you can just do things (ft. anson & hudzah)

When I talk to young people, they often ask about finding time for creative work amid an accelerating rat race. When I meet folks older than me, they lament that the kids are cooked—AI and TikTok have rotted their brains. While these crises are real, I dislike narratives of inevitable societal decline, and retain stubborn faith that technology can be expansive if we learn how to wield it. Alongside products and policies, that means developing new personal ethics for living alongside AI—new cultures and attitudes to accompany new tools.

I invited my friends Anson Yu and HudZah to chat, as young people thriving amid these ~unprecedented times~. Our conversation very much centers on agency: how it’s nurtured by family and by community; how it’s shaped by stories, books, and films; and how AI can instill it or take it away. Despite my usual distaste for Silicon Valley’s buzzwords du jour, I haven’t been able to shake this one. Agency, free will, creativity, or drive; whatever this is, I believe it’s the critical resource of our time. And in addition to being generous, warm, and fun, Anson and HudZah are two of the highest-agency people I know.

Quick bios: Anson and HudZah are both graduating seniors at the University of Waterloo. Anson studied systems design and HudZah math. They're both involved in organizing Socratica, which recently rallied 2500 young people around the world to fill a hockey stadium for a giant demo day. They also have a million other projects under their belt. But to highlight two, Anson spent six months on a road trip filming Unstuck, a feature-length climate documentary with a friend, and HudZah is perhaps best known for using Claude to build a nuclear fusor in his kitchen. Now, they’re cofounding an AI-for-hardware company together.

I recorded this conversation to learn from them, and I think you will too.

Watch/listen by clicking above; or on Spotify, Apple Podcasts, or Pocket Casts. A full edited transcript is below.¹

Full transcript

Jasmine (0:00)

To start, I would love to hear how you each met each other and what your impressions were.

Hudzah (01:31)

It's pretty stupid. It was 2022 and I was on LinkedIn. I found Anson's LinkedIn and she had made this website called Reflect which was, funny enough, a website to block your social media use. I thought it was great, so I messaged her and we had a call. Then I met her at a Socratica thing after; I didn't know if she remembered me or not.

Anson (02:06)

We met as all great friendships start: in the LinkedIn DMs.

Jasmine (02:10)

Anson, did you remember him?

Anson (02:13)

He threw a barbecue and I came by and was like, this is actually very well organized. At that time I was flagging every person that I thought was mildly extroverted in this very introverted town, and putting them on a flag list for myself of potential people to be friends with or to organize things with in the future. When the barbecue was actually fun, which is kind of a crazy thing in this town, I was like, noted.

But I actually think that he reached out to me wanting a referral for an internship.

Hudzah (02:46)

That's also true. I wanted to work at Notion.

Jasmine (02:54)

I thought this was going to be a lot more whimsical than a LinkedIn internship referral.

Hudzah (02:59)

What you'll realize is that we were quite cringe actually. You have to go through the valley of cringe before you reach some place of enlightenment.

Where agency comes from

Jasmine (03:20)

The first question I wanted to ask each of you is, what was the first moment that you realized that you had free will?

Hudzah (03:28)

When I was 12 or 13, a friend who was very into horticulture—like growing tomatoes in his backyard—asked me to help him build a website control system to grow a bunch of plants and track them. I was very confused, but was like, “Sure, let's do it.” Slowly this friendship evolved to the point where we became business partners. We actually started a fully fledged greenhouse company, and we were selling lettuce, tomatoes, and all sorts of vegetables all around the country.

I remember one day we were buying a bunch of metal to build this greenhouse, and we hired this random dude off the street to put it in his truck. I was just sitting in the back of his truck on a highway with a bunch of metal on top of me. And I was like, “Holy shit. What is happening right now?” And that was probably the moment where I was like, “I did not tell my parents this. Nobody in the world knows I'm doing this. And I'm just here with a bunch of money and trying to do this insane random thing that I probably should not be doing and that was very much not legal.”

That was the first moment that it clicked—I could just kind of do things. It was quite formative, and I was lucky to have experienced it at a young age.

Jasmine (04:51)

Where were you in the world?

Hudzah (04:53)

I was in Sri Lanka, so that's probably why I was able to do it. If I was in Canada or the US, there's no way I would have gotten away with that. Someone would have shut me down. The police, my parents, or someone else. But growing up in a third-world country where parents let you roam around the street and buy groceries and bike on the highway, the risk tolerance is much higher. You see people doing a lot more things that kids here now would do at a much later age. So naturally you grow up knowing I should be doing this on my own and that I'm in control of my destiny.

Jasmine (05:37)

I don't think I could have started a greenhouse company without my parents knowing.

Hudzah (05:43)

Or the law or something. It's definitely not legal. We had no permits or anything. We didn't know what a permit was. Part of the pros of not having rules is that you can do a lot of things. Of course it's good to have those rules, but as a young person who's ambitious and curious, it can be an inhibition to your growth when someone tells you no. But it also can build rebelliousness, which is also quite valuable—in good dosages.

Jasmine (06:16)

I joke that my libertarian origin story—and to be clear, I'm not a libertarian—is that when I was a kid, right behind my house, there was a really big park where they held the international softball world championships every summer. I had the idea to have a lemonade wagon instead of a lemonade stand. I thought, these people are watching the games, and rather than going to the concession stand, it would be better and more lucrative if I dragged a wagon full of lemonade and ice to the audience areas and asked the coaches, "Wanna buy some water? Wanna buy some lemonade?" It was working really well, until then the park people found me and asked, "Do you have a permit?" I said no, and they were like, "Well, you have to leave. You can't be selling inside the park. You can go sell on a side street, but here you're competing with the official concession stand."

Anson (07:02)

Jasmine said, "Let the free market decide."

Jasmine (07:19)

Right, they were literally inhibiting innovation. The innovation being Uber for lemonade. I was DoorDashing that shit.

Anson (07:31)

That’s funny. I love the mental image of a bunch of kids in Sri Lanka running around with pieces of metal, schlepping them into greenhouses.

The first time I felt like I had free will was when I was quite seriously running this charity in high school. Everybody has their high school nonprofit era, but I took this so seriously that I definitely spent the majority of my waking hours on it. It was for helping children with disabilities, through organizing music and art therapy programs.

We were going on this massive fundraising bend because we wanted to hire full-time staff. I remember setting up job descriptions, writing them by hand and then putting them on Indeed. I was 16 at the time, and some of the applicants had Microsoft Word experience longer than I've been alive. I had to interview them and I would deepen my voice on purpose. I would wear a blazer, hop on the Skype call, and conduct an interview with a 40-year-old. Then, when I was raising more money than I've ever seen in my life: emailing property companies, with sponsorship documents and realizing that you can just be things.

That was the first time I realized that the world treats you with the seriousness that you attempt to treat it with. You can be serious and then the world will be like, this is a serious person, and then they will treat you as such.

I think a lot of Canadian young people can attribute a lot of their origin stories, especially if they're involved with philanthropic efforts, to Mark and Craig Kielburger. Their organization got into a bit of financial trouble, but they basically held these megachurch style concerts for grades three to seven. How do you get a generation of kids really excited about can drives? You bring in Selena Gomez. They did cross-country tours with celebrities talking about how you should do mission trips and sell bracelets. For all of its flaws, it was one of the first times that I felt an organization take people with youth seriously rather than using it as something to their detriment.

If you have less rules, more people will do higher risk things. But also if you seed people with the potential that they can do things, they will grow into it too.

Hudzah (10:09)

I thought you would mention the blueberry wine label.

Anson (10:13)

That's a good one! I forgot that.

I grew up on a blueberry farm. If anybody looks at my website and says, oh, she has a lot of projects, nobody has met my dad. My dad is the king of projects, much to my mom's dismay. One of his many projects as people on a blueberry farm is, what do you do with all this fruit after the season? You harvest it and make wine. So he started a mini-winery and needed someone to design a label. Who better to do it than his 11-year-old daughter? So I designed it in a dim sum restaurant, he printed it, and he made me write the descriptions too. It was before GPT. So you had this 11-year-old writing about the rich taste of the blueberry wine before I'd ever had a sip of alcohol. When it was printed out and on hundreds of bottles of wine, I was like, wow, I didn't know that was going to happen.

Jasmine (11:11)

I'm trying to identify causal relationships: not having rules, having parents who do projects, and having someone invite you into a project with them… Are there causes that you can attribute—whether role models, mentors, books you read, whatever—that made you believe in your own capability to do things from a very young age?

Hudzah (11:39)

I've noticed that some people act like a bank of agency: a BOA. You can kind of lend it and you don't have to return it back and it just grows and grows and grows. It self-replicates through many people.

For me, it was actually my dad too. My dad also would be like, "I just changed the house layout today." Or "I built another floor." Just seeing him do random things, like things that he shouldn't be doing or out of his scope, was just very amusing. Also seeing my friend Munir put on the suit and go to speak with some random hotel CEO. It felt like their agency just got lent to me, and now I can give it to other people. It’s a bank that has an unlimited supply. You just need one really good friend or person to light that fire up.

Anson (12:30)

I think that's actually really accurate. It's the gift that keeps on giving. For me, when I couldn't access it through people, although I very much did, it was through fiction.

Jasmine (12:41)

Were there specific books?

Anson (12:42)

Geronimo Stilton.

Hudzah (12:50)

I love Geronimo Stilton. Curious George, bro, that guy was high agency. Johnny Test? That guy was the highest agency person ever.

Anson (13:03)

In the media, we do a really good job of nurturing agency. But in reality, with how people raise children, it's less so. I know the word “infantilizing” describes literal infants, but even up to ages 15 to 20, these are the ages where people were previously sent off to fight wars. And now they still have to get permission to do certain things. I think people are much more capable than we give them credit for.

Hudzah (13:28)

I find that in different countries, it has a different effect. In Sri Lanka, even if you're 25, if you have no degree, no one's going to take you seriously. But at the same time, a lot of children when they're 12 or 13 will be doing groceries—they have responsibility, but only when it comes to manual labor. But if you're 14 years old and have a business plan, no one's going to care. So there is permissioning, but up to a certain extent. Whereas in North America, it's quite normalized to have 18-year-olds want to raise venture capital because they have to for the sake of university. If you don't, you probably can't put anything on your university application. Versus in a more third-world country, there's a lot of tolerance for young people doing manual labor, but not for actually using your brain.

Jasmine (14:28)

Even here, it depends what community you're part of. If you grow up in Palo Alto, that's where people know that you can do things like get internships in high school. But the vast majority of places, even in the U.S., the idea of starting a company or getting an intellectual job as a teenager is not super common.

Anson, how did you figure out how to raise money for your nonprofit?

Anson (14:58)

Bashing head against wall until it worked.

There were three ways that we tried. The first one is what everybody does—donut sales, the Krispy Kreme classic. But we realized that the margins are really slim. We had to recognize that the organizing that we were doing was valuable: people wanted energy, people wanted to contribute to social causes, people wanted tax write-offs. There were large companies who had funds that you could reach out, allocated budgets for charity. One of our mentors told us that instead of trying to get things out of their operational budget, try to get it out of their marketing budget. As a 15-year-old, I had no idea that these distinctions actually made really big differences. Also, public libraries and community centers were aggressively helpful. They had resources about large federal grants that you could apply to.

It was a combination of brute forcing, understanding that the funds we need were more than what we had immediately available, then having to get creative. I didn't think about it like this at the time, but it was looking at who the capital allocators were and trying to figure out ways that we could fit into their plans—finding the people that we needed to reach and then how to make it work for them.

Building creative community with Socratica

Jasmine (16:16)

I assume that for many people, each of you have now become the agency bank, right?

One thing that I appreciated when I was reading Anson’s blog and the Socratica site is that it represents “agency” as nurture, not nature. I don't like when people assume that there are “high agency” people and “NPCs.” This feels dehumanizing as a concept, especially because there often are cultural barriers or different ways people grow up. So I'm curious what you have learned about how to nurture agency in the people around you.

Anson (16:57)

I really agree with not liking agency as a nature thing—not necessarily for moral reasons, but because it's boring. It's really boring to assume that 1 out of every 10,000 people is the chosen one with the mandate of heaven, and everybody else, well, sucks to be you guys. I think it's so much more interesting if you think that there's a little vessel that is in every single person that's going to burst if you don't find a way for it to get out.

Humans are really good at copying what is around them. Some people describe this as the “adjacent possible,” where if you see someone in your peer group that is similar enough that you can see yourself in them, doing something out of the ordinary, then it’s much more possible for you to feel like that you can as well. That's really important. Then in Socratica specifically, we've gotten to the point where there's a lot of tacit knowledge that we've accumulated that we need to impart. For every project that we do, we try to bring on an intern to osmosis the knowledge.

Oftentimes people see particular groups as, these are the people that can do certain things, so they try to get into those circles. In reality, a lot of good things come from planting your feet, being a really good collaborator with your peers, and co-developing that within where you are. And I think that is much more emotionally fulfilling and gets you much further than looking over your shoulder all the time, seeing what circle you can join.

Hudzah (18:29)

On the topic of nurturing agency and Socratica, a lot of things we do now are not even about the project. It's like, we know this person is really good and they could exponentially become better. We just want to show them the way. I think of this as throwing someone into the deep end of a pool, and also then being the lifeguard to rescue them if things go haywire. Giving ownership to people is just so good, especially if you know someone is capable and can figure it out—you give someone trust, and they’ll go very far.

Anson (19:28)

For Socratica, there are no applications for anything. Literally, you show up and you keep showing up and then you get the job or become the host by just doing it. I think that's really important because it filters for the type of person that isn't looking for prestige. And it also signals some sort of flexibility, willingness to put in the work, and innate desire to want this thing to exist. It reminds me of how people say that sometimes the people that you actually want to be politicians never run, because they don’t want to go through elections and campaigning.

As hosts for this organization, one of the important things that we do is finding people who stay late, pick up the trash, and want to contribute to making random PRs on the website. Then we give them more responsibility rather than opening up an application form, spending a lot of time reviewing applications, all to do the application cycle again. We don’t see applications as a good use of our time.

As Socratica scales, this might break, but that's what we found to be really effective right now for having a truly delightful team of extremely enthusiastic hosts.

Jasmine (20:45)

I really like the metaphor of throwing someone into the deep end of the pool, and also being the lifeguard to come in and save them. Do you have a specific example of how you might approach that?

Hudzah (20:56)

For Symposium, that was the case for a couple of people. One of our friends was doing audio/video stuff. He was really enthusiastic about it, but was just telling me how he felt imposter syndrome. So I was like, "Do you want to just own the whole thing?" He said, "I don't know." And I was like, "Oh, just do it." He did really well, and he also had one of our other friends, Alex, with him all the time. So he wasn't alone, but at the same time, had so much ownership. He was able to do things how he wanted to. In that case, the lifeguard was Alex. They would also hang out in person all the time, so there was just constantly tacit knowledge being passed around. He was teaching Alex things, Alex was teaching him things, and it was very bi-directional. It wasn't like, this guy's the intern. He's contributing to a core part of this event that we absolutely need.

It’s similar for a lot of how we bring people into Socratica, and how we run the Symposium. There are people who are enthusiastic or who are very good at something. We just see how fast we can onboard them. We give them as much information as possible to absorb all the material. The person who really cares will read every Slack message and will go years back, just to make sure they know what they're doing.

incredible symposium video

Anson (22:20)

The other part of it is that for Hudzah and I, neither of us like other people telling us what to do. We have deep empathy for people on the team, where if you want to build someone who feels like they have a lot of free will, you actually can't tell them what to do. So how do you create an organization where they are incentivized to do things? We’re still figuring it out. We'll take tips as they come.

Jasmine (22:46)

At a lot of college campuses where there’s a culture of trying to gun for internships and a lot of ambitious students, people are reluctant to spend tons of effort and time on something that they can’t put on their resume. And one thing that from afar has really impressed me about Socratica is everything seems really high quality—it’s professional quality design, AV, everything—but it's not attached to titles or internships or things like that.

How do you get people to try so hard without those incentives?

Hudzah (23:21)

We try really hard to fight against it. It's not like it doesn't exist. For the longest time we were like, absolutely no LinkedIn. Last year we budged and made a LinkedIn, but even for that, everyone has to put “Host” as their title. So it's all equal. There's no roles, there's no hierarchy. There are implicit ones, but there's no explicit ones.

The way you actually get into Socratica is that you just show up. You show up on a Sunday morning, whether it's raining or snowing, and people recognize your face. Because there's no application and it's fully through effort, that weeds out that type of person. It’s hard to fake that you care for so many weeks per term.

Anson (24:13)

What people see publicly of the Symposium is part but not the whole of what Socratica is and stands for. We have to start at the roots: it really formed at the beginning as a place for people to do things that they personally really cared about, outside of school and work. It wasn’t big, it wasn't flashy. It was friends in a classroom. And it didn't matter if nobody showed up except for the two people, Aman and Adi, who first organized it. It was that beacon that they set out that attracted people with strong, similar values.

As Socratica grew, we integrated that philosophy that you would rather have a much smaller group that really feels this than try to grow as fast as possible. The community dilutes when people do things that they half-care about because they come for socialization. The events that people see now are three years in the making, of developing this counterculture. There were demo sessions and a small symposium and then a slightly bigger symposium and a bigger one and a bigger one. Along with different programs, like film festivals, summer camps. Hundreds of different people have touched Socratica and pulled in their equally committed and enthusiastic friends, having that flywheel develop. If a community is a really well curated basin, then you are incentivized to bring in people with similar values.

These values are very strong. If you are deterred by them, it's not the place for you, and you can leave. There's this concept of evaporative cooling in communities—that if you open it up, you should expect a lot of people to come in and the quality of the organization will deteriorate. But actually, if you keep it open, so that people can flow in and people can flow out, then only the people that really want to be there will stay. Those people are usually very committed. And then you can give the people that are committed more responsibility, and they can use this as a platform for their visions. Some people’s vision was a symposium. And some people wanted to do design for the symposium. Some people wanted to get 3000 shawarma for the symposium. And it was the compounding of many different individual visions over time that made this possible. So it became possible slowly and then all at once.

Jasmine (26:46)

Communities are like that. You have to plant a lot of roots before the action shows up.

One book I think about a lot is Twitter and Tear Gas by Zeynep Tufekci. She focused on Gezi Park, Tahrir Square, and how social media introduced a new model of organizing protests. While social media makes it easy for everyone to show up and create this giant spectacle, they immediately disperse, so it became much harder for these protests to build lasting institutions. Tufekci introduces a concept called “network internalities,” which are built when people have to show up in person and print out a bunch of flyers or go through an Excel spreadsheet and make phone calls together. A lot of this operational, boring, logistical work strengthens the internal bonds within the network, making it more possible to build strong institutions. You don't see network internalities from the outside, so they reveal themselves in what happens after the disruption. Are you actually able to build something long-term?

It sounds like by having weekly co-working sessions and smaller events, those bonds are strengthening and making it possible to do something big and visible and flashy like the Symposium.

Hudzah (28:06)

I think we first-principled exactly that. Anson was running these big postering campaigns across the university. That was actually one way we just brought in so many people, because we randomly messaged people on our Discord, and 10 people would show up out of a few hundred we messaged. They would do the boring work, but those people became very invaluable. Like if you can do the shitty grunt work, you'll be around for the fun stuff—and you deserve to be.

Jasmine (28:32)

HudZah, you mentioned that Socratica was the first club or organization that you ever joined. What was different?

Hudzah (28:40)

I only cared because I was able to go freely. Every student club requires an application, and that is immediately a barrier. I don't want to fill out applications. I'm not good at applications. I was always turned off because those incentivize people who know someone to get in, and at that time I didn't know anyone. For Socratica I just showed up every week. I joined and met friends and those people became my roommates and it went from there. If there were a barrier to entry, I would not have come.

Socratica also brings out strange people on the edges of the University of Waterloo. It's the people who are on the internet a lot. That also strongly resonated because it was like, wow, I found my people finally.

Jasmine (29:40)

What is the hardest thing about nurturing Socratica as a community?

Hudzah (29:47)

People are really hard to find. You see this echo in different ways, like even companies trying to hire people. It's hard to find the person to lead the charge or do the shitty logistical thing when somebody graduates.

There’s also the issue of not being able to hire—trusting people to do this, on their own time as a volunteer with a job and school on top of it. It's a bunch of responsibility, and it's an important responsibility, but the ball cannot be dropped.

Anson (30:19)

I only half agree. People are important and they're hard to find. But I don't think that we have a shortage of great people. It's when these are all strong-willed, idealistic and visionary people within one organization, how do you keep that cohesive enough that you can actually sustain an institution for a long period of time?

An organization like this doesn't fail at once. It fails three years down the line when something hasn't moved for a long time, or a core piece slips and there isn't a person there to catch it because everybody's a volunteer.

You buy a double-edged sword: You want to be a platform where people can do anything, but because of that, you are a platform that can do anything. When someone asks, "What is Socratica?" there's no answer, or everybody gives a different answer. That it feels like its most likely downfall. But it’s also the magic. So you got to pick the battles.

Jasmine (31:12)

It seems like demos are an important part of Socratica and what it does. I’m curious why.

Anson (31:25)

I have a personal strong affinity to demos because I likeearly computing culture, where community was built through demos. For a lot of people, it's much easier to talk about their work than about themselves. To have a third thing to look at helps people bond because you're sitting shoulder-to-shoulder instead of face-to-face. It’s the same advice you would give someone on a date—go do a first date where you don't have to be staring at each other all the time. You're at an arcade, or you're watching a movie. It's much easier to break that socialization barrier with a third thing.

On top of that, with companies, they say you build culture by who you hire, fire, and promote. In a community, you can kind of hire people by tapping people to come in. You can kind of fire people by telling them what they can’t do. The only way that you can promote people is by giving certain people more visibility or airtime. By being selective with who demos, you can curate the rest of the community by selecting for who actually gets to present.

So demos are really important for gathering people, for giving them an excuse to talk, and also for tuning how you want the group of people to be.

Using the internet to find your people

Jasmine (32:43)

It reminds me of what you said regarding peers who show you the adjacent possible. Demos are the way to say, these are some people you could model if you are looking for types of person to be or types of thing to do.

I want to talk about internet stuff. So people talk about the internet expanding agency or creating surface area for luck. For me, I often think of it as a way to manifest opportunities. If I say to the world, I really like this and I want to do that, you can be magically connected to people who care about the same thing. I really love the @Henrik Karlsson essay about a blog post as a search query to find your people.

I know that the internet has been important for each of you. I am curious how you view the internet, but also how it's changing. For example, for me the critical platform was Twitter. I wonder, is that still the place to go? What do you recommend?

Hudzah (33:45)

I think it's actually still Discord. Especially in my circle, a lot of people get exposed to Discord because of games. That gets people super comfortable talking to random people online because you're just playing video games and hopping on a VC with them, like a voice channel. Initially I used Skype, and then TeamSpeak, and then Discord eventually blew it out of the park as a way better product. I didn't use Twitter very much until the last year. Even when you first start college, you join the class Discord, and probably not social media.

Jasmine (34:28)

That's crazy. We didn't have Discord. We had a class Facebook group, but it was when Facebook was already dying, so it was very ineffective because we hadn't had a replacement yet. So we were on Facebook, but no one actually used Facebook.

Anson (34:44)

The West Coast lagged on its Facebook evacuation. We still had our subtle Asian dating groups.

I have this concept of “beacons and basins”: the beacon is the lighthouse where you put out a signal to the world saying what I’m interested in, what I like. The basin is a way to collect it. A beacon could be a personal website, your social media profile, or putting out messages in a channel; and the basin is the way to have an excuse to meet those people again on a recurring basis. So a book club, or a Discord channel, or a forum, or an email chain even.

In many ways the internet is worse, but in a lot of ways it's the same. There is still hopefully a person behind every username and every email. A good old cold email still goes crazy. The pattern of reaching out and having a way for people to find you will never fail. You might have to tweak it a little bit for the platform, but that’s the core of it.

During COVID, I tried to meet a person and have a call with them every two or three days. A lot of my formative friendships came from a secret Discord link at the very bottom of someone's blog post in the backwater of their personal website, which led me to a different Discord channel, which led me to a Slack. Now these are some of the closest friends of my life.

Hudzah (36:22)

In terms of also how platforms are changing, on Twitter you now have to pay to message people and that kind of already kills the whole thing.

Also, what Anson mentioned about hopefully the person has a face behind their account—I feel like that's not going to be true. If I had to be worried about something, it’d be an increased amount of fake people who are good at pretending to be someone.

Anson (36:43)

I do think that the retreat of the internet from open forums, even from feeds to group chats and closed discords, is one that concerns me for the sake of finding strangers that would have resonated with you otherwise. But in that case, you gotta adapt to the game.

Hudzah (36:51)

Gotta be in the group chat.

Anson (37:05)

The best way to be in the group chat is to start it.

Hudzah (37:08)

There was a really good blog post by @Sriram Krishnan on how group chats rule the world. And there was a recent post about a group chat between Balaji and Joe Lonsdale. There are these tech group chats on Signal and WhatsApp that are very dominant. That got me thinking, because everyone flooded away from public forums, so a lot of people who make decisions are now in more private circles.

Jasmine (37:32)

It was the Ben Smith Semafor piece on the tech right group chats. It was really interesting.

The way that I think about it is you still do things in public, whether that's Twitter or blogging or posting, in order to signal to people that you should be added to a group chat. One way to make your way into private spaces if you don't know people is to do things in public that show what your values are, and then that leads you into the hidden spaces afterwards. Having some proof of work in the world.

I'm also a huge cold email fan. If someone writes a blog post I think is good I'll just email them and be like, "I think your blog post is really good. Do you want to hang out?" I've actually made so many good friends this way. One lovely thing I've noticed is that my friendships where both people have read each other's blogs and meet that way—they’re always really good in person too, because you have built so much context about who they are. After you connect on the level of thousands of words of brain material, it's almost never the case that then in person you have nothing to talk about. So I will keep doing things in public because I want to signal to the world, these are the things I care about, come talk to me about them. I guess that's a beacon in your framework, Anson.

Hudzah (38:41)

I preferred your older metaphor of “bat signal.” Bat signal is a better word than beacon. It makes way more sense.

Anson (38:56)

We'll workshop it.

Liking someone's blog post is second base.

Jasmine (39:08)

I keep hoping that I'm going to find the love of my life via a reply to one of my newsletters. It hasn't happened yet, but someday.

What do you think are mistakes that people make when they hear, “You should be on the internet more”?

Hudzah (39:24)

I think it's very obvious you're trying to clout chase or farm likes or something. The two ways to go on the internet are that you either post something insightful or post something entertaining. If it's not either, people are not going to pay attention.

Anson (39:40)

That's almost exactly the framework I use. Every time I post something, I'm like, is this useful? Is it funny? Or will it make you feel something? If it's not one of the three, it's cut. I think the thing that people get wrong is when they copy the pattern of somebody else's speech or language. Sometimes you'll meet people and realize, you don't talk like this. It's like the writing test of, if you were to say these things out loud, does it actually sound like you? I would venture to say most people have pretty good personalities. And it's not knowing how to convey that properly that is to people's detriment. Actually, you could be good at the internet, but you're trying to be something that doesn't really exist or that people don't want.

Hudzah (40:33)

Also, I'm not a wordcel. I can't deliver insightful things. I'm not going to type up something that'll make people be like, "Wow, that is so deep." My insightful thing is "I made a thing" not "Here's what I think about the economy."

Jasmine (40:49)

I'm sure you have great and deep thoughts about the economy.

Hudzah (40:52)

I'm not as good as phrasing it for attention. I don’t know, there is an art. But I think you should probably write a lot. That's probably something everyone should be doing.

Jasmine (40:56)

I do think it's a great time to be a wordcel. But also, if you're not a wordcel, what's your way of writing a lot? Sometimes I worry that the advice biases too much toward, "write all the time," which I really like because I like writing and I'm better at it than I am at making things. But sometimes I meet people who are forcing themselves to update a blog because they heard that it’s professionally useful or it's a thing that smart people do. But that may not be the way they necessarily feel most natural presenting themselves. I think everybody has a medium or form of expression that is most authentic to them.

One thing I've liked, and maybe you two do this, is a personal website where instead of a blog with articles, you just list projects or apps. A portfolio as a blog, and that's your mode of self-expression. Or if you're an artist, you might have visual thing. I almost worry that writing is too valorized in Silicon Valley.

Hudzah (42:04)

Anson and I both have that for our websites. Mine is more like perfumes I like and photos I take and projects.

Anson (42:17)

A good way to think about it is that your website can just be emissions of life. Like here are my thoughts from while I made this project. Our friend Ivan literally posted his ChatGPT history, so you can search through his queries. Our friend Riley makes a bunch of things related to running. Making is a way of self-expression and it can come in the form of music, videos, writing, code. People underrate engineering projects as a way of expressing values onto the world.

It's the act of exercising, this thing didn't exist, I had an idea and now it exists. That's a muscle that's going to be much more valuable and salient in the next couple years rather than any specific skill. I think a full stack person is not just going to be front- and back-end, but "I had an idea. I was able to sketch it out. I made a prototype of it and now I'm able to talk about it." That is the real full-stack in the next couple of years.

Hudzah (43:17)

My friend Xavier listened to an album from every country in the world and posted that on ratemymusic and got a lot of attention. Or he makes Spotify playlists for niche artists like Death Grips, and they have hundreds of followers. People underrate the people who make those Spotify playlists. Someone has got to do it. And it's another basin.

Anson (43:41)

I'm really excited for this era of no longer feeling inhibited by tools, like code is not scary to me, I can make anything. I yearn for the day that my mom makes an app and she's like, "Look at the thing that I made, I can track my plants now."

A lot of this stems from the idea that people have something inside of them that is intrinsic and special, and the tools and people around them can enable that expression. That's a through line of Socratica, of our other work, of our involvement on the internet. It’s a philosophy of life.

Formative storytelling

Jasmine (44:16)

My gosh, so much good stuff.

Anson, we've talked moving between storytelling mode and building mode, which feel like different parts of the stack. You've mentioned that sometimes you feel like storytelling is more natural, but it's really important to also be building. For example, this company right now. I was curious if you could say more about moving between those worlds and why you see them both as essential.

Anson (44:38)

I'm very curious how you think about this because I find it very hard to talk or write about something when I don't feel like I'm in it. So in order to anthropologize well or to document well, you actively need to be participating in it. I realized I was storytelling and writing to ask questions, and I was engineering and doing operational execution to actually try to answer those questions.

Jasmine (44:40)

Maybe it has to do with the tacit knowledge embedded in “doing.”

I spend a lot of time thinking about tech journalism and why everyone feels dissatisfied with it. Part of it is, if you're a super outside observer, you miss a lot of tacit knowledge that only comes via being part of the process. I worked on product at Substack pretty explicitly because I had been blogging about social media dynamics and the sociology of platforms. At a certain point I needed to try building things in practice. You can have the most beautiful concept in your head that sounds really awesome and uses all these theories and is really ethical—and then you ship it and nobody uses it. Then you have to look at that and, well, it doesn't matter if you had an awesome concept. The DAU numbers don't lie.

Now I'm back in writing/theorizing mode again. But I try when possible to be close to people who are building. I try to use all the tools. I feel like you have to try.

Anson (46:08)

It's a really vulnerable thing to make something and send it out. Doing it yourself helps you develop deep empathy: in addition to knowing the actual details, you feel what it's like to be a person trying to make something real.

One of the main qualms I have with tech and science reporting is that you have two modes:

1. The first mode is, "This seaweed is going to save everything, it's going to solve climate change." That’s the silver bullet.

2. Number two is, "We're all doomed. This is super unethical." Someone releases a technology that is genuinely going to help a lot of people, and all people can talk about is, "This doesn't hit this very specific data standard.” But maybe I don't care because it can now diagnose a medical problem that I could not figure out beforehand.

Both modes misunderstand that whenever you're trying to do something different, big, or new, you have to get to a 70% point that will probably fail before you can get to the 100%. For the previous Starship launches, when they failed, that's all people could talk about—rather than the fact that this specific engine worked for the first time, or this booster had never been tried before and was actually able to hit X range.

I think that if you want to tell the story of science and technology well, you have to be in it. Because then you can actually empathize with the 70% steps that are necessary to get to the 100%.

Jasmine (48:10)

Also, things are almost never silver bullets in general. There is no panacea anywhere, technology or otherwise. All of social change is different incremental interventions that make things incrementally better and occasionally worse again. There's never been a panacea in history. So I don't really like "This is not a panacea" as a dunk.

Hudzah (48:18)

I'm also wondering how much of that is incentivized because obviously you want to report on the controversial thing or the bad thing. I feel like regardless of whether I was within the thing or not, I'm going to do the thing that makes me money, if it's my job.

Jasmine (48:40)

Both are definitely true. There are incentive problems and also there's an attitude thing.

I think there are values that you only gain an appreciation for through doing things. Building in and of itself encodes a set of skills and values, having to carry a thing to completion. For example, “scoping” is important because every project can get infinitely large. If you're building anything, whether that's hosting a conference or shipping a product, you will not be able to do everything, because there's a timeline and you only have so many resources. So you have to think through trade-offs and P0s and P1s and P2s. That’s something I feel like I gained appreciation for through being a builder or operator that pure writers don't always have an appreciation for—because when you're writing, you don't have to scope. You can theoretically want everything to do everything. It's not that I think the writers are being unfair, but just literally, if you've never been in the position of having to make trade-offs and scope, you don't recognize trade-offs as a value.

Counterfactuals are another one. One of my blackpilling moments around journalism—which obviously didn't blackpill me enough to not do it—is that when I was in college, I interned at a climate modeling startup. We worked with city governments to build simulations of how natural disasters would impact city infrastructure. It'd be like, if a flood of this severity comes, these are the buildings that we think will fall first, and this is how evacuation should work. There were machine learning and climate science people who would try to model this out. Then the New York Times, some Pulitzer winning journalist, did a piece on the startup. It ended up being a hit piece. A lot of the critiques were things like, they made a map of Seattle, but there was a missing Costco. This was a really big part of it. They made this thing of like, then emergency planners will get things wrong if you have this rate of errors, so this is really bad. But we had FEMA people inside our organization, ex-FEMA heads, and we worked a lot with planners. And they were using paper maps beforehand and drawing routes—color coding literal paper maps. Like, okay, we missed the Costco. But counterfactually, compared to paper, it’s way better. There are all these thinking styles or values that you literally don't learn if you’re never in the position of building a thing.

Hudzah (51:18)

Wow. Poor Costco.

Anson (51:20)

Somehow I feel like a lot of the people making videos and movies actually get it more. I think Cleo Abram does a fantastic job, Jason Carman, our friend Donald. Video, I think, is very similar to engineering in some ways, where you have your hardware equipment and you have to make a lot of trade-offs. And it's physically taxing. I find closer corollaries there. It's interesting to see how the form shapes the output.

Hudzah (51:50)

You also get to be there, you see things firsthand. Like, wow, this looks awful. This is so difficult. I love watching the behind the scenes of movies, because you gain an extra depth of appreciation.

One of those especially is events. Now that Anson and I have done big events, it's like, wow, that trussing looks insane. How did they move that? Or how did they get those lights up there? It's just a very deep appreciation for things that people don't think about, like concerts, how venues are set up, how a stage is set up. It's quite insane.

Jasmine (52:29)

Anson, you seem to really like video and film. I'm curious what you appreciate about the medium.

Anson (52:38)

It's pretty. I'm kidding. Film was for me some of the first pieces of media that made my jaw drop and my stomach turn. I very distinctly remember watching the AlphaGo documentary. I was like, I should be paying attention, this is very important.

I don't have a very intellectually stimulating answer other than it's really cool.

Jasmine (53:04)

I literally have the AlphaGo documentary open in a tab right now. People keep mentioning it to me and I'm like, I gotta watch it.

Hudzah (53:07)

It’s a banger. And the crazy thing is that it wasn't really advertised. They just kind of dropped it. I don't think Google even cares about the documentary or the impact it's had. The documentary has probably had way more impact than AlphaGo itself. There’s so much media similar to that. A lot of what VICE does, the old Johnny Harris videos on Vox. Those are impactful to a lot of people who knew nothing to figure out what’s going on in the world.

Jasmine (53:38)

I did really like the VICE videos growing up. I know people critique some of their journalism as being sensational, but I enjoyed that their journalists went places and met people and took that pseudo-anthropological approach. It’s what I like and admire most in observers of a scene—you got to go to the place and hang out with the people.

Hudzah (54:03)

One piece of media that was very formative for me was air crash investigations. I was so obsessed with air crash investigation videos; I would watch them every day at 11pm on National Geographic.

Anson (54:20)

Can you talk about that?

Jasmine (54:22)

I've never seen that in my life. Tell us more.

Hudzah (54:26)

So it's historic plane crashes. One of the worst air crashes after 9/11 was an Air India bombing that was going to Toronto from India. A bunch of terrorists had put a bomb on it, and it blew up mid-air. 300 people died. They would do the post-mortem and show the whole story behind why this happened. I was always impressed by the animation. Seeing really good CGI in 2010 was insane. Star Wars in particular, Pirates of the Caribbean. Those were just very interesting to watch the behind the scenes of.

Anson (55:03)

Wait, Jasmine, what are your formative media?

Jasmine (55:06)

A lot of it was pretty classic stuff. I was a real big Harry Potter kid. I read every book between 9 and 15 times each. I had a poster where I kept track. The first time I read them, I read them out of order because I was a library kid and they weren't all available at the library in the right order. So I just read them in the order that I got. Ender's Game was another I read at least 10 times. The Mysterious Benedict Society.

I feel conflicted because I wonder what messages get ingrained when you read too many “Chosen One” books. I read this Tanner Greer essay in City Journal called "Escaping Only So Far." It talks about a set of tropes within a certain generation of young adult literature where there's some powerful dark cabal of adults who are evil and bad. Then some child is this prodigy “chosen one” who gets sucked out of their normal life, and they reckon with the morality of these adults wanting to manipulate them. Do they succumb to that or do they start a revolution instead?

For better and for worse, a lot of my formative media was things of that type. But I don't actually know if it's good for building free will. Chosen One stories are very anti-free will in a lot of ways. And it sort of makes you conspiratorial, assuming that there's always another room that things are happening in, rather than developing the sense that this is the world as it is, there's not a secret cabal, and anyone could be the chosen one. Everyone has the vessel inside of them that can be developed. But it was all escapist and fun and I certainly enjoyed it a lot.

Hudzah (57:05)

I never thought about that. In Harry Potter, they literally segregate people by whether you're a muggle or a half blood.

Anson (57:16)

In recent years, I've been looking for the stories of squads. I want to hear about 12 people who work together and have friendships over the span of many years. Those are so interesting to me, but exceptionally difficult to write. So whenever I find them, I’m like, I’m tucking this one away. One that was recommended by a friend is The Mandarins, which tells a story of six people throughout time, through revolution, through their careers.

Hudzah (57:52)

I think Enid Blyton wrote a lot of those, like the Secret Seven. I read a ton of that genre—a band or a ragtag group of people coming together and solving a mystery, versus a singular person. It was quite good.

Anson (58:05)

Maybe this is a dating app question: What is the psychology that has been instilled in you by your early childhood media?

Education in the LLM era

Jasmine (58:15)

Since we're on the topic of childhood development, I'm curious how you think that education ought to change as the world changes in the next 5 to 10 years. I remember reading on your blog, Anson, that you thought a lot of your engineering classes weren't very well-done. In the ideal world, what ought education—broadly defined—look like?

Anson (58:45)

What education right now is really good at is providing a lot of content. You end up with a thin layer of content, and hopefully you can create this grand narrative. This is how the circuits fit into XYZ or whatever.

But what I would have really liked is an even thinner layer, and then a T, where for this very narrow sliver of a thing, you need to learn all about it. You need to build it from scratch. You need to figure out how to sell it. Go through that entire process. Because especially now, you're not lacking any content. We have on-call PhDs in our pocket, which is really awesome. What people need is confidence and how to have content interact with the world. School should be a place for you to embark on these things with people that are also doing it at the same time rather than being primarily a place for you to absorb content.

Hudzah (59:46)

A lot of my views on this focus more on the educators. The subjects I've always cared the most about are taught by people I've really liked. With a future where education can be so personalized, we should be taking advantage of a person's personality, the way they learn, and the content they care about. I think you can teach anything to anyone because children are naturally curious. The common argument says we should be teaching how to do taxes in school, but nobody gives a fuck about that. No one will do that class. The problem is that we inhibit people's curiosities early on and shut down a lot of things. All my favorite subjects in high school correlate to having a really, really good teacher. For me, it was computer science and chemistry. And I care about those to this day. They've been very, very impactful for me.

The way they teach the content also matters. Something that has changed my perception a lot is the "A Mathematician's Lament" essay by Paul Lockhart. That was a very interesting approach to how math should be taught.

Jasmine (1:00:56)

What made those computer science and chemistry teachers so good?

Hudzah (1:01:00)

Their first class was how to learn. He taught me very early on the actual way to learn so that you don't suck at exams. The second is that he had a very practical approach to things. Our chemistry class was a whiteboard, seats, and then all around us were chemicals. If the teacher wants to show us something, he just grabs a reagent and starts pouring things. It was just so cool, seeing stuff, and very fun. And he very much encouraged deep questions. Naturally humans don't want to answer all these questions. AI is a very good application there. But he just accepted a lot of questions and debate. He allowed people to be curious, especially the ones who cared a lot.

Anson (1:01:41)

A big thing for kids especially is responsiveness. I would settle for a worse answer if it came to me faster. And I think that most people feel that as well. Having access to LLMs now, especially over the past year, I feel like it's made me more curious. I feel myself taking my phone out and asking, "What is this thing on the lamp pole? Tell me about this bug." That feels distinctly different from 10 years ago.

When you have an abundance of responsiveness and information, then the things that become really important are emotional connection and care. If you free up the teacher's ability to actually see if the student is struggling because they are frustrated, that becomes a much more important part of education. Also, making friends. You bond with people over arbitrarily complicated assignments and staying up until 3 a.m., trying to get something in on time. Those are the parts of school that are intractable and shouldn't go away.

But I'm excited for the day where you feel like you can learn anything to a much greater depth because you actually have someone to ask questions to, to reframe it in the ways that resonate with your brain, and to not feel embarrassed or scared. So many people I know were completely deterred from subjects because they felt stupid asking questions.

Jasmine (1:03:06)

I totally feel you on the speed of response. I've noticed the same thing with myself for LLMs. Sometimes people will say they provide only a really low-level summary, or they’re not as good as reading a whole book. But the rate at which I ask questions is 10 times as much as I used to, because I was never going to go read a book about my stupid question. I was buying oranges from the grocery store and I really wanted to know the difference between navels, cara caras, and blood oranges. I was talking to ChatGPT about my oranges, yet there's no world in which I would’ve read a book about them. Now I feel like every time I'm very mildly curious, I know I can get an instant answer.

Anson (1:03:48)

I feel like an annoying kid again. I'm like, why? Why? Why?

Jasmine (1:03:52)

Yes! Why, why, why? Why is the sky blue? Those kinds of questions. Sycophancy problems aside, I love that this thing will never get angry at me for asking really dumb questions.

Hudzah (1:03:56)

The flip side of the coin is when you're sharing your screen, you should not show your chat history. Anson and I were on this call and I was showing my ChatGPT history and it was like, pufferfish mating ritual, or some shit like that.

It also seeps into the real world. We were talking with a friend and Anson was like, "Dude, you're prompting her." Because I was asking her to explain it in a simpler way. I don't know if humans are going to become more receptive to answering lots more questions.

I don't know if I'm in a bubble or not. That's my biggest fear. Is the strategy to get all the kids early? Do you want to give your kids a teddy bear LLM?

Jasmine (1:04:49)

I've been asking parents this recently. Because I'm writing an essay about people using LLMs to code things for themselves or their family. I've been trying to do more parenting examples and learn how parents are thinking about exposure to LLMs.

Anson (1:04:51)

What is their response?

Jasmine (1:04:52)

A lot of the parents are like, "We can use the LLMs together, but I don't want them to be left alone with the LLM too much."

One parent was like, "I don't want to have too much of an instant gratification loop around, for example, image generation where they never learn to draw by hand because they just think that the LLM can instantly create a drawing that is better than something they could do by themselves. I'm worried that it might demoralize my kid to not want to learn because the LLM can do it better."

Another parent who has a seven-year-old son was like, "I've been encouraging my son to use them. We're gonna create a video game together." He had first vibe-coded a game for his son to play 20 Questions, because his son loves 20 Questions so much that he will play infinite numbers of times. Then the son got really into it, and now they're going to make the next game together. That sounds so fun, you know?

But it is a hard question, because if that expectation of responsiveness affects how you expect humans to be, I could see that becoming more problematic. I love that the LLM talks to me all the time, but because we grew up in a period before them, I know that my friends will not text me back instantly, and will not always affirm me all the time.

Hudzah (1:06:19)

That's a good point.

Anson (1:06:21)

That's a really interesting prompt about what a child-locked LLM could look like. Maybe slower response times or "You sure you want to look at this?" or the equivalent of SafeSearch. There's probably a whole company in here around how to child-proof LLMs.

Hudzah (1:06:26)

I wonder if this would tie into interpretability research. You change the neurons that it needs to be more child-safe, or to respond in a way that encourages thinking and pushes curiosity as opposed to providing an answer and stopping. It's like, "Why do you think the sky's blue?" It could guide you through a process as opposed to instant gratification loops. But it's also not that hard to download ChatGPT.

We were also talking about these LLM teddy bears. I wonder if those memories from whatever LLM is inside that will grow up with you. Once you've upgraded to ChatGPT from your teddy bear, will it have all that past context? Almost like a nanny growing with you, if it remembers all your quirks and things from 20 years ago. It's the ultimate context.

Jasmine (1:07:24)

If you switch therapists, your old therapist has a big packet that they can give to the new therapist. My therapist had some rule change and she was like, "If you want, I can introduce you to a new therapist. I have a file of everything you've ever told me and I could have a meeting with the new therapist and also send them all the files."

It's also interesting to have interoperability regulations for LLMs. I could totally see the case made so that you don't get locked in just because you grew up with one company and that company becomes exploitative or enshittifies or stops improving their product. Don't you want a way to transfer your data and those memories to a different LLM? That seems like a super legit thing to be concerned about.

Anson (1:08:17)

On a somewhat consistent basis, I download all my chats. That export option feels like a ground zero—everybody should be able to do that no matter what. My grand hope that I don't think will happen is having a local model and a pool that all of your queries are passed in through that contains all of your context, like a data lake feeding it everything. But I’m dubious whether local models will be high quality enough where people would choose them for the sake of privacy and interoperability.

Hudzah (1:08:53)

I think it's less about how good local models are and more that it's so big, in terms of storage, that no one's going to download one.

There was this recent tweet that I was losing my mind over. It was the “Sign in with OpenAI” thing. It just makes sense. In the hands of a singular company is kind of sus. But we do that with Google and with Microsoft. Is that the future we're heading towards? And if so, is it that bad?

Jasmine (1:09:05)

I think people feel the pain of all their stuff being with Google. Lock-in is real. What is the “Protocols not platforms” model of AI? I would be very excited about that.

Mapping AI dystopias

Hudzah (1:09:25)

I'm more concerned about if kids grow up with LLMs like that, are they gonna talk to anyone? I've been lurking on the Character AI subreddit for a few years now. Whenever Character AI goes down, everyone loses their fucking minds. You've never seen this much PMF ever. People are spending 16 hours a day on Character AI. If you nurture that from a young age, you're kind of fucked. At least for children, it seems like a very obvious path to have LLMs be controlled.

Anson (1:10:09)

I'd be really curious if there's a parallel to how a lot of the people working on social media apps don't let their kids actually use it. I don't think that's the case right now because to LLMs are a really interesting piece of technology where the benefits are so clear. But I’m imagining the Character AI lawsuit for the 14-year-old that passed away, and Character AI then advocating for First Amendment rights for their models.

Hudzah (1:10:47)

The Character AI thing concerns me. And I think it concerned a lot of people at Character AI for them to leave and go back to Google.

My concern about LLMs is not AGI destroying everyone—it's society getting fucked up on a mental level and developing a great overdependence.

Jasmine (1:10:55)

One of my biggest pet peeves right now is how everyone's like, "Oh my god, the new 4o update is so sycophantic. How could this ever happen?" I'm not saying that AIs aren’t misaligned with humans, but the market is not always aligned with human values and corporate incentives are not always aligned with human values. That problem is hitting us now, way before the AI itself gains goals and starts to deceive us. Sam Altman had that tweet, "algorithmic feeds are the first at-scale misaligned AIs." I think that's very true. The thing that the platforms want is different than what you want, or your short-term values are different than your long-term values. Right now I want to watch a TikTok video, but in a broader, longer term sense, I don't want to spend 10 hours a day on TikTok.

I wonder how many of these people spending 16 hours a day on Character AI would say that they in a longer term, high-level sense, feel satisfied with that, versus just constantly satisfying the short-term desire to talk to this thing.

Anson (1:12:03)

It makes me think about what you said before—that you can create a beautiful system that's super ethical, but not everybody will use it. You have to toe the line. If you’re Claude and make something that's too safe and curated and moderated, then you won't get wide-scale adoption. You create this perfect thing that exists within a bubble, but it doesn't achieve your goal of safety because everybody else is using a different model.

Jasmine (1:12:34)

There is an extent to which companies inevitably are going to pursue company goals. I'm not saying they're evil, but that I don't think we should rely exclusively on self-regulation or pressuring the companies. OpenAI is going to tune back the sycophancy, but fundamentally, they are still measuring how much time you're spending on the product. When I put my PM hat on, the killer feature that Character AI is already doing, but OpenAI hasn't yet, is to have AI message you proactively. The AI should be sending you push notifications to start chats about things it knows you are interested in. That will be such a sticky loop. I'm sure that the PMs really wanna do it, but they're like, "Can't go there." But we as a society should be prepared for the fact that this is going to happen. There's no way that that doesn't happen.

There are different forms of building social resilience. One is personal ethos. For me and my family, I might try to teach a certain set of discipline in the same way that healthy eating is a discipline. Then there's broader social norms. You might have rules at the level of schools, like no phones in classrooms. And then legal regulation is the strongest form of it. What’s on my mind is the question of, how do we build social resilience to the fact that technology will give us everything we want really fast and cheap?

Anson (1:13:56)

My friend Donald and I were talking about the concept of cognitive security. You would have good security hygiene by not sharing your passwords, using a password manager, and using 2FA. So what is the equivalent of that for your mind? Maybe it's having a group chat where you can actively debate things, or having deliberately built in pauses where you question whether or not things are correct, like auto fact-checking mechanisms. As much as I don't want the onus to be on the user, it definitely will be a skill to cultivate to have really good security for your brain.

Hudzah (1:14:31)

The thing I think about a lot is that it's hard to get people to combat this because all the people using TikTok are, at the end of the day, fighting a war against a research team getting billions of dollars of funding to make it addictive. Breaking out of the loop is hard because in every single way, the whole goal is for you to keep using, regardless of whatever apps you're going to use to block it. So an LLM prompting you, I mean, it's already happening. ChatGPT prompted me the other day. You can set a recurring task for it.

Jasmine (1:15:05)

I started setting these tasks for myself, which is how I got on the topic. And then I thought, wait a second, they're definitely just going to start doing this for everyone. That's what I would do if I were them.

Hudzah (1:15:13)

Some people are trying to argue against that. Like, "OpenAI is not going to do that." But it's in every incentive—your customers want that, your investors want that. It is inevitable that it starts messaging you back.

Anson (1:15:30)

Beyond that, a conversation that HudZah and I had before was about how people are finding different products or services now is not through Google SEO but through ChatGPT mentioning it. And you can just tweak things ever so slightly to, say, feature Waymo cars more. This is the new form of advertising: it just appears higher up in your response packets and we don't have any visibility. That is a slight fear that on a system level it can prompt consumer behavior on this.

Jasmine (1:16:02)

OpenAI just introduced shopping. I saw a Greg Brockman tweet; they rolled out shopping embeds.

Anson (1:16:09)

In spite of everything, I'm still optimistic that these technologies can be the most transformative thing of the past couple decades. HudZah and I are personally super dedicated to making sure that the people that want to become more embodied or happy or whatever with these products should be able to do that. I have a lot of faith that these things can be really, really good. And if we don't push for it, the default is going to be that people are psychoanalyzing in their homes and talking to their AI girlfriends, but it can be so good. It can be so good.

Jasmine (1:16:42)

I agree with you. This is why I'm like, I don't want to slow down the progress of the development of a thing that I think is really amazing. I want to speed up the way that we as a society are having conversations about how we will adapt to it and build cognitive security and build resilience.

Hudzah (1:16:58)

I just had a look at the shopping thing. I had no idea this was launching because some people we know are working on it. I wonder if we're getting back to the slop era of, "These are all the shitty products you need to buy because people use the OpenAI ad waterfall to sponsor it." I mean, advertising is such a good business. Google AdWords makes so much money.

Jasmine (1:17:20)

We wouldn't have LLMs if advertising wasn’t so lucrative. All that research was funded by ad money.

I have an idea now. Right now, these AI companies have safety teams that are quite influential, plus product teams that are full steam ahead. That friction results in not going too insane. But I think it would be a useful exercise to map out in a world with no safety teams and just pure Silicon Valley product thinking. In that world, what do we think the next five years of OpenAI as a product company would look like? What are the things that they would build just based on good product ideas and market incentives? That might produce some ability to anticipate products that roll out. Again, it was surprising to me how surprised people were by the sycophancy update. How could you not expect this? How could you not expect shopping? There's a lot of researchers in AI, and because they're more academic types, I don’t think they anticipate the strength of market incentives.

Hudzah (1:18:30)

The other thing you'll also notice is that with 4o the sycophancy is there, but for o3 the sycophancy is not there. But no normal person is going to use o3, no normal person knows any model exists beyond 4o. And so if you know about it, then you're fine. You never actually see the bad side of things.

Anson (1:18:45)

It does feel like there's a chasm where if you cross it, you're good, you're safe. You use the more advanced models, you had a period of cognitive development before these became really popular. And you almost have vaccinations against it.

I think the OpenAI product map is going to map pretty closely to that of specifically Google and Facebook. Facebook now is working on Facebook dating. That'll probably come up. Marketplace shopping, single sign-in with your memory to different apps. This is the new bingo board.

Hudzah (1:19:14)

I think the next will be travel.

Anson (1:19:30)

Maybe. HudZah, our friend Brian, and I, have a predictions list of things we think will happen, and a point system attached to it. If it happens, you get this many points. So we could do something similar.

Hudzah (1:19:31)

Just look at what Google has and then OpenAI will have it soon too. Travel makes sense, payments, you link your credit card to OpenAI and instant buy. Shopify pages may not exist. It'll all just be direct embeds.

Anson (1:20:00)

I feel like people are misunderstanding what is actually going to suck about AI. The enshittification of the internet is probably up there for me where you don't need a human interface layer. In Minneapolis, they have a walkway and then a highway and it's completely separate. The humans walk on a layer where they interact with no cars and the bottom layer is all cars. There's going to be something similar where just, "Here's the ether land where all the models interact and they have their agent to agent interactions. And then here's the cozy palace where humans interact." You have your group chat layer and then you have your public open web layer that is just all agents. For websites, there's probably going to be like a llm.txt where it's like "I want you and your model to read all of these things about me, here's a preloaded context and my payment details, just take this and run with it."

I think companies are catching on, where Cloudflare has the ability to send bots and agents into an infinite loop to trap them. The internet itself is kind of resisting it, but everything else is pushing towards this bifurcation.

Hudzah (1:21:10)

Another product that they've already replicated that went over everyone's eyes is OpenAI Canvas. That is literally a replacement for Google Docs. You can bold. You can add headings. When is the Excel part coming out? If students already are writing their essays on ChatGPT, you might as well let them edit and comment and share and collaborate and do everything on there. It'll be a very beautiful experience because it's all in one place. That feels like the everything app, ChatGPT.

Jasmine (1:21:37)

Or I'm thinking about ads, but instead of scrolling through influencers on Instagram selling you stuff, you're scrolling through pictures of you that look exactly like you, but in the clothes holding the thing. You don't even have to do projection and aspiration, which is what happens on Instagram now, where you think “What if I was that influencer?” The feed will automatically show you as that aspirational person and exactly what to buy to achieve that lifestyle. God, this is coming.

Anson (1:22:08)

A microsite that I think that you should make is the “alternate universe app store” where you vibe-code small versions of these apps that people can go in and play with, for good or for bad. There are fashion apps now where you just dress yourself in a doll. And some people really want that. I think that's fine, but one of the best things that communicators can do now is help people not feel surprised.

Hudzah (1:22:40)

But nobody reads anymore.

Anson (1:22:41)

That's why you have to make fun microsites.

Jasmine (1:22:43)

That's fun. Before there was dystopian speculative fiction, but now it’s like this future AI internet. And you could go around the little world and experience that. Hopefully people would have reactions of, "This is really fun, I really like that" versus "This makes me feel icky and gross."

Again, I really like LLMs. I do not want progress to stop. But it's like the Three Mile Island thing where if people aren't prepared or a really bad mistake happens, there will be a winter. People will overreact. There will be social unrest and I don't want that to happen for a million different reasons. The more prepared people can feel, the more resilience can be built in advance.

Hudzah (1:23:30)

I just had this idea and it's like, one of the biggest things about ad waterfalls is targeting, right? If you have everyone's memories and their chats, you know exactly what they like. Facebook's whole algorithm was that it just tries a bunch of things and then eventually finds the right demographic. Now OpenAI just knows who it is, and directly targets that person, irregardless of demographics. You're one prompt away from finding your right audience. I wonder how people would A-B test prompts to run through an ad waterfall. That’s probably a very profitable idea for someone to build.

Anson (1:24:07)

Free ideas, guys. Free ideas.

Jasmine (1:24:09)

You write out your dystopian prediction. Then the aspiring YC founders swarm in and use it as their list of what they’re going to build. Every one of the dystopias becomes somebody's new project.

Hudzah (1:24:23)

I mean, is it a dystopia if somebody gets the product that they wanted?

Jasmine (1:24:29)

I think that's a hard question, right? The thing about the market plus technology is that it satisfies your short-term wants at faster and faster speeds. It is giving you what you want. It's not tricking me, you know?

But are there technologies and products that serve our long-term wants and have business models that do that? I think there are. For example, gym memberships or anything subscription based tend to be more long term because you're paying on a yearly basis. So subscription business models are more aspirational. When you buy a year long membership at the gym, you're saying, "I commit to wanting to be the kind of person who goes to the gym." When people pay for a Substack, you're aspiring towards being the kind of person who consumes that news because you paid for it. That is very different than an impressions ad based model, where it's like, "Did this content satisfy me in this half a second moment?” versus “I want to read this for a year because it will enrich my life long term.”

Hudzah (1:25:32)

Regardless, we're going to move towards whatever we're imagining. Human consumerism. It's too strong, and there's just too much money to be made. But it'll be interesting to see.

Jasmine (1:25:48)

That’s the theme. We are in interesting times. There are ways for this to be amazing and fulfilling and very creative and fun. And how do we build the will, the community, the infrastructure, to take advantage?

Hudzah (1:25:48)

I think it'll still be more good than bad. Like Google, they do ads. Made trillions of dollars off of it, but I'm pretty grateful they exist. I like YouTube a lot. The fact that we’re able to get this thing that can spit out a bunch of words and give you a great answer is incredible. But we will definitely go through the transition.

Research rabbit holes

Jasmine (1:26:21)

Cool. The last question that I always ask folks is: What is a research rabbit hole that you are exploring right now?

Hudzah (1:26:26)

There's so many, where do I begin? The first one—a friend brought it up to us in DC—is that there's potential for a ninth planet in our solar system. They're installing a new telescope in Chile that's gonna start looking at it. It's the biggest one, bigger than the Hubble or the Webb, to look at this part of the night sky. Soon enough we'll know if there's another planet in our solar system lurking around, which is insane. One of our side projects was we wanted to go find this planet, through looking at the data and see what we can do. That's my main one.

Do not go through my prompt history, though.

Anson (1:27:00)

Anything about the Concorde. Anything about 1900s Chinese history. For a while, HudZah and I shared a Claude account and it was really funny. Mine were radio, radio, radio. I'd try to make cookies and be like, “How do you brown butter?” And then HudZah would come in and be like, "So how do goldfish?" It's really fun to see someone else's mind like that.

For me, the biggest one for the past year has been radio—getting my ham radio license, and playing with the software-defined radios, tuning into different frequencies. It's been fun in a way that feels early internet. You're tuning into something and you're like, "Wait, there's something there." It makes me respect the engineers from the 1950s so much because how the hell did they figure this out?

Hudzah (1:27:50)

I have this very long notes app with things to build, and it's hundreds of ideas at this point. I can't wait for the day that things get so good, I can give the whole notes app and everything gets made. That would be my dream. I want to make these things but it's just not high enough priority, or it’s too expensive or something.

Jasmine (1:28:02)

Sometimes I feel like I get sniped into wanting to be a product manager again, because I'm like dude, I could make a killing off ChatGPT wrappers. I sometimes think about my alter egos. One of them is a ChatGPT wrapper founder where I become a real Nikita Bier type. The other is doing crisis comms for large, evil corporations. I think I would be quite good at that.

Hudzah (1:28:42)

This is dark Jasmine Sun.

It’s probably worth mentioning the thing Anson and I are working on so we don't sound evil. We're trying to work on a hardware copilot. We're trying to solve problems that are immediately useful, just because I think anyone watching this will be like, wow these guys are really stupid.

Anson (1:28:59)

Which could be true, but we're trying.

Hudzah (1:29:06)

We have a lot of ideas. And it's good to be an idea generator in this day and age. The idea guy is going to win.

Jasmine (1:29:10)

Huge era for wordcels. Huge era for idea guys. Amazing time to be alive.

Cool, this was so much fun. Lots of tangents, but a really fun conversation. I appreciate you both coming and chatting. Maybe a year or two in the future, once your startup is up and running, we can do another session and talk about that instead.

For listeners, you can access the transcript, the links, and be notified of new episodes at jasmine.substack.com. And if this was interesting or useful or funny or insightful, please share it with a friend. Thanks so much!

Links & more

• Geronimo Stilton, Curious George, Johnny Test

• “The adjacent possible” (Stuart Kauffman)

• “The Evaporative Cooling Effect” (Hang, 2010)

• Twitter and Tear Gas (Zeynep Tufekci, 2017)

• “A blog post is a very long and complex search query to find fascinating people and make them route interesting stuff to your inbox” (Henrik Karlsson, 2022)

• “Group chats rule the world” (Sriram Krishnan, 2024)

• “The group chats that changed America” (Ben Smith, 2025)

• Videos by Cleo Abram, Jason Carman, Donald Jewkes

• AlphaGo - The Movie (Greg Kohs, 2020)

• Videos by VICE, Johnny Harris

• Air Crash Investigation (National Geographic)

• Harry Potter, Ender’s Game, The Mysterious Benedict Society

• “Escaping Only So Far” (Tanner Greer, 2021)

• The Mandarins, The Secret Seven

• “look what the cat brought in” (Anson Yu, 2023)

• “The Mathematician’s Lament” (Paul Lockhart, 2002)

• “Protocols, Not Platforms” (Mike Masnick, 2019)

Thanks for reading or listening!

Jasmine

1. This is my first time doing a 3-person podcast on Riverside, and I think some of the audio/video ended up kind of choppy as a result. Hopefully it isn’t too obtrusive; I’ll figure out how to edit better next time!