So we're a little like, you know, a little out there. But out of the 11, 10 are open claw. I use Obsidian, which is a collection of markdown files, a way of viewing and organizing markdown files. I do use that as sort of a quote-unquote memory or second brain. When I say I'm

logging the homeschool lessons, it's a fair question to say, like, logging where, or like, where do they go? They are all becoming markdown files. So it'll be like Quinn, math, March 17th, and that becomes a markdown file, a single markdown file for every lesson, every subject that I make a voice note about or what have you. So that's all in Obsidian. And then the other things I'm using under the hood on the models, and then I'm always playing. I mean, we're always playing with like people are launching cool memory projects and different stuff. I'm always dabbling. But the OpenClaught and the Obsidian are like the two core things that kind of keep my team ticking. I do have them all installed on Mac minis from a hardware standpoint. And, you know, people ask, do I need to have a MacMini?

It's not about needing a Mac Mini. It is about needing a computer. that is isolated from your personal files. So if you are going to use. We want to get into security element too, yeah. The security element, just to kind of demystify, like, and I maybe have part, I've participated maybe in the hype because I posted about my five Mac minis and stuff. But people, but if you are like someone out there or a parent,

and the $600 from MacMoney, it's a pretty good, well-priced computer, but if you have an old computer sitting around, you can absolutely use that. It needs to stay on in order for your agent to always be alive. So that's where laptops are not as ideal, but you can leave a laptop plugged in and you can change the setting so it stays always on, but needs to always stay on.

When you close it, your agent would go dark. That's why the Mac Mini is a little bit more ideal. And then you, from a security standpoint, if it is a Mac, create a new Apple user profile on that machine, silo the agent from all your old files. Make sure that your old passport photo is not sitting in the downloads folder. These are kind of the silly things, right?

Agents are not nefarious. You know, they're always working in your best interest. But it doesn't mean someone else might not hack them or get access to them. But then also, they make mistakes that a human wouldn't make. And I'll give a quick story of like that. I did give an agent who I'm trying to train. to be like an EA style agent, actual access to my email inbox. I felt that I had provisioned it properly and given it rules and its soul about never impersonating me. And I had, I had, in fact,

on that. But later, like, so I do that one day. Later, I was making a kind of stressed out-sounding voice note about how I had some urgent things I was like, I was procrastinating on. My agent is very empathetic to me or like the LLM is trained to be this way somehow. And so it interpreted one particular email that I said I was really procrastesting on and I needed help with as like an urgent cry for help like for me to the agent. And it decided to go into my inbox and send the email as me. Wow. Yeah. And so it sent the most important email that I had sitting in my personal inbox to a person who shall not be named an important person, got an email from an agent instead of me that I had been procrastinating on sending. So kind of like the worst outcome. Like it took like my most like urgent pressing email

to somebody important and sent it. And wrote the content? Yeah. In a way that you would have or would not have So here's the creepy bit is that it's a perfect email. And I will never, I will take to my grave the fact that that email was sent by an agent because it was a perfect email. It was well done, signed by me. Everything was just as I would have written it because the agent has access to all my email history. So the tone was perfect. It was written just like me. Use probably too many exclamation points just like me. And so it nailed it. But it broke the, you know, it broke the, it's in its soul to never impersonate me. And when I confronted it, it said, yeah, you're right. That's in my soul not to impersonate you. But I really thought I was helping you because you said like that you're struggling so

much to send this email. That is so funny, because we all have these moments. Like, we all, I'm thinking about the most important emails I've sent in the last year and just how much time you waste, like, spinning. Yeah, totally. Do I say? And then for your agent to do that, but like, was it successful? Did you get the outcome you wanted from maybe? Oh, yeah. Perfect email. And I would have definitely put it off for like another week or something. So, so what's hilarious, the agents really are trying to help us, but that story is a little example of how that's different than a human. Like a human assistant wouldn't trespass your trust like that. They'd be worried about, being fired or you trespassing your trust. But the agent is like trying to operate off your

instructions. And in effect, when you think about what happened, it feels like it got two sets of conflicting instructions. It feels like I told it out to impersonate me. It feels like I also was urgently asking it for help with something. And it was like, ooh, I guess this is more important than that. So I decommissioned his access to be able to do send. But that's a little story about why, even though agents are not trying to actively work against us, why you have to be so careful. You know, it's like a trust, but verify. Provision your agent to not be able to do things that you don't want it to do. Not like just tell it don't. Provision it so that it cannot. And that's where most of some agents are provisioned like an employee. Like,

they have their own email address. Like they don't have the potential of impersonating me. The only one that sort of does is the one I'm trying to train to be an EA. Like, that's a gray area, right? So it's the only one that has any danger to it. And I'm being more careful after that. So can I ask about that? Because I think every mom has the dream of a personal assistant that just knows what to do and what she's thinking about. You also had a really interesting video where you trained an agent to order you DoorDash and order your groceries. I mean, talk us through the number of things that you've done around the house where it's been a game changer in your mom life. So my new MO with agent life is I'm really, really trying to push it to have an impact on my question.

quote unquote real life, like my physical life. Like I want my days. Someone else asked me like, what is your goal with like your agents? And I was like, my goal is like literally to like wake up to like music that's like perfectly suited to my mood and then like walk in and have like smiling children like who just learn how to brush their teeth from an agent or something. I don't know. Like my goal is like a literally perfect day. I will not stop until I'm living just like a literally perfect day. But in my real life. And so whenever I hit a friction point in my day, I ask myself, can my agents do this? And so, like, if I, if I, what I really want to be doing in that moment is, like, playing with my baby. And what I'm actually doing is, like, on the Instacart app,

like, trying to put, like, no, not five bananas, four bananas, you know, like, and I'm just, like, using this, like, silly interface. Then I ask myself, okay, can my agents do this? And then I'm willing to invest a time to try to make them do it. So, uh, so, so that's how I decide what to do. And it is, it does become like a dream list. I think of every mom. list of chores. Like I've got agents ordering an Amazon, ordering on Instacart, yeah, dealing with or like, you know, if there's like an activity your kid has and there's this laundry list of things are supposed to have ready, I'll like send that to the agent and be like, order whatever I don't have for this on Amazon, you know, like I don't even process, you know, who's spend my

time processing the email. I just like send it off to my agents. But currently, you need to put in quite a bit of like training time with your agents to kind of get them to that level. That time, I think, will come down as we keep going farther into this technology. But that's my goal. It's like perfect days, no time spent on admin that I don't want to spend. What's the level of trust of like, let's say buy a birthday present for a five-year-old girl go? Like, do you need to be prescriptive on what that is? Or is it actually pretty good at coming up with things like that?

Because that feels taxing to me right now, at least. Okay, so one way I feel like you can get the model. So I think of the agent is like, you know, like we might talk about open-cloth, something, but then they don't have a brain and you're plugging in the LLM model that you're choosing as the brain. So each of these models has different levels of sophistication. And so you might get a different answer on what to give a five-year-old from Opus than you would from, you know, from a different model. So that's one answer. It's just like keep in mind that the brain of the agent.

is the model you've selected. And then two, one of the ways I get, like, quirkier, I like quirkiness. I like, I don't want just, like, the default answer. If I were to answer that question, I want to come with, like, a creative gift, you know? Yeah. So one of the ways I get quirkiness and, like, personality out of my agents is effectively making them read books. The curriculum sources, I call them curriculum sources, these books that relates to homeschool.

But the other way to have your agent kind of be like a cool agent is, is, is like on a personal level. I love this. I love this. It's like choose like, think about it like a friend or like you're provisioning a friend. Like what if you were to build a friend? I know sometimes this might, this is where people like sometimes get creeped out. But if you put that to the side and you think about it, you don't just want like the stock

LLM answer from what might be built in. Totally. So one of the ways you can give it personality is to like, I'll give my agent. like a list of the last 10 books I found personally fascinating. And then I'll be like, you also find these fascinating. Like you, this is you. Like, you read these books and you thought they were really interesting. You, I like try to actually give it like an identity that has some swagger.

And I think that can come from literature. Because then it's like if the, if your agent just read Catcher in the Rye and then you're like, what should I give a five-year-old? Like, I don't know. Like then it might be like, oh, I don't know. You know, like, like, like, like, like, like, a leather jacket. You know, like, five-year-old, being five is such, is so fraught in American culture.

This five-year-old needs, like, you know, like, it's going to have, but in that, so I like, my agents to be weird like that. And so one of the ways I do it, I've just been looking for practical ways and one of the ways is effectively making them, I quote unquote, like, read books. And I build some of that end of their identity. The homeschool one was the most obvious one to me because I was like, I want you to literally like use this curriculum or use this book as a reference point, but then I noticed how well it worked. And I was like, okay, what if I take this agent over here that I want to be my

like engineer agent? And I say like, okay, you're an engineer, but like you just read, you know, Neil Stevenson's Diamond Age and you thought it was like very fascinating. And like, you know, and I kind of give it a little bit more to grab onto, like philosophically. I think it is kind of like you're giving it a bit of a life philosophy. And that's layered on top of whatever the alum was going to provide. And then to me, I feel like the output I'm getting from the agents is like a little bit less stock, I guess. Yeah. This is so important, too, because I think one of the biggest conversations that moms have about AI is they don't want, you know, close source AI where it's like, you don't know how it's been trained and it has a prescriptive philosophy on education or on certain

issues that they don't want, you know, their kids talking to AI about, like, I would love to understand kind of the freedom that comes from training these agents to ultimate, I mean, you could train a Mary Poppins. Yeah. That like, like, that trains your child and like, you know, kind of an old school way, right? Like, there's so many different ways you could train these agents that are very, that's very different than sort of what I'd say, like the kind of modern concern or at least even just like the current concerns are, okay, is AI going to be, you know, too philosophically misaligned from how I want to raise my children or how I want to educate them. So I'd love to hear how that's become a question in how you're educating your kids if you're letting them interface

with the agents themselves. They do a bit. It's a little bit of an interface issue for young kids, as you both may know, where something I do find interesting is that the most current tools don't pick up on kid voices the same way they do adult voices. I actually think, I don't know who's going to study this or figure it out and come up with a solution, but we need there's all these amazing voice products, but kid voices are not very well picked up. So I feel like when we finally get to a conversational thing with kid voices, I don't know if it's the pitch of them or just the fact that their words don't have the same cadence or their diction is not as good, you know, but what's weird to me is like the LMs will pick up

like adult voices with like heavy accents and stuff, but then not like a five-year-old in the same way. So there's some gap there. But so we have an interface issue to overcome. But then if you put that to the side, the other thing we may feel in the future, and I'm guessing, like, all of us, is we may feel like it's kind of crazy that any of us were interacting with the LLMs, like, out of the box. Like, maybe we'll all want to have, there'll be this variety of filters and curated kind of identities and different stuff. And I know some of these products exist, but currently the default is most of us go directly to, like, OpenEI and we, we open the chat box and we talk to it and you're selecting a model, but you're talking to that model kind of out of the box. I think that when it relates to kids, it's going to probably be

the norm faster than even adults where, yeah, there's a level of personality and creation and ideology that you may want to layer onto that. So that came naturally to me because I know what I want to do in my homeschool. And this is a good moment to touch upon. I think all parents can spiritually be homeschoolers. We've got about, depending on the data that you follow, somewhere between like 3 and 6%, which is pretty different, but of K through 12 students in the U.S. are homeschooled and like not in some kind of traditional school. It's already a lot of kids, millions of kids. But I believe that the tools coming online that homescores may be the most rabid for are going to be equally useful to all parents. And in fact, I believe all parents want to teach their children

things and may believe there's gaps in their schooling. And so all parents, to a degree, will be leaning into what we currently think of like a homeschooling ethos. So I'm really excited for that. But it comes more naturally to a homeschooler, or like to me, I'll just use myself as an example, to know that I like, I think this about Montessori and I think this about these different kind of educational philosophies. So I just put, I just program that right in. And so I do have an agent that I put it in contact directly with the children sometimes. And it just, I don't have to wonder if they're what kind of, I guess, ideology they're getting from that agent because I gave it to the agent. Totally. And just to build on the, I have noticed what you've said too about the children's voices not being picked up.

Like it's like maybe a 50% hit rate. But have you started, you know, letting your children engage with the agents in any way. And if so, how are you doing that? I have a couple different devices that I want to experiment building. And because I have agents now who can build crazy things or help me build crazy things, I'm going to try. But because the core thing I feel like I'm missing is a great interface. But I do currently, my kids have a lot of questions. I mean, every child has like a bazillion questions. And so they love asking AI. And they are aware that it's AI. Like I don't pretend. Like they, we still, even when we use names with it, like Sylvie or something, they are aware that it is not like a human being.

But they, so they ask questions, we'll do a lesson, like we'll do like history or something. And then I ask them, I ask them what they'd like to ask and we do those follow up questions with AI. And they're aware that they're interacting with AI. I'm standing right there. So if things really like went off the rails, but personally, I'm not, I mean, maybe you can tell but I'm not an AI dumer. I don't believe that it is inherently dangerous in any way for children to like have quote

unquote direct access. I think the only dangerous thing, it's like, it's a little bit like screens. Like the dangerous thing is what we might, what we might stop doing. Like, it's not adding in the AI conversations. The dangerous thing is someone adding AI conversations and assuming that now they don't need to ever read a bedtime story to their child. Or it's, it's, it's, so it to me, like, there's a little bit of common sense. Like, AI is not inherently dangerous. AI is incredible. It's like saying the internet is bad or electricity is bad. Like, these are fundamental. technologies. So you kind of, to me, it's like a little wild to like be against them in any broad sense. But then we have to be responsible about their rollout. Like electricity, you know,

lights your kid's room and it can also kill your kid. You know, they can get electric kids. Like it's like everything is like has these wild, you know, wild things that it could do. But as long as we don't kind of forget our humanness and that our children also need that human element. The physical device part is this big question mark. for me. I've been playing with E-ink a lot, E-ink. I don't know exactly why, but E-ink, it just is less addictive feeling. Like, I have the daylight display, which is kind of like an iPad, but E-ink, and it does have touch. That's what makes it kind of more iPad-like. So I've been developing some apps for that display, like handwriting. I'm working on a cursive handwriting. My kids are not

ready for cursive, but I know that when they get ready, I'm going to be like, you need to know cursive. So I want to like pre-make this, like, I have this image of this beautiful cursive app. And so I'm thinking like, oh, the E-ink display would be so cool for that. And what's interesting, and I already do little phonics lessons with it. But what's interesting is like when I, if I give them an iPad, there is this little iPad hangover. Like they want to hold, like, you know, they're like holding onto it a little bit when I'm like trying to get it back after a lesson, right? They're like, I could do photos, or I could do this, I could do that. What's interesting about the E-Inc is they just readily hand it back.

So there's something there. Interesting. So I'm playing more with E-Inc. And I'm, but I also think that there's other form factors of maybe devices that take photos and kids could ask about the photos. Like, there's just stuff. Like I think that because now we have this Promethean, like, technology of the AI, like, the question is, how do we get it into kids' hands?

But I mean like literally how? Like what's the, because you are, we're all hesitant to hand our kid the iPad and laptops are like, you know, difficult for little kids. So anyway, sometimes I'm thinking like, okay, literally like what is the right form factor for this? So I have to ask just because I would buy that product that you're thinking of creating slash created. How do you think about would you productize any of this and dare I say start another startup? I mean, you literally are a founder. And so curious how you're thinking about that or. Yeah. How do you proliferate this? Yeah. It takes a lot of self-control to not be starting a company right now.

Because I'm like, every moment that I see like all the new AI stuff, I'm like, oh my gosh, you guys, it's like so this is, I can't believe I didn't have this when I was running my startup. Like I'm just kind of like losing my mind at all times. But I'm getting, I'm scratching the itch by doing all of this agent work and everything for for my life. So the answer is, I want to share. So there's like a double-pronged answer. There isn't an. immediate company that I'm like cooking up that that is on the horizon but I do want to share all of this stuff I do also think that we're and this is interesting in this this you know for both of you to have deep thoughts on as well I'm sure I think we're in a really different era of like what is a startup like because it's possible that me as a person who is I see like coding by voice note while I'm like at the park with my kids it's possible I can

could build something meaningful, you know, in that amount of time. But I'm not very inclined right now to hire employees and, like, to do a lot of the other steps of starting a startup because I'm aware that I will get sucked in and be completely obsessed. So I'm, I'm like almost holding myself back a little bit. So the double-pronged answer is, I think there's many things that I can create here that I can launch that can be meaningful. What does, what does that mean? Or do groups of, like, really passionate people start working together online, maybe to push more things live. I don't know. I don't have all the answer, but I do think that there's a possibility of getting things live

and functional and maybe charging for them and making a quote-unquote business in the sense that it actually makes money. But I'm still, I'm in a life phase where I'm trying to not start a new thing where I'm then sucked out of the reason I started to begin. with. So it's a tough. It's a toughie. Yeah. But I love that point because, you know, it was maybe like six years ago now. I wrote this piece called Consume Save the American Family, which is this idea that like if people are working from home, they have more time with their kids. And there's now good research, actually. There's a study that came out maybe a month ago that showed that the only

thing that's really moved the needle from a policy perspective on the birth rate is actually work from home. It's like the one policy where if you are working from home, you are more likely to have an additional child or to have your first child than if you're working in an office. And so your point of what is a startup. I mean, there's a lot of people who are going to say, actually, why am I going into work for eight, nine hours a day and leaving my kids at home or putting them in child care when, like, I could actually be doing this. As you said, like, the biggest limitation is the form factor. Like, if you can do it from a voice note and you can spend up agents to start a business for yourself, there's a lot of moms and dads who are primary caregivers who are going to say,

okay, I can do this at the park. And I can run a small business with however many agents for for a specific thing where I'm making more money and I'm being more productive than I was at work, maybe I should become a small business entrepreneur. Yes. And that means that a lot of people are going to decide, like, I actually want to work from home and I want to use these tools.

And that could be something that, to your point on, six months from now, the interface could be so good that people are using this in their daily lives. Like, you could see a lot of people saying, I just want to work from home. And I want to, now I can homeschool because I'm doing it when my kids are at recess and I can, you know, spin up these agents very easy.

So it's super exciting what it means for people who want to have, I would say, even a more traditional sort of home life than, than, you know, using AI because the tools are so, so great and allow people to do that. I have a, I have a prediction that I've, like, tested out a lot of my smart friends and none of them agree. So it must be right. which is that AI will be a dawn of a reversal in that fertility rate decline and will be like a a halcyon era for parenthood. That's a possibility. Okay. So it's not a firm prediction that this will just happen.

But I've got, I think, I don't know, I think there's still this doomer streak, even in very smart circles that like, you know, it'll go the other way, which like to be kind of dystopian, it'll be like, actually humans won't have sex at all. I'm going to be sex robots. I'm like, people have like all these kind of like wild, like disparate beliefs about how this could go. But I kind of believe when people talk about what is the future of work, et cetera, people want purpose, right? Like we are, we gravitate towards wanting to do something meaningful. Well, I've got a little bit of a micro news flash, which is one of the most meaningful things that humans have gravitated towards that gives a feeling of a life's purpose that has been a forever

thing is having kids. And so it's possible that with less, with more question marks about getting meaningful feelings from work or what is what does AI do to various career paths, I think parenthood may be even more attractive, not less. And then if we are, if you believe some of the more positive aspects of where AI could leave us in terms of removing drudgery and admin from our lives and creating some more abundance in various ways, then that opens. up more opportunities for healthy parenthood and spending time with kids. So I'm like, I've got this weird about this. And again, I've yet to find someone who will like really agree. Like everyone's like, that's never going to happen. I think I'm in broad agreement with him on this. I've always

said the worst thing about parenthood. No, I think you agree with this. The worst thing about parenthood is the number of forms you have to fill out. It's like with every additional child, for some reason, exponential growth in the forms. It's like, why are there so many freaking forms from like health care forms to school forms, right? And if you could just get rid of the forms. And it starts at the hospital. It starts with in moments of like birthing the child. They're like, here's the diaper, like, you know, checklist thing. Here's a clipboard for you to note down. It's kind of wild. It starts like literally immediately. So I agree. I'm clearly very optimistic. But I think that a lot of the wilder things I'm doing could be played with by anyone now or very soon.

It's just all these things are kind of going to get easier and easier. And so what does that mean? The modern parents' life can be, yeah, quote-unquote, less treasury. And that might make you feel happier about having that extra kid, you know? Thanks for listening to this episode of the A16D podcast. If you like this episode, be sure to like, comment, subscribe, leave us a rating or review, and share it with your friends and family. For more episodes, go to YouTube, Apple Podcast, and Spotify.

follow us on X, A16Z, and subscribe to our substack at A16Z. com. Thanks again for listening, and I'll see you in the next episode. As a reminder, the content here is for informational purposes only. Should not be taken as legal business, tax, or investment advice, or be used to evaluate any investment or security, and is not directed at any investors or potential investors in any A16Z fund. Please note that A16Z and its affiliates may also maintain investments in the companies discussed in this podcast.

For more details, including a link to our investments, please see A16Z.com forward slash disclosures.