Linus [Processed]

Haroon: [00:00:00] Welcome to Humans and AI, where I interview world class AI leaders and creators to uncover best practices and insights for navigating the world of. I'm host Torin Audrey, the Human Behind, not a bot and founder of AI for anyone. Today's guest is Linus Sem, one of my favorite builders in the space. Linus is a co-founder of Bedtime Story and he's a CMO of Con and he's also the proud father of two girls, hashtag Girl Dad.

Haroon: Linus and I chat about his journey to becoming an AI creator, projects he's worked on and the projects he's most excited about. His favorite AI tools and how he finds them, and much, much more dived, right.

Haroon: linus. Welcome.

Linus: Thank you. Thank you, family.

Haroon: Of course. Super excited to chat. So before we get started, do you mind just giving a quick overview of your background, what you do now?

Linus: Yeah, sure. So my background is . . I call myself a [00:01:00] digital janitor. I've been doing a lot of things in, in, in the digital realm. I've been on the internet for the past 20 years. Briefly, I got started in edu, like agency land. I worked for a bunch of smaller firms and worked with clients and took myself from there to start my own studio.

Linus: And back in 2012, I sold that studio started my first SAS business. Failed miserably at that building on top of Instagram's not public API at the. But it was really good that kind of pushed me into the trajectory of being in early tech, like whatever's happening. I've been close to the drama and the fire so fast forward a few more years. I went to Typeform. I was one of the first employees or first design employees there. I got hired directly by one of the founders and then I've been doing a few other startups. Recently I was at floes com and another. Rather successful SaaS business in the email space.

Linus: And then last year I decided to stop [00:02:00] working for others. I've been going back and forth being an employee and starting my own things. And yeah, I decided to go full on, do AI stuff. I didn't really know what I wanted to do. I wanted to build with the recent, the latest stuff out there. Yeah.

Linus: And that, that pushed me into this trajectory where I am now currently building a suite of tools not necessarily. Many tools, but I'm building a bunch of things and the most prominent thing that's public is bedtime story ai, which is a story generator for kids, essentially, or parents and kids.

Linus: Yeah, that, that's basically it.

Haroon: Yeah. That's awesome. And so when you decided to make that jump I didn't realize so you completely left everything and you just said, I'm just gonna work for myself now, did you have a plan at that point, or did you just land on the whole AI tooling?

Linus: No I didn't really have a plan. I didn't have a plan the first time I started something either. I was just like, fuck it, let's go. Sorry, that's a swear word. But I felt I was done with the whole kind of working for someone else, like selling my time. It's this time that I can't get back, I just got, got my first kid and I was like, [00:03:00] I need to do, I need to figure out a way to work for myself and I take whatever it takes, right?

Linus: And I've been like, I've been fortunate in my career. I've been fortunate to work in great companies alongside great people. I've managed to find myself financially in a good position where I can take leaps of faith. Not worrying about paying my rent and putting food on the table for the family.

Linus: So a bit of smartness along the way have put me in a good position. So I wasn't really worried, but yeah, I didn't really have a plan either. I was in the mood of just let's build stuff, throw it at the wall and see what happens. Yeah, that's kind, that summarizes my line of thought,

Haroon: Yeah. Interesting. And given your background, it seems like you already had a lot of the, I guess you could say, the skillsets that are required to just build software. And so you mentioned you were designer for part of your career. I'm assuming you did front end design

Linus: Yeah. So yeah, I've been a product designer, so ux, ui, the whole term, like for me it's like I've been a designer taking stuff from blank canvas [00:04:00] to finish product. So these days you might not find many of us, but we are there, there are definitely more skilled people than me out there as well, but like to be able to go from ideation working alongside teams, building, with the necessary tooling that you need to do and, all the way to the, getting your first 100 or a thousand customers.

Linus: So I think that's been, it's really hard to put a finger on okay, I'm good at. That particular job in design, I'm more of realistic person. I think that's my power or the, what I'm good at. But sometimes that's also bad because there are plenty of companies that don't have space for someone.

Linus: That's more of an entrepreneurial person than someone that fits the mold and can go particularly into a niche problem and be spearheading like a specific task in a team, for example. That's not me. I'm kinda the jack of all trades that just go in somewhere, solve big problems and then hopefully by the end of it, still enjoy it that much that I want to be around.

Linus: Yeah, so again, very fortunate to have been able. to to see a bunch of companies go [00:05:00] from very small size to large size, but also been solving like some really heavy problems in these companies. So they're super excited about that.

Haroon: Super cool. Yeah, and that's really interesting that you say that too, because I was chatting with Masad, who's c e of Replit just a couple days ago, and one of the things that he said was that the entrepreneurial types are gonna be the ones that are gonna succeed in this changing world.

Haroon: Because you had a lot of skill sets to start with that. difficult to attain. Super powerful product design being one. Just being able to take an idea for an application and being able to build it out. I think that's enormously powerful. And in the age of ai. With all of these AI tools emerging, it seems like it's becoming easier and easier, and the mode of, having a software engineering background or having expertise and knowing a particular tool or whatnot, that seems to be chipping away.

Haroon: Is that what you're seeing as well? or are you seeing it differently?

Embracing AI

Linus: Yeah I think I am, for the first time in my [00:06:00] career, like honestly, hands to heart, like the first time in my career I'm really worried about actually being out of a job . So my response to that is to try to embrace this new technology shift and try to get as close to the action as.

Linus: Possible with what I know and my skillsets. I'm not a researcher, I'm not a machine learning expert. I got no idea what these guys and girls are doing in this field. But knowing what I know and applying what I have learned so far in my career, and I can see like connecting the dots, I think I feel empowered, but I also see that This is potentially what's gonna get me out of a job.

Linus: So I better just like strap onto this road, getting all done for dear life. And try to come out on the other end. Still in the game. I, I this even 20 years of experience in, in tech and I still feel every day, so I knew Dawn, like excited to build and like I'm more hungry now than I was when I was 18, which is, It's rare.

Linus: It's very rare. I [00:07:00] think maybe not in tech, but like in general, people are quite tired at working when they're hitting their forties, so

Haroon: Yeah. No, I completely agree with you and from my perspective, you're doing such a wonderful job of this at, of just embracing the technology and just moving forward. And it's one of those things where it's like everyone's gonna have to adapt and so it's better to just dive right in.

Haroon: And that's exactly what you've been doing. And so I wanna talk about that a little bit. What was the whole process of you, observing what was happening in the AI space and saying, Hey, like this is actually something I want to dive head first into. Was it more of a gradual process or was it like, oh shit, this is happening.

Haroon: Let me just go all in before it takes my.

Linus: I think it's a bit of both, , the gradual shift that HA has happened over the past, let's say a year or so. For me, like a bit more a bit more than a year. I've had friends that started building really cool stuff using some of the underlying APIs and technologies that are now commonplace.

Linus: And at first I didn't think much of it. I'm [00:08:00] like, oh, that's cool. That's a very rare use case and that's cool. Probably a lot of people will find that helpful. and then like somewhere when G P T three was announced and we saw some of the early things being built with that, I was like, alright, wait a minute now.

Linus: Now we're entering some really interesting territory here cuz I could see the end usage of this being so power, like so empowering for a lot of people where it's just a simple natural language interface. And then, people started building more things with this and we saw stable diffusion pop up and we saw Dolly and Dolly two pop up.

Linus: A lot of the things just happen ally, like next to each other. And if it's one thing that I've witnessed over and over the last two decades, it's like when things compound and when these Like paradigm shifts happen. They usually don't come alone. It's not actually one thing, but you see like it's building on top of each other and then all of a sudden it's just like the snowball is rolling so [00:09:00] fast that you cannot pick you, you basically kind jump on without just getting hammered onto the snowball and you're just like rolling along with it.

Linus: And I saw that early with web, that's called web two, but social as well. I was, I think I was one of the first. that had Facebook in Sweden it's really weird. Like I literally managed to get like a edu email to get Facebook. And then that snowball happened and now we have some of the biggest companies in the world are these social companies.

Linus: Seeing the same thing now with AI early on, and we're still early on, like we are a bunch of nerds in this field. Even though if Shaq G P T went viral and took over the world and managed to get a hundred million people using it, it's still fringe, like for most people in tech, they've never touched ai, they never touched any of these APIs or, yeah, so just, yeah, that's, it's been gradual, but also there was a pivotal point.

Linus: and I think now we're past the pivotal point. Now it's just okay, it's full on. It's gonna happen. And we're seeing it and being embraced by all the big companies, and it's [00:10:00] gonna get omnipresent, like having an AI assistant of some sort or having trans, like a transformer of some sort available at your fingertips at any given point in time.

Linus: It is just the way it's gonna be. Like there's no going back.

Haroon: Yeah, no I totally agree with that. I think we definitely have hit that tipping point, like the, all the momentum's behind it right now, and to a point where people are wondering, is this a bubble? Is this a fad? Which I have my own answers for that. I personally don't think so. And I would bet that you also align with that as well, but it's definitely got eyeballs from everyone at this point.

Haroon: I, I wanna rewind really quickly. So you said you were one of the first people in Sweden to have access to Facebook. That's really interesting. How so how did you manage to get that edu email address? And also I'm curious are you just generally an early adopter when it comes to technology?

Early adopter

Linus: I'd say like I'm a hyper early adopter. If you go through the, if you comb find, comb the internet for like my writing or articles, you'd see, like I was bullish on VR way too early and like I own that. I, that's fine by me. [00:11:00] I'm still bullish on both VR and ar. I think from a paradigm shift in the way that we're doing computing right now, we're conducting this interview.

Linus: I'm looking at a small little screen and it's the biggest screen I can get from my laptop, but it's still small. And like we will enter an error at some point where we will do embodied computing, but I think me and a lot of other people were just too hyped up on it. We all drank from the same Kool-Aid, and I'm not likening the, what's going on with ai, with vr, it's very different.

MIT story

Linus: Very different. But yeah, in general, I'm an early adopter. And to the point why I got early access to Facebook, I had a friend that went to m i t and I managed to, somehow acquire. Some old students email loaned it and then managed to set up an account cuz it was still very hard to get a Facebook account if you didn't have a college campus email address.

Linus: And then it took

Haroon: An MIT address

Linus: Yeah. Never went mit. Yeah, but It is interesting [00:12:00] though cuz I, I feel my proximity to being early in tech in general has just led me to have this magnet for things. And I also sense it in this community. I see a lot of familiar faces that's been around the block a few times.

Linus: Yeah, it's gonna be interesting to see if we are, if we're dead wrong or if we're right this time.

Price to being an early adopter

Haroon: Yeah, that's an interesting thing. Whereas, being an early adopter, I guess the price you have to pay is that you're gonna have a lot of misses potentially, where you might invest a lot of time into learning a technology, a tool that just ends up going nowhere. The tool is deprecated or the technology goes nowhere.

Haroon: And so I'm curious did you run into that a lot in the past? Not deter you. How do you keep that mindset even despite potentially it leading you maybe to a dead end at CER at a certain point.

Linus: That's a really good question. I think Let's put it this way. I have a very playful mindset when it comes to new tech. Again, general, generally very playful mindset. I have very little ego in [00:13:00] my work and usually just like I want to feel like when I was a kid and I went to the playground with my friends and we all just played.

Linus: I wish every day was like that. I wish . We go into an office working with people, doing creative things. I don't want it to be unstructured and there's no goal or vision, but I want the work to feel like play. Let's just rewind it a bit.

Linus: So at Typeform, Circa 2016 or 17 when the Vibe headset came out, I managed at the time to get signage from the CTO to hook me up with all I needed, like a custom-built pc, like all the hardware to set up like a VR experiment in the basement. And I did, and I managed to personally walk through all 250 employees at the.

Linus: in vr, they all booked sections with me. I all did all the demos just to get people like excited about the future of vr. That led me nowhere to [00:14:00] stand. I still haven't built anything in vr. However, I think I changed those 200 and p 50 people's opinions about vr. They got a taste of what's coming and they all get very excited about getting this experience.

Linus: And it's funny how like none of them would've thought to have done the same if I wasn't there. So it's in a way I just want to figure out ways to have fun. And VR is a lot more fun than AI in that sense, but there is so much we can do with AI that it feels, it's like an abstraction layer that lives in the 2D pain where we're already living and experiencing internet in general.

Linus: and it's a lot more obtainable for people cuz like VR is this big, bulky thing. And like even now with Quest headset, it's like very isolated. You do something and you're alone. But here, like doing stuff, putting it online, other people can try it out themselves. And this flywheel of just everyone riffing off everyone is, I haven't seen that before.

Linus: In, in, [00:15:00] not at this scale, this fast at least. Yeah.

Finding AI tools

Haroon: Okay. So one of the big questions I have for you is like, how are you dealing with all of this?

Haroon: Influx of AI tools right now, what is your process for vetting and for testing them out? It's like drinking out a fire hose, right? So what is the process that you go through to identify the tools that you're gonna test out and potentially use for applications you're building?

Linus: That's a very good question. Like I, I've start, like for the stuff that we're building, the pipeline is quite easy to handle because we have some specific needs or problems that we're trying to solve. And we're very looking at very specific tools to solve those problems. I can give you a clean example is , we're looking to figure out something that can do text to speech really good and with a set of parameters that we like, have decided are important for us.

Linus: So then it becomes easy to kind of source and matrix decide, put everything in a matrix and start going one by one and putting pros and cons. What's more difficult is, I'm found myself in this peculiar situation where I amassed a bunch of followers on Twitter and I'm doing I'm sharing my [00:16:00] journey as I'm learning and discovering things there.

Linus: And then it becomes a lot more tricky to understand this is something that's useful for others. This is useful for me. What are the implic. is this a tool that could blow up and, have 20 million users in the next 12 months? It becomes a bit more tricky to figure that one out.

Linus: And my inbox as anyone can imagine, a bunch of people that are doing what I'm doing as well. But there are so many companies building really cool stuff and they all want to pitch, what they're doing and why we should try their things out and why that's the best. So it's becoming, increasingly hard.

Linus: So in, in a sense, it feels like there is a bit of a bubble in tooling, but I think that's very common. With all new like shifts that happens that everything bubbles up and now everyone's building on the same stuff. And we're getting like 25 text editors. There won't be 25 text editors in a year from now.

Linus: There will be a few that are really good and the rest have. But [00:17:00] we've seen that repeat repeat over in the course of the internet history before. So that's just the way it's gonna be. But what also is gonna happen is when I have all this technology baked in the OS anyway, so do I even need text editor that has AI built in?

Linus: I don't know, but that, so it's like trying to work out what to test now that might be useful for me and others. and also trying to vet okay, this is solid idea, one, one year down the line and how much time am I gonna invest into this? I think one thing to mention there is like every image generation in general, mid journey, stable fusion, dolly if, or Leonardo, and all these other tools that are popping up as well.

Prompting will go away

Linus: Prompting is one thing that I think definitely will go away. So it's if you're focusing too much as a person now learning. Prompt engineering or using the right keywords to get what you want, you are wasting your time. I think it's more important to work on learning the techniques or figuring [00:18:00] out principles that will give you persistent output or like thinking, how can I communicate with ai even though these things might change.

Linus: But we're already seeing like half of the documentation from Mid Journey that was actually three months ago is now depre. . So it's that the space is moving so fast, there is no point in, becoming like a, an expert in mid journey prompting, for example if you don't have the near term goal of selling prompts or something, but, two models down the road, prompting might be a thing of the past.

Haroon: Interesting. So where are we heading? If we're moving away from prompting, like what

Linus: I think more natural language, less, less kind of keyword prompting. Cuz now if you look at it's difficult to like, also obtain like what, where are we going? I think it's a combination of key like removing keyword prompting and going more natural language and what we want describing, but also more image to image or like using source material for your prompting.

Linus: We see that with control net, for example, that just got released the other day. And a [00:19:00] combination of things I think. Equate to how we're producing generative art or generative images or content in general. It's not gonna get locked into a handful of people that are learning how to speak AI per se. It's gonna be become that because then it's gonna be a barrier of entry.

Linus: And I think we wanna try to remove any type of barrier.

Haroon: that, that makes a ton of sense. And in terms of like how you're searching for tools now versus how you search for tools before this massive generative ai. Boost or, just wave. What's the big difference you would say?

Linus: I think there's a lot of aggregators already popping up. So like before there wasn't a lot of choice in terms of finding SaaS tools. There's just a few handful comp like companies that are curating and posting SaaS businesses in general. Now there's a bunch of aggregators that are.

Linus: The tools gets listed. You can filter, you can find stuff, and the fire hose is just pointed at you in a different way. And everyone is look like there's a lot of eyeballs, right? So everyone is trying to get in front as many people as possible. [00:20:00] So it's almost like a self, like the top of the funnel is not the problem.

Linus: Like the fire hose is not the problem. There's just a handful of these, like there's probably around 10 different websites that are listing AI tools at the moment.

Linus: And I think there's a. maybe combined two, 2000 ish tools at the moment that are worthy of actually being looked at, which is quite a lot. And if you start looking into different segments, there might be like 20 to 30 tools in each vertical. But again, I'm not too fussed about all the individual tools.

Linus: If you go to, if you go to hugging face or if you go to replicate. , you can devour yourself in models, in AI models that are doing all sorts of weird things. Ultimately, I think, we'll, we're gonna have to consolidate a lot of this. We can't have maybe we can, I'm rambling a bit, but maybe we can maybe there can be tens of thousands of different things, but I'm sensing that's not where we're heading.

Moving towards genarative AI models

Linus: I think we're heading to. [00:21:00] A place where we have more generalized models that can solve more complex problems that doesn't require you to switch between, let's say a plethora of models, but rather just reduce a handful and then those five models will get the work done for you. One way or another.

Haroon: totally agree with you there. And in the meantime, actually we're putting together our own. Tool aggregator. And I totally agree with everything you described there.

Haroon: It's it's really interesting to see. Just the proliferation of tools right now. It's just, again, it's like drinking out of a fire hose and there's a million tools for the same job. And so it's like, how do you surface the one that is the most tried and tested and offers the, slight advantage over the other tools.

Linus: It is funny, this whole thing with the aggregator is funny cuz I did work for a year or so with a good friend of mine that started an aggregator for direct consumer brands. So thing testing.com and I can see a lot of similarities. Like now [00:22:00] happening in, in the AI space and like the amount of tools that are coming out in the AI space, similar to how there is a lot of direct to consumer brands just popping up like mushroom everywhere.

Linus: So that's an interesting also like connecting the dots going into a completely different markets marketplace, like direct to consumer physical goods and looking at software and, AI tools and how there. , there's 200,000 sock brands, but yet, everyone wears socks.

Linus: But there's currently probably around 2000 different AI tools that are doing wrapping chat, G pt oh G P T three.

Product Hunt for AI creators

Linus: One benefit of having been an early adopter for this long and have got gotten burned and not got burned. I've had the opportunity to be in many different markets or many different areas where, you know, like I've not just been in, in B2B SaaS and I'm not just niched in on that, but I've been across the board trying different things.

Linus: And I think that's also Part of [00:23:00] why I got drawn into this earlier, I think I just like, like I said, I'm a magnet to this I'm drawn to anything that's not new and shiny and fancy, but that's oh, there, there is something here. So yeah, it's like I, I hope that there is one massive aggregator that wins like one Craigslist for AI tools, just like we have product on for.

Linus: I don't want Product Hunt to actually go and, list a bunch of AI to, because literally in the past few months, like Product Hunt has become like an AI launch platform, like every day. This is like AI tools from a bunch of in makers us included , but I wish there was like another player. We need a bit of, they need a competition as well.

Haroon: Yeah. Yeah. I would love. With it. I would love to chat with you separately about that at some

Linus: Yeah

Haroon: ai, I'll share a link with you and I would love your feedback.

Haroon: Awesome. Linus moving on to the next segment. Bedtime stories like that was. First time that I heard about the work that you were doing, and I checked out the product. And [00:24:00] funny enough, only a few days before I was thinking about this, I was like, all right, so one of the really interesting use cases here is like children's books.

Haroon: At the time my wife was pregnant and so obviously, children's related stuff was top of mind. So can you share a little bit more about bedtime stories and the motivation for it? The story of putting it together and the direction you're gonna take it in?

Bedtime Stories

Linus: Sure. Yeah. So to, to just, yeah. Start with a bedtime story. Born by. So it's a small little, it was a small, little idea or side project that me and my colleague Brian, and my partner Jenny, started working on just to get our feet wet in terms of doing something practical and combining multiple AI models into something that we actually also wanted to use ourselves.

Linus: Essentially, our problem definition was that I'm a I'm a parent. My partner is a parent, Brian is a parent. We all have kids. We'll read a lot to our kids. We have a lot of books. I think it's like the, if we look at our budget, it's like [00:25:00] book purchases is probably one of the items that, takes a lot of the volume or the amount each month.

Linus: And I, I don't mind that, I want my kids to have books and I want to read from my kids. However, we are also finding ourselves like reading a lot of fiction, like making up a lot of fiction stories. Or we might try to incorporate something that happened during the day to help my two and a half year old, like grapple with something that she had to go through and we're making it up as we go.

Linus: And then I tell the story and then when it's my, my wife's time to tell the story, she doesn't know exactly what I. So then my daughter or our daughter would go, that's not what dad said. You, I said it differently. And I'm like, wow, this is a painful problem. And what we're just, we don't know what we're doing in general with parenting, so we're just like, we mix it up and we're try, we have the same playful mentality there.

Linus: It's just as long as we do more good than bad, I hope we end up on good on the, out on the other side. We end up well. So with that problem taken of like having. [00:26:00] A lot of books, buying a lot of books, they become quite repetitive. They're not personalized. They don't update. Once a book is there in the bookshelf, it's there and at some point it gets boring.

Linus: What if we can just leverage AI to. Help us. Not take over was never the idea that I wanted what to make myself out of the equation, but I want to augment my abilities to tell stories to my daughter. And I wanna make sure that me and my partner tell the same story if we start, if we start telling them stories.

Linus: And then we started and it became very obvious that this is a really good use case. A children's book or a children's story is quite short. It has a very strict narrative that it follows. It's usually the same pattern and so obviously something that a and a large language model love to work with cuz it's, can produce this output persistently and with good quality.

Linus: We went ahead and we started to do some really early tests and just oh, this is okay. This could work. And at the time I was [00:27:00] also playing around with Dolly too, and out painting, and I published some photos on Twitter way back when I didn't have any followers on Twitter. And one of the open AI founders, added like he sent me a DM and it's this is a really cool use case.

Linus: There's so much stuff we can do with this. And I was like, okay, if he thinks it's cool, then maybe, maybe I should try, let's try to incorporate more things here. And then that's how we got started. And then it became very apparent that, this is something that more people are gonna try to do.

Linus: And there is a bunch of, there's a bunch of good solutions out there. I think none of the solution has taken the approach that we are taking, but. . Yeah. And then I think a more. . So a guy for, on Twitter he had done like a mid journey book where he, made everything like by hand and put some effort and time into making a book.

Linus: And I just, dropped a link there saying Hey, this is what we're building with bedtime story. Didn't expect anyone to go to the website. And then all [00:28:00] of a sudden it was just like went viral and we had don't know, a few thousand signups in one day. And people were obviously on the fence, both.

Linus: We hate this, this is the wrong thing to do. You're like, this is disturbing. What kind of dad are you? What's the future you want for your kids? Like, all the sorts of weird things on that end. And then on the completely other side, you have the people that are, oh, this artwork, this is stolen from someone.

Linus: And it's, and then in the middle we have some enthusiasts that are just like, oh, this is super cool. So I think like it's started a healthy debate around. , what should be done and like, how can this be? And yeah I understand all of that, but I just wanna build stuff, right?

Linus: So I just kinda I slowly back out from the whole fist fight and just get, okay, this has some merit to it, people like this, so let's just continue to build. We're still in that mode where we're just like exploring and building and trying to build more of a platform rather than just like a single.

Linus: Like a gener, like not just a single generator for a single story, but we're trying to build a platform around like publicly available stories that are indexed, searchable, filterable, [00:29:00] and just like medium is for like article writing, but for kids' stories. And the reason why we went with bedtime story.

Linus: Was just because it's a search term on Google that has quite high volume and it's fit the narrative of us focusing on something very easy to market and easy to build, and we don't have to go crazy and try to solve for everything and just like focus and hone in on one thing. But it was always apparent from like the first story that we generated.

Linus: This could be used to do a lot of really interesting things from education to solving, if you have. If you're going through rough times and you might need help with coping with that. Like it could be something, a tool that you could use together with a youth therapist or it's a lot in that space, like in health, but then all the way to fiction and bedtime stories.

Linus: Just, it's just [00:30:00] imagine what kids can do if. They are augmenting their ability to output, their thoughts. So it's like this whole chain of oh, we can do a lot of things with something that came up really early. So I think ultimately, once we are like. We've figured out what our what we are good at and what, like the stuff that we're doing behind the scenes to like glue everything together.

Linus: We know what that is and how have that nailed down. I think we will see like bedtime story going a different direction, like becoming, I think our kind of internal wish or dream for it is to become a narrative company where. , you come here to produce narratives, whether or not it's for kids or it doesn't matter.

Linus: Like we, we see ourselves as a narrative company. And I think that's a very like, lofty ambition and it's gonna be very hard to pull off. But it's also interesting cuz we then have a moat. become a tool and not just someone that's wrapping another service [00:31:00] like in, in a little neat shell, but we actually want to.

Linus: Some of our own IP and we actually have a vision that's longer and bigger than just bedtime stores. However, bedtime stores is a very cool space to be in,

Haroon: So interesting. You touched on so many really interesting topics there. One of 'em just how polarizing some of this stuff can be. AI is obviously having such an outsize impact on the world, and it's scared a lot of people, right? Rightfully

Linus: rightfully I'd say yeah.

Haroon: Yeah. And so when tools are created, a lot of folks, the first thing they think of is, all right, so who does this adversely impact?

Haroon: And how can we prevent those adverse impacts? And is this in that positive? Is it not? But I think that's a builder's mindset the one that you described. And I think in the end, technology isn't gonna stop for anyone, right? Is my own

Linus: Sad, sadly. So back to the fact that I think I'm gonna get automated out, and this is my reaction to it, just like

Haroon: I think everyone's gonna have to be reactionary.

Haroon: And, going back to your, like the entire use case that you focus on I think this is such an interesting one because especially in the educational space, [00:32:00] I was talking with one of my friends who is a vc who invests in at tech startups.

Haroon: And one of the things you told me that was really surprising was, Literacy is probably the biggest problem in the education space in America today. And that surprised me because I'm like, it seems like literacy isn't that big of a problem in the us but he said, a lot of parents are finding out that their children are reading, but they're not actually learning how to read, right?

Haroon: Like they're just, it is more of a superficial type of reading that children are doing where they're not truly understanding how to read to, to, to the level that the educators want them to. And. A big potential solution to that is more personalized content, right? Is stuff that they're may be a little bit more engaged to read, rather than reading about Jack and Jill reading about themselves in a story or a situation which interests them and learning that way.

Haroon: And I think, the educational. Angle is such an interesting one. I think there's huge potential there, and I'm so excited to see [00:33:00] what bedtime story and whatever else is gonna branch off of that. Wherever you decide to take that I think there's just massive implications here.

Educational use cases

Linus: Yeah. I think there's a bunch of people that are already working on, like specifically solving. educational store creation or even learning mark is one of the founders of us two games, or like one of the early earlier employees. I don't know if it's a founder or not, but I know that they are working on a tool like this for education that's like targeting kids with a diagnosis.

Linus: It could be ADHD or it could be other like learning disabilities. And it just, exactly what you're saying here is. . Having the ability to personalize the content on the fly, having it delivered personalized on the fly, having the questionnaire or the like what is that you are learning to be also personalized.

Linus: It. It's gonna be a paradigm shift in how we learn, and I think this is one of the [00:34:00] fundamental. Changes that large language models will bring to the next generation. Just the breadth of information and the richness in how people are gonna be able to inquire information is just gonna make Google look like an old landline telephone.

Google is going to be outdated

Linus: It's just, it's gonna be crazy when we think back at this and go you actually had to Google and it was a skill to Google things. Did you live in the ice age? Did you guys have moments it's just gonna be very strange. And even to the point where like my two-year-old, her definition of Google is an omnipresent voice in our house.

Linus: She doesn't know that Google is a search engine for her for what all she cares about. You ask Google to play songs on Spotify. Or you ask Google about the weather and it's there, it's in her room, it's in the living, it's everywhere. So my definition of Google and the age old, like you [00:35:00] have to Google stuff to learn and her definition of Google and what she will do to learn, that's just gonna be very different.

Linus: So yeah, I'm super excited about this and I think I, we just hope more people can put on. Excited mindset. And of course there's a lot of problems that we need to solve, like with any new tech and we need to be vigilant and not fall in the same traps as we did with social. Cuz social is gonna have impacts far longer than what we anticipated and it's gonna be something that's we gonna have, have to deal with as a society.

Linus: But I think like you had, don't get too cautious. Don't put up roadblocks just because you want to put up roadblocks. Yeah. It's, super excited about this. I could probably talk about implications for the next generations for hours, literally for hours.

Haroon: Yeah. And I think that's a great segue too the closing here. If you had a megaphone to the next generation, what would you tell them specifically in [00:36:00] regards to just preparing and adapting to how the world is changing as a result of.

Next generation needs to learn how to ask really good questions

Linus: Yeah. That's a very good question. I think. What the next generation need to become very good at is asking really good questions like the, and then that's true currently as well, but don't think we had the same issue. Like the, I think next generation needs to become expert, inquisitive. They need to be really good at just asking very good questions.

Linus: And if they can ask good questions, I think AI will be, A, a very powerful liaison in, in that relationship in the future, cuz AI is not going away like the gene is out of the bottle. We can try our best to regulate, we can try our best to sort this out. But, more slow is continuing compute power is, on, it's still exponential.

Linus: People don't [00:37:00] get this like wrapping your head around all of that. Whatever, get good at asking questions that's not gonna get old. Even if we have AI that's gonna be good at asking questions, you as a person or as a human our ability to imagine and rethink and reform and also adapt, I think, yeah.

Linus: If we're, if there's one thing the next generation needs to be good at is asking questions.

Haroon: Love it. Great advice. And last question for you. If there was three tools that you could, three AI tools of course, that you could recommend to folks, if they're just starting to dabble into the world of AI creation, which three tools would you recommend they start with?

Linus: If I if I would give advice on three tools to get started with ai, I'd say go to chat g p t experience what, the best of the best is in terms of a large language model and a user interface. And then mid journey, even though it's a difficult to maybe get across the discord [00:38:00] situation, but then you get a good idea of how diffusion model works and then text to speech add.

Linus: 11 labs to get a sense for how powerful the current state of the art, voice synthesis is. So let's leave it at that.

Haroon: Perfect. Final question. Where can people find you? How can they follow all the great work that you're doing and learn from all of the wonderful prompting and the projects that you're building?

Linus: So people can find me. I'm more, most active on Twitter Linus EAM on Twitter and then. I write a CK as well, so it's linus eam dot.com and yeah, I got a website but has, if I haven't updated my website in since 2017 and I keep getting like tens of dms a week now.

Linus: It's like you, dude, you need to update your website. I'm like, whenever I have the time, I'll do it. I'll update

Haroon: Whenever you're done building

Linus: whenever I'm done building or changing diapers.

Haroon: Yeah, exactly. Yeah, I know how you feel now. Linus, I'll add those links in the show notes, but thank you so much for your time and [00:39:00] thank you for being the inaugural guest for the Not a Bot creator series. Super excited to

Linus: Perfect. I appreciate it. Good chatting.