She Left UWaterloo Without Graduating. Now She’s at OpenAI

I definitely don't want to stop learning. That scares me quite a bit, especially since I left school. Didn't graduate from my undergrad. So, I'm here with Brooke Joseph.

She taught herself differential privacy as a teenager and joined a 12 person startup called Sky, which 6 months later got acquired by OpenAI. Take me back to that moment. " What makes Sky feel magical? What makes like this software magical?

We had this one feature that I think Conrad built. It was called the Skyshot and now it's in Codex. It's called AppShot. Everyone should use it.

Like you said, you picked Sky for the people. Ari and Conrad. Tell me more about them. The team was just like super talented.

When we were at Sky, some of the team figured out how to do background computer use. So you could have multiple computer sessions at once on your computer. We're the first team to figure out how to do that. I just threw myself into things without being scared and trusting my ability to figure it out along the way.

Hi, Brooke. So, I'm here with Brooke Joseph. She taught herself differential privacy as a teenager, built ML for Volva Seammens in Sweden before most people finished undergrad and joined a 12 person startup called Sky, which 6 months later got acquired by OpenAI. Now she works on computer use post training at OpenAI.

6, the thing she believes is one of the most important skills on the road to AGI. Welcome to the show. Hey, thank you. By the way, where are we right now?

Uh we are in Washington Square Park. I think it's Washington Square Park in San Francisco. It's like my favorite area of the city. Is there a fun fact about this place?

So, I was told I actually don't quote me on this because I haven't searched it out since I was told this, but Marilyn Monroe was married in this church, I think. It's pretty cool. That's so cool. Yeah, there's so many gorgeous places in SF.

I'm so glad we're doing it here. Now, let's first talk about a little bit about Can you tell me, Brooke, about your first full-time, like first job? Not full even full-time, like first ever ever job. Yeah.

So, when I was younger, I don't even like actually remember the specific age I was when I started working at my dad's store, but I that was like my first job. I worked with my dad. He owned a produce business uh in the small town in Canada I grew up in. Uh so, usually like on weekends or breaks from school, he would have my brother and I wake up very early.

m. and then go to the store and uh we would work with all the employees and we would get the store ready. Um and it was a lot of really hard work. Uh but yeah, that was like my first job.

What did that like teach you now that you still kind of use to this day as a you know engineer and then like now researcher? Yeah. So, it obviously didn't teach me like much technically, but I think it taught me a lot of really important soft skills that I just like carry forward or just I guess maybe even ways of being. So, I think it taught me a lot about how to be hardworking and persistent.

" But I think now that I'm older, I have so much appreciation for it because it taught me a lot about just like working hard and just, you know, uh sticking through with things. Um so I think it was like really important like foundation that I it taught me and also just all the people my dad hired worked super hard. So that was like the only example I had of like this is how we operate. We work hard and um and yeah and have fun doing it too.

So wow. Well, you have like you had so many amazing cool chapters in life so far. I mean you did federated learning in agriculture, differential privacy in New York, web apps in Toronto, edge ML in Sweden, Sky and OpenAI. So all of this before most finished undergrad.

If I asked you what all those chapters have in common, like what would you say? Um I think it's probably like putting myself in situations that I'm like on paper very much unqualified for. So, for example, the first job I had in Toronto, which was like my first like real kind of like internship, um I was like very unqualified for it. I never built any web app before.

I didn't know React. Like I knew nothing and at the time, uh LLMs were good, but it was definitely no like codecs um for example. So, uh and also like the company didn't really pay for a lot of like these agentic tools. So I was sort of like left on my own to like actually figure out how to like write React for example and how to build web apps.

And I think I just threw myself into that situation and then I was like I cuz I really want to be excellent and do the best work I can possibly do. I was like I need to just figure this out. So I uh basically would spend weekends trying to like watch videos on React and build like like little little um sorry little React web apps. Uh and then yeah and then I would like try my best to just show up and build web apps for the company and that really turned out well.

When I moved to Sweden, I had done like sort of related research. So I think technically I was sort of qualified on paper but I was very much unqualified to move to Sweden. I had no idea where I was moving to in Sweden. Like I couldn't locate the town on the map like until a week before I actually moved.

Um and and um also like I didn't speak Swedish. I didn't know a single person in Sweden. I moved to like the small like super small town. " Um, and yeah, I was like very unqualified to move to Sweden and I did it and I learned a lot through that experience.

When I joined Sky, the team was just like super super talented and um like some of the best engineers to this day that I've ever met and worked with. And uh I felt very unqualified to be there and and I just was like, "Okay, I have to keep up. I have to learn and you know try to do excellent work and meet expectations and I I think for the most part I was able to do that and so yeah I think the commonality between all of them is I just threw myself into things without being scared that and trusting my ability to figure it out along the way. Um and I think it's worked out.

Yeah. Awesome. So what would you tell someone who would um who feels like they're going for a job that they feel unqualified for? Uh, I think I think this comes with experience, but like having confidence that you'll be able to figure it out as like like time goes on.

Um, I think like as you gain more data points of like, okay, like I put myself in this situation and I figured it out. Uh, I think like you become more confident in yourself. So, trying to actually like find those experiences where you're like, damn, I don't know how I'm going to uh like meet like level or rise to these expectations. and then kind of like forcing yourself to do that anyways.

Um, so yeah, I think it comes with time. Like the first internship I had in Toronto, I I definitely was like very scared and I think like as I did it more and more, I become like less scared and more confident that I can figure it out. Um, and I still think I'm learning that. But I would say that's like my biggest my biggest Yeah.

And out of all places, why Sweden? What's in Sweden that was like super important for you to be? Um, well, I really loved the founder of the company I worked at. He was like so incredible and he was like my he was so kind and he was like my point of uh contact and I um I was just really excited to work with him and I always try to like optimize for great people when I work places like smart people who are also just like really kind humans.

Uh and they met that profile all of them. Also, I was sort of debating at the time between an uh an internship in San Francisco or Sweden. And I was like, I'll go to San Francisco later. I can never move to Sweden.

And honestly, it was like a great decision because I got to experience what it was like living in a foreign country. Uh I met so many wonderful people in Sweden, uh who I still now speak to and and so overall was a good experience. Yeah. I'm really curious like what's a how do you call like what's kind for you?

Like what does that mean? Like what makes a person kind? Um yeah, I think I think like people who are just like selfless and like want to see you grow, want to like uh like care about you being curious and your passions and want to like help you um just like improve and become better. Yeah.

Uh, I've been really fortunate in that everyone I worked with in my career has been really kind and like met that I guess profile. Um, but yeah, I would say like that's the biggest thing. Um, and then in Sweden I think it was like people were very cared a lot about me as a person and less about me generally and like my career which I think was also really nice. um they cared like how are you doing?

And also like I there was a situation where I like I signed up for this gym membership uh very far from where I lived and then I needed to like cancel it and then I I tried doing it and then they like only spoke Swedish. So then I was like damn, I'm screwed. And then I told one of my co-workers and then the next day he drove me there and then spoke to them in Swedish for me and helped me cancel the membership. And uh like just like stuff like that I felt was like you really did not have to do that and that was so kind.

Um and then I think the Sky people as well were also just like very kind people always making sure you're okay and and just like care deeply that you know you're like you're on track to your goals and that you yourself are good too. So um yeah, I think it's many things. That's amazing. and a little bit more about kind of your work in Sweden.

Uh you were kind you were uh compressing models onto edge devices for valance semens quantization distillation. Break that down for me for like and for anyone who doesn't know what those terms mean. Yeah. So quantization is like sort of um I guess like decreasing the amount of information that we like train on because we want to we're working with like super small uh microcontrollers.

Um, and then the distillation work I did, um, actually don't know how much I can talk about it, but it was like sort of, um, the general idea was like taking learnings from like multiple models and figuring out ways to like have these models like learn from one another. Um, so yeah, that's probably I can talk about it. Awesome. Perfect.

And then was there anybody like anyone notable on this team of PhDs that you kind of were just like really inspired by? Um yeah, I think like all of them were really inspiring um and very very just like smart and kind. Um, I don't think there was like one particular person that was like that stood out, but a lot of them were really good mentors and taught me a lot about like how to like set up experiments and how to think about things more scientifically and like how to start with a hypothesis and then like um I guess like from an ML perspective set things up so that you're like getting what or trying to like reach the goal of whatever the hypothesis is like setting out. Yeah.

Um, so I think it was like a really good experience to like lay the foundation for like how to be a scientist. Um, yeah. Nice. And going into a little bit more about now Sky.

So before Sky, like you mentioned in one of your Substack articles that you really love like moving fast, like breaking things and Sky would spend an hour in design uh meetings thinking about like the minuscule details. Like when did you stop thinking that was crazy? Yeah, I think it was like probably a few months into Sky because I remember joining and then there was like these design meetings and I was like oh my goodness we're spending like an hour talking about like very very detailed things. Um and then yeah like you said at first I was like why are we doing this?

And then I think I as I started using Sky and then started to see how those like hour conversations played out in the actual product I was like okay this makes a lot of sense to me because you just felt the polish of the product. you felt like how magical it was. It was such a a special like uh piece of software to use. Um so I think that was like uh I started to grow an appreciation as I started to like actually use Sky.

Um and then yeah see the the those choices play out. And what makes Sky feel magical? What makes like just software magical? Um yeah I think it was like a lot of things.

like we had this uh one feature that I think Conrad built um that basically it was called the skyshot uh and now it's in Codex it's called appshot everyone should use it uh and basically the idea is that you would take like um um an app and then pull that context into the chat so you would do like command command and then pull it in and there was this really nice animation that would happen and there was like a sound and it was just like so many things playing together and it was just like every time I saw people use Sky for the first time and then they did that. Uh you would just see their face light up. " Because it was just so uh well put together and like nice and beautiful. And I think that was just one example.

Um, we also when we b um made the first like initial computer use uh or brought computer use into sky. Um, uh, Dexter, who was an intern at the time, built this like really cool, uh, cursor that would like squish when it like clicked on things and then it would fly because like Sky had the it was like this command pallet, I guess, and then it had like this like sky logo uh, that was like always there and you could use it to toggle the window, but then it it also like when you use computer use, the sky logo would fly off and then it would like be your cursor and that was a really magical experience. I loved that. And then also just seeing Dexter specifically work on that was really special because he was so focused on all the details.

" And also you just like saw his work play in the magic of the product as well. So it was like many things but just everything was so intentional and and and nice. So, uh, yeah, that's amazing. And like you said, you picked Sky for the people.

Ari and Conrad. Tell me more about them. Yeah, I I mean, Ari and Conrad are really special people. Um, they're both just amazing humans, so selfless and kind.

Uh, Conrad is like uh what I the nerdiness I aspire to be. He's like very just passionate about computers. like like he just loves computers and he's always building side projects and crazy stuff and I always love hearing what he's building because um it's just like how did you even think about that? And then Ari is also very special human.

Uh he in a good way. Uh he's very curious. He's very kind. Um, and one thing I really love about Ari is that he's always um trying to like understand things and he has like zero shame.

If he doesn't fully understand things, he'll ask why a million times until he understands it. And then also as a founder, he was uh so good at being in the details of things. Um, one day hopefully if I found something I really want to be or like model what Ari was to Sky, which um, he was like in the details. He was putting up PRs.

He was reviewing code. He was like talking to investors. He was like in uh uh meetings about like design and he was like in everything and uh he didn't um I don't know he was just like really good at being present and you could tell that he just cared deeply about everything um and wanted everything to be like not perfect but like the best it could possibly be. Yeah.

Um so yeah, both great people. Yeah. And then um you kind of yeah I mentioned like Ari like says a why a lot and that like also inspires you to keep saying why like no matter what like um what's like the hardest part of that of like saying uh yeah I think it's a bit like sometimes you'll like ask your co-workers why and then I think some people don't want to always like uh continuously answer that. So it's like kind of respecting people's like time as well.

Yeah. Um, so now sometimes I ask why to codeex a lot. Okay. Cuz I'm like I can't annoy you.

Yes. Um, so I think that's like maybe a hard part of it, but I I do really like being curious and and um I think also I've always been like that. Always asked why. Like I remember being in high school and I was really into chemistry and then I Oh wow, not so many.

Um, I remember being really into chemistry and um, the teacher that I had in grade 12, he was like super he had a PhD so he was very curious person and I would go to his office and I would like ask him a million questions and like I would come with a notepad of like pre like pre-planned questions that I would ask him and then I would ask why so many times just because I was curious and I did that in uh, university as well. I was like always at office hours in classes I cared about. There were some classes I did not care about and I didn't do that. Uh, and I would ask like why um, so many times.

Um, and then seeing Ari do that as well as someone I like really look up to kind of like inspired me being like, "Oh, okay. This is okay. We can do this. " Uh, yeah.

That's so lovely. Okay. So, after 6 months, you joined OpenAI acquired the company. Take me back to that moment.

What was the feeling? Where were you? What was going on? Yeah.

Um, I remember Ari brought us all into a room and then he's like, "Okay guys, I have a special announcement and then he had like a nice like PowerPoint design put together and um, and yeah, I think like everyone was like mostly excited. You know, there's like some mixed feelings. I really had only worked on the product for 6 months. Um, I was I really loved Sky and I felt an attachment to it, but some people like the founding team were working on it for two years, so they had like a different attachment and feeling.

And so I think that they had like uh mixed feelings, mostly positive. I would say everyone was like excited. Um and uh yeah, it was really good. I remember we all went for a walk around Salesforce Park after.

Um we're all like talking about how excited we were. Um yeah, just good vibes. Yeah. Yeah.

No, it was Yeah. Unreal. Like it's Yeah. Probably it's very awesome.

Um and what was So at the time it was Yeah. 12 person team. Yeah. Okay.

So you guys were all there together then. All of you. All of you. All of you.

Awesome. It was really nice because I was an intern at the time. So it was really nice to be included in those conversations and we were uh yeah like Dexter who was the other intern there. We were like always like uh very grateful for you know being so closely a part of the team.

Yeah. It's incredible. Okay. Tell me more about So computer use is a is a big thing for you.

Tell me um tell me more about what that is for anyone. Like break it down in simple terms for anyone. Yeah. Yeah.

So computer use is basically letting the model uh perform click type actions on your behalf. Um so it has like anything that you would do on your computer. So the ability to like uh type um to scroll to click um and then also to um do like uh like key uh keyboard kind of like actions as well. Um so yeah that's sort of like what computer use is.

And then specifically Codex computer use for Mac is really exciting because uh when we were at Sky, some of the team figured out how to do background computer use. So you could have multiple computer sessions at once on your computer and also be able to do other things on your computer without Windows uh being brought to the forefront, which is to my knowledge, don't quote me on this, I think it's the first uh we're the first team to figure out how to do that. Um or at least like on Mac, I think which is really exciting. And I think that was like something really revolutionary and and made computer use a lot more of an enjoyable compelling experience.

Um yeah. So that's like what computer use is. Uh and it does so many cool things. Uh yeah.

Why is it important for anyone to know about this like in the space like as a founder maybe as a developer? Yeah. Yeah. I mean like it just enables you to do so much more.

6 is very good at coding um just as an example. Yeah. Um, and then I think like the next exciting thing is okay, how do we have enable our model to be able to use like any piece of software and that becomes or that makes like a whole other range of things accessible for people. So, I've been seeing a lot of demos on Twitter where people are using it in Blender um or like uh uh AutoCAD to like CAD stuff.

And that's like super exciting because now you have people who have never opened a piece of CAD software, designing stuff and then 3D printing stuff like the next day and like having codeex do it all. Um and it really feels like it's we're like zeroing in on closing like the end like pipeline of things which is really exciting. Um, yeah. And then it also like Codex computers use like makes me so much more productive.

Like I'll have it do things in the background for me while I'm like uh doing other things. Like I also heard someone the other day tell me that they had Codex computer use like go on for like an hour or not an hour, multiple hours while they were sleeping. And it was like trying to find them like the best hotels with the cheapest prices and very specific criteria. and then it came back with a whole spreadsheet of like a bunch of options that were really high quality and good for them.

Uh so I think there's like so many it enabled so many exciting use cases. Yeah. When you're coming into um I mean when you're learning more about like uh computer use what kind of things were you not expecting and uh like how did your opinion about like things like change your Yeah. How did they change when you're like more involved with it?

Yeah. Um I mean so when I first started working on computers I was at Sky still and I think initially I was a bit like confused what it was and at the time it wasn't as popular. I think now it's becoming more I see it on Twitter all the time. Okay.

Um but yeah I kind of was like questioning the utility of it. Like I went back and forth with some uh of my friends and I'm like do you guys like what is the point of this? 6 it's very good um for example and how powerful it can be and what it can enable for people I've become more excited um about it. Yeah.

Codeex can use any app a human can. What should founders build differently in a world where that's true? I mean, I think I think like as I mentioned, it's like Blender use cases or like using like more heavy duty software like even editing um like in Final Cut Pro and being able to do more stuff than just programming or that requires some like programmatic um approach to like doing it I think is pretty exciting. Um yeah, or like even like music uh music editing like you can literally do anything now.

like you don't really need to familiarize yourself with like is it like FL Studio I think they do for music or something like that's I think that's just like that's pretty exciting. Yeah. And so what like what's the distinguish between computer use as a scale we distilled into models and AGI itself and are we like on the road to that? What are your thoughts on that?

Um, yeah. I mean, I think like having the agent be able to use any piece of software is like really important on the road to EDI for sure. " And I think like we're getting to a world where we have just like one chat interface and then we can just like spin off like a bunch of agents to do things on our behalf. And a lot of things won't be possible without computer use.

For example, um the one thing that I'm thinking of uh the person mentioned that they were like had to edit this video or something. Um and then uh they had like computer use like go and like look at the video to make sure that it was like uh uh like edited correctly. And then if it was like okay please do this to edit it or like fix his bug or whatever and then he open his computer and it's just like computer is doing it. So I think wow um yeah I think it's it's like really important that we have models be able to like use your computer so we can close the endtoend pipeline and then all you need to do is like distill whatever you're thinking into like some prompt and then have like your agents do it.

Yeah. When when should you not do that? like when what tasks is it like you should probably always keep an eye out on or like um some things. I mean I think it depends on how comfortable you are.

I mean I personally believe that open takes security and safety very very very seriously. Yeah. It depends on like what you're Yeah. It depends on what you're doing and what you're comfortable with like what your knowledge is too.

I hear some people like give it their credit card information and like let it do trust. That's like a little bit beyond my scope of trust. Not because like I I like not because I don't trust the models, just because I feel weird about it still. I don't know.

Like I I even feel weird giving websites my credit card information, buying stuff sometimes. So um Yeah. Yeah. I think it like really depends on your level of comfortability.

Yeah. You're doing a really incredible things at OpenAI right now. Would you say it's true that like everyone should eventually like build their own company? Like is that where you're headed?

Yeah, that really excites me. Yeah, just being able to like own a vision for something and then like put together a group of people who are passionate about the same vision and then work and execute towards that. Yeah, that excites me quite a bit. Um I think eventually I would like to start my own company.

Uh especially since I kind of grew up with a bunch of entrepreneurs. You know, my dad is a a founder and then my grandpa and then my grandpa's my grandpa's dad. Uh, so everybody since they immigrated to Canada have all been founders. So I feel like that's in my blood.

I've always wanted to do it because I'm like I feel like I have to. Uh, not not necessarily feel like I have to, but like more more like it just feels on on brand. Um, and um, yeah, I mean I'm really enjoying my time at open air right now. I think it's a really good opportunity to like learn a lot and, you know, work with really smart people and um, yeah.

Yeah. And what do you still need to figure out before you can make make it happen? Make your own company like who founder? What do you think?

I think finding a problem that I have like a super strong affinity towards like I have general ideas uh or spaces that excite me, but I feel like the actual problem needs to be set in stone and something worth attacking before um actually pursuing that. Yeah. Uh so yeah, I think about that sometimes, but for the most part, I'm enjoying where I'm at right now. And uh also I think the world is moving very quickly with AI.

So it's really hard to like decipher like what are exciting opportunities for the moment versus like what will be exciting in 10 years from now as well. So sort of kind of Yeah. I think it's just like a really strange time in tech and also maybe even the world. Um so yeah.

Yeah. You think this like this time do you think like just this time in particular do you think it'll be like much more clear like down the line in a few years everything will kind of um honestly I'm like very unsure. Yeah. I guess we I remember being in 2024 and feeling sort of similar feelings and I'm like oh I think it will calm down probably and then here we are in 2026 and I'm like I don't think it's calming down.

So it's a bit I'm a bit unsure honestly. Um and I think people have very uh like varying opinions. uh a lot of really I've asked similar questions to a lot of very smart people I trust and there's no like similarities between the answers and so I think everyone is very unsure and a lot of people's predictions are kind of like scoped to like uh I guess like the data points they have and like extrapolating from like where we were like a few years ago and trying to like continue that trajectory. So everyone's kind of guessing in my opinion and no one really knows and I think time will tell on like where we are.

So yeah. Yeah. Yeah. And also yeah like you're this idea of like you want to you know find a problem where you're like kind of world class at it too and like something that you could really um know like yeah you know so much about and yeah this like a this state of like relevance like what is going to be relevant like now that's not that's going to be relevant in the future that's like yeah cuz you start something and then it might like change.

Yeah. Yeah. You mentioned to me before that uh you know in terms of career you never feel like satisfied which is a really good thing because I feel like you're always like you're always like growing. You're never going to you know you're you're never going to stop like you're going to keep striving.

Uh is this this like the striving of something and like is this like the goal? Are you more like chasing like some kind of like arrival anywhere? What's like your um I don't know satisfied like where's that come from? Right.

Yeah. I mean, I don't I don't think I'm chasing a specific destination. " And like what does excellence look like? And then uh just trying to like live up to that and then reaching that point and then kind of being like, "Okay, what's next?

" Um, so I think right now and probably for the next like 10 plus years, I probably will continue to chase that uh those experiences where I can continue to level up and grow and be better. Um, I think like ultimately I would would love to like maybe when I'm my parents age chill out and kind of enjoy life and just do pure passions and hobbies. But I think at the moment I really love working on difficult problems and I guess like putting myself in like uncomfortable maybe stressful situations and then trying my best to like live up to uh the expectations and the quality bar that I have for myself. Yeah.

Um yeah. What are some of those difficult problems for you right now? And yeah, what are you chasing the most right now? Um, I think like high level I'm chasing just being the best researcher I can possibly be for the position I'm in.

Um, you know, just like discovering ways to make our models better and better at computing specifically, faster, more accurate, more reliable. Um, those are kind of the problems that really excite me right now. And what's something you're still trying to like what's a ch a really big challenge for you? like what are you trying to like like trying to figure out and grasp like even outside like anything it doesn't have to be even like young anything where it's like you're really stepping outside of your comfort zone with this and it's like yeah I don't know I I think like one thing I I think is a bit of a challenge is like figuring out how to balance like relationships generally.

Um, especially now I feel like I work 90% of the time I'm awake, maybe more. Uh, and I think having relationships with people, like friendships, um, like relationships, uh, uh, like with boyfriends and whatnot, um, parents, grandparents. I think that's a bit I struggle with that sometimes like understanding like okay when is when is it appropriate to put the computer down and not try to like reach for the A+ and like just like enjoy moments with family and uh friends and that's pretty tough and it was part of the reason why I was really excited to move to Sweden because it felt like a good opportunity to prioritize that and I think I learned a lot of lessons through that but I still struggle with it um and honestly I think it's okay. Like I think it's okay, but it's also in my experience like important to figure out at least how to balance the family component because I think a lot about how my grandparents are getting older and stuff and that makes me a little upset and they mean a lot to me.

So, um Yeah. No. Yeah. No, I Yeah, cuz I think I think especially at our age like Yeah.

You know, a lot of people say like you really focus on like your your passions and and and like it's so important like you have someone who's like you can strive and reach towards goals with and like I'm sure like if you find the right person like they could like motivate you and like you both are like super busy and u yeah. What are like some of those like difficult um challenges or like difficult things that lessons you learned um in Sweden then? In Sweden. Yeah.

Um, oo, I feel like I learned a lot in Sweden. When I before I moved to Sweden, I kind of had this realization cuz I was in school and I was always like, "Okay, I need to build cool projects. " Um, and I felt like I was and also try to find internships because I went to Waterl. We both went to Wateroo.

Um uh and then yeah, going in to Sweden, I realized that I wasn't really like prioritizing friendships and hanging out with people. So, I was like, I'm going to go to Sweden. I'm going to make friends in this foreign country where I know nobody. Uh and I think that forced me to like get out of my comfort zone cuz I I remember I went to this small town and I was telling my co-workers, I'm like, I need to find people my age to be friends with.

And they're like, that does not exist here. Uh, and then they took me to one event and I met this girl and then she introduced me to all her friends and then I made all these friends in Sweden and some of them barely spoke English so when I hung out with them it was like kind of awkward because they would speak Swedish and then I would just be sitting there. Uh, uh, yeah. So I think I I learned that.

Um, what else? I did also a lot of solo traveling. So I had a lot of I think maturing through those experiences as well. I think before I started solo traveling, I was a lot more trusting of the worlds and people generally.

And then solo traveling taught me, especially because I was at baseline just like more on guard cuz I was like a girl traveling alone. Uh so it kind of forced me uh to just like be less trusting with people and I think that was the right approach and I think I've carried that like with me um after Sweden and and solo traveling as well. Um yeah. Oh, that's awesome.

Let's talk more about your love for math. You love math so much, right? Let's geek out about it. Like tell me like why and like what makes math so beautiful to you.

Yeah. Um I love this question. Um so when I was growing up, I had this teacher named Mr. White.

I loved Mr. White so much. Uh he was the best. And he was like this old, some people calls him grumpy.

I loved him. uh professor who was like very uh like had a deep passion for math and more importantly he really loved uh teaching kids math and he was just like the first person that I felt believed in me cuz I went to this math class and everyone was like really smart and you know their parents were kind of like forcing them to do math uh and they would do like math worksheets um like when they were super young and then I felt like I was coming in and I was like my parents don't care about this. They were very chill. Uh and so I felt kind of like unqualified to be there and he sort of reassured me that you know you're you you deserve to be here and you know you're good at math and he had appreciation for my passion and I think that just like carried on to the rest of my life.

I after finishing his uh classes I was really excited about doing math in school. So I went to Waterl and did computer science with like lots of math classes. Um, and then yeah, I also host these math nights uh in San Francisco now because I wanted to bring back what experience I had with Mr. White um back to San Francisco.

Um, yeah. So, I think for me it was always really fun to solve math competition problems and just like be around people who were very passionate and had a love for something and I think like math was that thing for me. Um, yeah. So, you did math competition?

I did math competition. How do they look like? Um, what do you mean? How they look like?

So, like what kind of math did you solve or like Yeah. What did you Yeah. So, I like high school math competitions are a lot more like um number theory and like um uh like like kind of like puzzles a bit and you have to kind of know the trick. I think math competitions generally are like this.

Yeah. Uh so yeah, those were always like super fun. In Mr. " Wow.

Oh my god. And uh or like the ones that seemed scary and intimidating and I knew no one else was going to do. And also it was a it was a good strategy because like I would always like get like maybe 80% of the way there and then I would get stuck and then I'd had every reason to like reason about this with Mr. white and so we would like try to solve it together.

Uh or mainly him yelling at me like no that's wrong. Uh he had a very funny like voice. He he was awesome. Um yeah I I don't know.

Um so those were like kind of the style of questions and then Mr. White also made us like write these uh contests that were like kind of out of out of distribution. Like we did this like one Russian contest called Tournament of the Towns. Uh I think that's the name of it.

I actually kind of forget. And that one was super difficult. Like I actually don't even think one year I didn't solve any questions and then another year I only solved one. Um and even solving one was good.

Uh so yeah, it it kind of varied, but those were like really difficult. I remember those being tough competition questions. Wait, can you like visual let's visualize that like what why are they difficult? Like what was like hard about it?

Um, what was hard about it? I don't know. It was just like there were like tough uh contest questions that you know probably didn't get like by just like intuitively reasoning or thinking about it. Um, or like you didn't know the trick.

I think that's like a lot of competition questions is like do you know the trick or like have you seen this style of question before and then know the optimal strategy for like solving it? Um cuz if not then like I would say most of the time you're screwed. Or if you have a question uh sometimes it doesn't always look the same but like you can reason about it using the same kind of strategies. Um so yeah I don't know if that answers your question.

Yeah that's amazing. And so like for me like math is such a it's such a scary topic and I think for many too maybe they're like it's a very challenging and so like I feel like my brain just like it just I can't like grasp anything in math. And so what do you like what do you tell people who are like scared of it or like are like they're having a hard time like like figuring it out like Yeah. I actually I don't really know.

I mean, I feel kind of bad saying this, but I felt like math always kind of was a bit intuitive to me. Okay. Like I just kind of Wow. Like I I don't know.

I I never really struggled with math. Yeah. Uh so yeah. So you don't really like know what it Yeah.

What it feels like to I also feel like the the one thing I really love about math is that you can always like go down like super fun like fundamentally like everything everything is intuitive. Everything makes sense. Uh Okay. Wow.

That's amazing. And this is awesome because I feel like there's just some skills that are like so much more intuitive to us to like some people and some to others and some how to just work like extra extra hard. Yeah. I don't really know where it came from.

Uh so that's great. Uh I don't really know where it came from. I mean my dad is a very like analytical person. He's very very good with numbers.

So maybe I got it from him. I don't know. Are you still in touch with Mr. White?

Mr. White passed away. Yeah, unfortunately I was very sad about that. " Uh, so I I there's um in MB3 there's a atrium named after him.

It's called Bruce White atrium. Heard of it? Yeah. And so I've taken a picture in front of the sign.

I'm like, "Oh, Mr. " And it made me cry a bit. I looked at our text messages after he passed away and I cried. He was such a special person.

This is someone who like made such a big impact on you. Wow. " Like what made Waterl Why was Water so special? It's just like because it's a very like big math school.

He also did his PhD. I think he did his PhD there and his undergrad, I think. Yeah. Um but it's the best school for math in Canada.

Wow. Um and then also I remember touring the campus and then in MB, was it MB2? I should forget the building names actually. Oh my gosh.

Uh but uh if you go to the staircase, there's like a bunch of like math and symbols. Do you know what I'm talking about? Um B1. I actually forget the name.

It's like the pink there's like the one where it's like Yeah. pink and then you walk and then it's like level three will have pie and then explains like uh what? And I was like I love this place. Oh my gosh.

Yeah. I remember that building too. I completely I completely like all the building names just like Yeah. completely.

Wow, that's amazing. Yeah. So, you just perfect school, perfect everything. Also, um your ML notes like they all they go all the way down to like calculating gradients by hand, but like tools like can do like math for you.

Would you say so you do it yourself because you really love like math? doesn't sound like or like or I mean I think like for these type of things understanding things like super fundamentally is really important to me because I feel like it's the only way that I can have like an endto-end understanding of whatever the the topic is and actually be able to convince myself that I fully understand it. Otherwise, I feel like I I don't know. I feel like if I don't go as deep as like in this case, um, you know, updating weights, uh, then I feel like I can't really confidently say that I have a strong intuition for how it works.

Um, yeah, I think I've always been like that. And I think it's like part of my curiosity as well, where I just sort of am always like, okay, you're telling me this, but why? And then wanting to go super deep until I feel like, okay, like I have a good like mapping of how this works in my brain. Yeah.

Um, and I think for ML specifically, it requires that much of like in depth going deep and understanding. Yeah. Wait, that's really good advice for anyone because especially now with like with AI and like you get get answers really quickly. I think like for us humans, our role is to really like understand it.

Um, and then you can like and use like AI to kind of almost almost like elevate that knowledge. And I think Yeah. Yeah. It's like this weird thing where I feel like now with like codecs and and like other tools, you can literally just uh be infinitely curios or curious about things and like ask or like open up millions of threads and ask millions of questions.

Um and I think sometimes I can feel like that can be a bit overwhelming and kind of there's only so much knowledge that I think you can absorb as a human. Yeah. Um so yeah, I don't know. I I struggle with that sometimes cuz I'm like so curious about a million things and it's like, "Okay, great.

" and um and uh and um and yeah, you have to really like you have to really figure out like what is like the priority of like what's like most important like what it's like and it's just like too overwhelming cuz it's like one minute I'm reading about like how our brain works and the next I'm like asking like about these very niche like uh like ML topics. Uh so it's like I have I have too many interests. Oh, that's amazing though. Like you go in so many rabbit holes and like that's so beautiful.

And um are there any like questions that like like general questions that you like ask like is there like some kind of framework of like what is like a question you always like make sure to ask? Is there something like that like in a toolkit of yours like like these like Yeah. Um I mean when I want to understand a topic I usually nowadays turn to Codex like first and foremost. Um, but I actually feel like Codex is kind of and not just Codex specifically, but like generally models are kind of bad at explaining topics.

Um, like they're definitely like for example, if I ask Codex to explain some like niche ML topic to me and then I ask my manager about it, he his explanation sticks with me much more than what Codex says. But nonetheless, that's another topic. Uh, for this, I just sort of start off by saying like, okay, I'm really interested in learning this topic. Please go very deep into the all the details um uh like highle details of everything I should know to like fundamentally understand this.

Yeah. Um and then from there I basically pick like one one point because usually it gives like 20 points. So then I start with like one point and then I kind of just like um start asking questions about that specific point. I'm like okay well this is interesting.

Can you explain this for me? And then in the realm of um ML there's usually lots of math and then I dive a lot into like the formulas because these formulas like they're just given to you and you know you can take them at face value and be like okay yeah sure but I think if you like dive into it and be like okay like what does this represent? What does this mean? Um or like what are like important maybe like statistics uh concepts that I need to understand to like really like fundamentally understand this.

Um, and then just like asking a million questions. Uh, and then yeah, I think and then sometimes I get too lost in the sauce where I'm like, okay, I'm at this like very unrelated topic now that is like just like completely down a rabbit hole that I I feel like is maybe not productive. So then I have to like reel myself out. But yeah, that's sort of how I I try to do it.

Wow, that is so beautiful. And so like I I'm going to use that like what you said with like with that I'm going to prompt it too cuz I I really want to like I'm really fascinated with like podcasting. We're going to go really deep into that, too. So, I'm going to like I'm going to really going to do that.

We should let it go. Yeah. And then and then what you said about like you like say and then there's like there's one little like you notice like a detail like one little Oh, like you ask why with that and then it's almost like this graph of like it goes. Yeah.

Yeah. It's like a tree. Um and then I I usually feel like so I try to have these like uh block off these hours for curiosity on weekends and I'm honestly unfortunately not very consistent with it because life and it gets busy and or I get busy. Um but when I do have the time for it uh and I have a free weekend, uh one topic that I feel like should take like maybe like an hour to like get an intuition for it.

Yeah. I usually spend the whole weekend diving into it or like multiple hours. Uh yeah. So I I yeah I think I I really try to maybe it's not maybe sometimes it's not super productive because then I like learn less but I do feel like I have like a good or force myself to have a good understanding of like one topic at least.

Yeah. And you just become like super good at learning like you never stop learning like Yeah. I definitely don't want to stop learning. That scares me quite a bit especially since I I left school um and didn't graduate from my undergrad.

I I I think it's really easy when you join a company to sort of just get sucked into the day-to-day and just kind of do whatever is asked of you. And I think a lot of times you're not really applying a lot of like I mean I think research is a bit different cuz like as the job title you are a curious person but I think engineering I think sometimes you can just get in the habit of like looking at linear tickets and just like doing that and not really being super curious about like other stuff. So, uh, at least when I was at Sky, I did this a lot. Um, and I do it now that I'm open at OpenAI, but I have more ground to be curious at my job now.

But then I would always like do these deep dives to like make sure that I'm like still being curious and and you know, exercising that uh because I feel like it's the best way to grow too. Yeah, that's so just like that like energy like you just like gave me some more of that like curiosity like that like deep diving. It's so amazing. And like so you have all this all these things you're learning like what what seems some of your advice of how do you like store this like information?

So like cuz you have it all in chats but how do you make sure like you where do you store it so you can look back or I usually do like simple pen and paper otherwise I cannot like absorb it and then uh after that without the use of like models I try to like distill my notes into like a a more like nice write up in notion or something um just so that I can like really confidently say that I understand something. Um, and then, uh, sometimes if it's like unrelated to work stuff, I will like speak it back to my boyfriend. " And then I will like like kind of like make him ask me questions to make sure that he understands it because if I can explain it to him, then I feel like I can maybe almost like confidently say that I have a good understanding of it. Um, so that's so cute.

Okay. Find someone who you can just like like sit down with and like exchange like do you talk a very patient person. He's great. Wait, that's amazing.

Yeah. Wait, that's such a good actually such a good quality to have in a partner and like and hopefully like anyone you find is also as curious as you. So like you can like both like and you guys have both like different interests, different like both like go into different rabbit holes. That's like a really cool date idea too by the way.

Like it is really fun. Yeah. Yes. So, um there's like a really fun event that my friend told me about called it's like a rabbit hole athon and uh people it's like um a hackathon but instead you just do like an idea.

So, you go super deep into a topic and then you put together a slide like PowerPoint presentation and present it to people. And I just like really love this concept. Um, so I mean I've been doing this more, but I've been wanting to tell my boyfriend like we should just like have an hour each. Yes.

And then just like rabbit holeathon and then come back and do our pres PowerPoint presentations, but he hasn't done it yet, so I need to like convince him more. Yeah. No, do it. Do it.

I want to do it, too. What the heck? Oh my god, that's so fun. Okay, wait.

This is so so awesome. Wait, because you're learning together. Oh, that's so cute. Um, okay.

Wait, I love it. But I think we covered like so much and yeah, especially love like talking about all those like like other passions you have like what are some like other like passions or like hobbies that you want to like get into outside of work and Yeah. or you're getting into? Yeah.

I mean honestly I feel like work is my life. But I I used to really love running a lot. Um, I felt like that was a really good way to clear my mind. Um, I also really like baking, although I don't really have enough time to do it.

But, um, I made like these Swedish buns at one point and that was like super fun and I have been like trying to find opportunities to like continue to bake. Um, I always joke with my co-workers. I'm like, my post AGI plans are to move to Paris and open a bakery. Oh my god.

Oh my god. But I don't know if that's going to happen. Wait, that's such a uh Oh my god. Yes.

Yeah. Yeah, I think I mean I think it'd be so fun. It reminds me so I actually don't really know how bakeries run to be very honest, but my very naive understanding is that it's would be sort of similar to how my dad's produce company worked with the super early mornings, very like laborious like hard like physical labor. Uh so I'm like I feel like I could handle it.

Uh but I feel like in Paris it's very competitive but that's exactly what I want. And I want to have like peak competition. So that it forces me to like be one of the best bakeries in the world and hire the best uh like what do you call those? Like bakers the best like best like Yeah.

Michelin star bakers. Bakers. Yeah. Or maybe I'll go to baking school.

Oh my god. Wait. And like France pairs. Exactly.

Wait. That would be such a cool thing to do. Okay. This is incredible, Brooke.

Thank you so much. We dived into so much and really excited for everything you do.