Interviewed by CatGPT: How I Rebuilt My Career Around AI

Transcript

[00:00] speaker_0: By the time Parth Patil met Reid Hoffman, he had already spent a year hacking alone in his Los Angeles apartment. A Bay Area native and the child of two software engineers, Parth had built a successful career as a data scientist. But when ChatGPT came out, he watched tasks that used to take him days get done in seconds. And in an instant, everything he thought he knew about work, intelligence, even his own job was upended. At that moment, the rest of the world was just beginning to hum and haw about what all this AI stuff would mean, but not Parth. Instead, he did something that most of us never will. He liquidated his 401K, walked away from a steady paycheck, and stepped into self-imposed isolation. For 14 hours a day, seven days a week, he worked alone in silence, trying to answer one question, which was, "How far can this tech really go?" For a full year, he built in the dark. Before anyone was talking about agents, he was building them. Before the word vibe coding was even a term, he was inventing the practice, all the while with no press and no funding, just pure obsession. Then one day, a mutual friend introduced Parth to Reid Hoffman, the co-founder of LinkedIn and one of the most influential minds in Silicon Valley. What was supposed to be a very simple meet and greet ended up turning into a five-hour conversation between Parth and Reid about the future of AI. And when it finally ended, Reid turned to his chief of staff and said, "Do whatever it takes to get Parth on our team." In today's video, we find out what Reid and Parth talked about in those five hours, what compelled Reid to hire Parth as his AI expert on the spot. We hear firsthand from Parth what he saw before the rest of the world did, what drove him to walk away from stability, why he believes playing with AI is more important than working with it, and why he says even our wildest dreams are not big enough for a future with AI. Before we continue, I'm pleased to present today's interview with no sponsors and no ad breaks, so if you haven't already, please like and subscribe this baby-growing YouTube channel. And with that, enjoy the show. Hey, guys. How's it going? It's Katherine from KatGPT, and today I'm joined by my friend Parth Patil, who is basically the resident AI expert that works directly with Reid Hoffman, who's one of the co-founders of LinkedIn. Parth, thanks so much for joining me.
[02:25] speaker_1: Oh, thanks, Cat. Yeah, this is gonna be fun.
[02:27] speaker_0: Parth, in case you weren't aware, you are kind of an anomaly. I remember when we first met, like, one of the ways that you described yourself as being red-pilled. That's a term that you coined, not me, and obviously it's a nod to The Matrix, but would love to hear in your own words, uh, what does being red-pilled mean to you, and how did you become red-pilled?
[02:45] speaker_1: Every- everyone has a moment when they're interacting with AI. It happens at different moments. I think for Kasparov, it was when he played Deep Blue-
[02:52] speaker_0: Mm-hmm.
[02:52] speaker_1: ... um, where he was like, he could see flashes of a program that was operating at his level or potentially beyond. And for me, that was at my last job. I was a data, data analyst, data scientist, and ChatGPT came out around then. And I'm watching a language model do this, you know, 100 times faster than me and, like, with answers that I don't even have. It's able to reach into parts of, like, the software stack that I've never had exposure to.
[03:15] speaker_0: Yeah.
[03:16] speaker_1: Well, so that night, I had an anxiety attack.
[03:18] speaker_0: (laughs) Okay, valid.
[03:19] speaker_1: 'Cause I was kind of like-
[03:20] speaker_0: So valid.
[03:20] speaker_1: I was like... Yeah, and I was like, I was like, "Yo, I thought I was pretty good at this thing, like this whole da- I thought this was my career. This is the thing that I'm really good at, data analysis." And so I woke up in the middle of the night. I was like, kind of like, "What am I gonna do about this?" Next day, I sit down, and I'm like, "All right, let's just... Okay, I'm just gonna replace myself." And what does that mean? So I sit down. I'm like, "Okay, let's build a bot that can use, you know, ChatGPT and talk to a database." And, uh, and, you know, just talking to ChatGPT, it showed me how to go from like natural language to SQL. It showed me how to make the API calls, and I had it working on my computer. Actually, that wasn't even the first problem I asked it. The first thing I asked ChatGPT was, "Teach me how to clone my voice."
[03:57] speaker_0: (laughs)
[03:57] speaker_1: And, uh, and it wrote the first version of that program. And then I was like, "Teach me how to run this program." And then it showed me how to set up my computer.
[04:04] speaker_0: And, wait, this is really important, because you were not a, a programmer.
[04:09] speaker_1: Yeah, not in a, not in any... I was a data analyst, so we use-
[04:13] speaker_0: Right.
[04:13] speaker_1: ... we use SQL and Python to do work, but it's not the same as building entire applications. That's, like, a couple more programming languages. And my buddy was like, "Parth, you should be talking to GPT-4." The Echo Hive, he was like, "You should be talking to GPT-4 and have it teach you programming, 'cause this thing is better than most programmers we know."
[04:28] speaker_0: And this is when you were using it to just, like, clone your vo- Like, this is-
[04:31] speaker_1: Yeah.
[04:31] speaker_0: ... after you had the panic attack, and you're like, "Oh, my God."
[04:33] speaker_1: Yeah, the first thing was, like, clone my voice, and then it writes the program. And I'm like, "This is supposed to take six data scientists a year." And now because the language model has already studied all the existing code-
[04:44] speaker_0: Mm-hmm.
[04:45] speaker_1: ... like, you can... If someone else has done it and then the language model has been trained on that data set, you can basically reach into that capability even though you didn't invent it.
[04:52] speaker_0: Yeah.
[04:53] speaker_1: So you're basic- you're able to, like, get these Lego blocks from across of all of what humans have worked on.
[04:57] speaker_0: Mm-hmm.
[04:58] speaker_1: That's general knowledge, right?
[04:59] speaker_0: Mm-hmm.
[04:59] speaker_1: So, like, the, the model has studied everything humans, like, a lot of what humans have worked on.
[05:03] speaker_0: Yeah.
[05:04] speaker_1: So when you ask for something, you're probably not the first person to do that thing, and there might already be a solution inside the model for that thing. So you can basically ask for the model to reverse engineer things that are possible. And so that was voice cloning. And then, and that, that only took me two hours, and I was kind of just like, "Okay." Like, "This is kind of a cheat code."
[05:21] speaker_0: Like, that was too easy.
[05:22] speaker_1: Yeah, yeah, yeah, yeah, yeah.
[05:23] speaker_0: Yeah.
[05:23] speaker_1: Yeah, yeah. So the second prom- the second question I asked ChatGPT that day after clone- teach me how to clone my voice, the second thing was, "Teach me how to build a GPT-powered chatbot."
[05:33] speaker_0: Mm-hmm.
[05:35] speaker_1: And it wrote the first version of that program, and then I got it running. And there's a video on my Instagram of the first time I made that call so you can see me. It's like, "What are you?" And it replies. It's like, "I are a large language model developed by OpenAI."
[05:45] speaker_0: Right.
[05:45] speaker_1: And it was running on, like, it's, like, 20 lines of code. See, when you have GPT, you can make things much more easily, including chatbots like ChatGPT, you don't-
[05:53] speaker_0: Right, it's so meta.
[05:54] speaker_1: It's so meta, right?
[05:55] speaker_0: Yeah.
[05:55] speaker_1: It's like sticks and stones aren't gonna tell you how to make fire.
[05:57] speaker_0: Mm.
[05:58] speaker_1: Right? But th- but this is intelligence, right?
[06:00] speaker_0: Oh, say that one more time. That hit.
[06:01] speaker_1: Like sticks and stones aren't gonna tell you how to make fire.
[06:04] speaker_0: But?
[06:04] speaker_1: But intelligence is entirely different. You could just ask for things-
[06:07] speaker_0: Come on.
[06:07] speaker_1: (laughs)
[06:08] speaker_0: That's insane, dude. When you put it like that.
[06:11] speaker_1: Yeah.
[06:11] speaker_0: Okay. Okay. Okay.
[06:12] speaker_1: Yeah.
[06:12] speaker_0: So, so I'm, I'm sorry. I keep bringing you back to this moment 'cause I wanna like imagine being there with you in this moment.
[06:17] speaker_1: Yeah. Yeah.
[06:17] speaker_0: This is like, it's happening. You're getting red-pilled.
[06:20] speaker_1: Yeah.
[06:20] speaker_0: You're in the, you're in the Clubhouse group, you're with the Python guys-
[06:22] speaker_1: Yeah.
[06:22] speaker_0: ... and you're like, "Oh, my God. Everything's changing."
[06:24] speaker_1: Another way to say it is like, it's like, this is feeling the AGI is like what other people might say.
[06:29] speaker_0: Mm-hmm.
[06:29] speaker_1: When I showed these capabilities to my manager at my job, um, you know, as we were just looking at these programs-
[06:34] speaker_0: Yeah.
[06:34] speaker_1: ... I was like, "I just tell it to work on it, and then I go sit by the pool for a few hours, and then I come back and I'm here, like, 'Whoa, I didn't know you could even do that.'"
[06:40] speaker_0: Yeah.
[06:40] speaker_1: So like language models are showing me what's possible-
[06:42] speaker_0: Right.
[06:43] speaker_1: ... outside of what I've been personally exposed to and the projects I've worked on in my life.
[06:45] speaker_0: Right. Right.
[06:46] speaker_1: And he was just like, "Oh, okay. This must be an AGI paradigm." And I hadn't heard AGI outside of science fiction.
[06:52] speaker_0: (laughs)
[06:52] speaker_1: And I was kinda just like, "Are you ser- Oh."
[06:54] speaker_0: Uh. Ugh.
[06:55] speaker_1: Oh. And then it's like, "Oh, this is 100 years early."
[06:58] speaker_0: I remember when we first met, one of the things that you told me in kind of this path, uh, we were talking more about kind of like how you went from data science into this red-pill moment into like the next step that you make, made. And one of the things that you told me is you sold off your 401K and you sold all your crypto.
[07:14] speaker_1: Yeah.
[07:15] speaker_0: Because you had realized that you needed to exist outside of the economy. Bridge those two for me. Tell me what it actually means in practice to exist outside of the economy. And why did you, why did you feel like that was necessary?
[07:28] speaker_1: Yeah. Yeah. Okay. I don't recommend this to everyone.
[07:31] speaker_0: (laughs)
[07:31] speaker_1: And honestly, after I say it, some of my friends that are smarter than me, like, "Dude, you, you didn't have to."
[07:36] speaker_0: (laughs)
[07:36] speaker_1: "You could've, you could've like taken a loan."
[07:37] speaker_0: "A little extreme."
[07:38] speaker_1: "You could've like moved back home. You could've done that."
[07:39] speaker_0: (laughs)
[07:40] speaker_1: Uh, bi- but basically, okay, so I was working at Clubhouse and I di- that's when I discovered the models-
[07:45] speaker_0: Yeah.
[07:45] speaker_1: ... and, uh, how powerful they were. But then, uh, we had layoffs and, uh, I was kinda laying, like the day I got laid off, I was laying there in bed, looking at the ceiling. I was like, "Wait, this is great."
[07:56] speaker_0: Mm-hmm.
[07:56] speaker_1: "Now I get... Now I'm not moonlighting this. Now I get to do this like th- every day-"
[08:01] speaker_0: Yeah.
[08:01] speaker_1: "... with every single ounce of like my waking energy."
[08:04] speaker_0: Yeah.
[08:04] speaker_1: Like so I, I was like, "Okay, what am I gonna do tomorrow? I'm gonna wake up, brush my teeth, order coffee, sit down, and then talk to the language model and figure out how to make it useful." But if you do that... And, and I didn't have, uh, a job, right? So I was kinda like, "Okay, well, I'm gonna do this every day, Saturday, Sunday as well."
[08:20] speaker_0: And were you like, "I'm just gonna do this while I, like, look for a job and I'll still find a job, but like I just wanna do this while I have the time," or-
[08:26] speaker_1: So we had severance. It was pretty generous.
[08:28] speaker_0: Okay.
[08:28] speaker_1: I had like four months of severance.
[08:30] speaker_0: Okay.
[08:30] speaker_1: And so I was like, "Okay, I will at least do this for four months."
[08:33] speaker_0: Okay.
[08:33] speaker_1: And I figured four months doing this every single day, that's like, it's like 120 days, right? Like at the end of that, like it's like 100-day sprint. You're gonna sit there, you're gonna sit down and you're gonna just like ask the dumb question. That's the other thing is like you get to ask dumb questions.
[08:46] speaker_0: Oh, yeah.
[08:46] speaker_1: You, no- it's never gonna judge you, right? So I'm like, "Okay, I don't know how GIT works. Teach me."
[08:50] speaker_0: Yeah.
[08:51] speaker_1: And then like, yeah, maybe I might be ashamed to ask that of my super smart engineering roommates from back in the day.
[08:55] speaker_0: Totally.
[08:56] speaker_1: But language model doesn't judge you. It actually just tells you how the world works.
[09:00] speaker_0: Mm-hmm.
[09:00] speaker_1: And so you should ask a lot of dumb questions.
[09:02] speaker_0: Mm-hmm.
[09:02] speaker_1: And-
[09:03] speaker_0: Preach.
[09:03] speaker_1: ... I figured 120 days, I was like, "Okay, at the end of 120 days, I'll probably have a clearer picture of like where this tech is going and how to use it," but yeah.
[09:11] speaker_0: Do you think you would have had the courage to do, take the leap if you hadn't gotten laid off?
[09:14] speaker_1: No. Uh, and because I did that, like s- eight months later, one of my best friends from college, uh, Saad, he called me, he's like, "Parth, I realize y-" In his case, he's like, "I realize you just can't moonlight this, so I'm quitting my job."
[09:27] speaker_0: Shut up.
[09:28] speaker_1: Yeah. And now he works at Anthropic. He realized like working on this after work wasn't enough for him 'cause he was starting to get more pas- he was drawn into this as a passion. And then, yeah, a year later, now he works out of Frontier Lab. I mean, after I ran out of severance, so I was like, okay, that was four months in. Then I was like, "Yeah, I'm not really done."
[09:45] speaker_0: Had you applied for any jobs at that point?
[09:47] speaker_1: No.
[09:47] speaker_0: Okay.
[09:47] speaker_1: No. I was... Look, because I would talk to people that I, like, uh, like I would t- s- I talked to CFOs, CEOs, I'm like, "You know you could, you could do this thing and like all of a sudden, that whole, like your, your entire data set, like now you have semantic meaning. You understand who the customer is because language models can interpret that."
[10:03] speaker_0: Hm.
[10:03] speaker_1: "Like it's not just how many customers you have. It's like, 'Oh, this one's a small business. This is a vegan restaurant.' Like there's so much more nuance."
[10:09] speaker_0: Sure. Yeah.
[10:09] speaker_1: But they couldn't see that, and so I was like, "If I work for you, then I'm not gonna get to think like this. Then I'm not gonna get better at thinking like this. Then I shouldn't work for you because actually, I think you're gonna agree with me six months from now anyways." And that has continued to play out. Almost every day I get a text. I'm like, "Oh, dude, you were, you were right." It was at the end of those four months that I, a lot of my friends, people I used to work with, like the data analysts, the data, all the data people in my life, I had already told them and, and we, we've been talking about this. We're... And like people I used to work with, Olivia Chen, uh, I used to work with her at Clubhouse, and she ended up working at Replit and then building Replit into an agent. But because I was working, uh, with her at Clubhouse, we saw the power of the language models.
[10:47] speaker_0: Mm-hmm.
[10:48] speaker_1: And we would text each other and we would be like, "Isn't it obvious that this is going to change all of data analytics, data engineering? All of that is actually well within the realm of do- like, like the data, the AI power data analyst is going to be 1,000 times more useful than like a normal data analyst."
[11:02] speaker_0: Right.
[11:03] speaker_1: And she... So we would... So I would get these texts. We would have these conversations, uh, 'cause I'd go to the mirror and I, and I would just be like, "Are you going insane?" Like this is... You know, no one in the real world's like really figuring this out except for a few people. But then my friends were just like, "I think what you're doing is better than the way we do things at, in the real world anyways. Like this is better than the way I do my job, so you should keep pulling, you know, turn over some of these cards."
[11:25] speaker_0: Well, ho- and hold on. I'm just gonna stop you there-
[11:26] speaker_1: Yeah.
[11:26] speaker_0: ... right there for a second because most people would be like, "I just ran out of severance."
[11:31] speaker_1: Yeah.
[11:31] speaker_0: "Uh, I, how am I making money next month?" Like were you scared? What, what... I'm trying to get inside-
[11:37] speaker_1: Yeah.
[11:37] speaker_0: ... like your-
[11:38] speaker_1: Yeah.
[11:38] speaker_0: ... calculus here of like-
[11:39] speaker_1: Yeah.
[11:39] speaker_0: ... life planning.
[11:40] speaker_1: Yeah. Well, either... Okay, this is the gamble it was like I'm making is like either a year from now, I have no money and I know how this tech h- works, and then I'll be able to go join teams and deploy this tech. Or, you know, worst case, like maybe it's not the future and maybe it was a complete waste of time. And then my friends will just hire me anyways.
[12:00] speaker_0: (laughs)Period.
[12:02] speaker_1: So like, like you, how it, it's like, you, yes, you could, you could have zero cash.
[12:05] speaker_0: Yeah.
[12:06] speaker_1: But if you have a ne- your network, right? Like your network is your net worth, right? Like, though, you, how can you possibly go to zero if you know a bunch of people and they're, like, you know, working hard in their own careers and they're learning? Like worst case, you're just gonna ask them for help and then, like, someone might help you. You're not actually alone. So all that was like how I justify it. Also, I have two cats, so I'd go to the cats and be like, "The boys ain't gonna starve."
[12:28] speaker_0: (laughs)
[12:29] speaker_1: Like, "Trust, trust," right? "The boys ain't gonna starve. Like the boys are gonna have wet food once again."
[12:33] speaker_0: (laughs)
[12:34] speaker_1: And so you come up with these, like, mantras, right?
[12:36] speaker_0: Yeah, Daddy's gonna take care of you.
[12:37] speaker_1: Yeah. You come up with these mantras because you're like, "Yeah, most people can't understand this." And I have a few-
[12:42] speaker_0: Yeah, did you feel like so alone?
[12:44] speaker_1: Yeah.
[12:44] speaker_0: I mean, everyone's looking at you like, "You're wrong. This is crazy," whatever, and you're like, "I really think this is-"
[12:50] speaker_1: Most people would think that, but I only care about a few people's opinions anyways.
[12:54] speaker_0: Whose, whose opinions do you care about?
[12:56] speaker_1: Uh, the people I've worked with that understand me.
[12:58] speaker_0: Okay.
[12:58] speaker_1: Like they understand my obsessing. Like I have an obsessive personality.
[13:02] speaker_0: Mm-hmm.
[13:02] speaker_1: So like they, they understand my quirks. And then some of, like the people that are smarter than me that I always ask for, for advice, right? They might not be in the same like risk-taking mindset as me.
[13:11] speaker_0: Mm-hmm.
[13:12] speaker_1: But when I, I'm like, "Hey, I think this is the future of analytics. I think this is the future of software engineering," even though it's like kind of jank and it's like very early stages, but v- it's obvious that I would rather speak to the computer in English and have it start working, um, than for, than for me to like try to do every single thing by hand, um, the old way. And those people would be like, "No, you're right. That's leverage. Like this looks like it's more leverage than that, uh, than the way we currently do things." So if, if the people smarter than me are like, "No, you should keep, you know, turning some of these cards over," then it doesn't feel like a gamble. Like yeah, like watching the net worth's number goes, go down is kind of insane. But, um, uh, but like, uh, you know, like I spent a month just talking to language models about music, and I knew I wasn't gonna make money that month. And it was h- it was like, you know, one of my friends was like, "Pars, just pick up some coins along the way, man." And I was like, "No, I just don't think it's time yet."
[14:04] speaker_0: Mm-hmm.
[14:04] speaker_1: Because I'm over here asking it, I'm like, "Yo, teach me about..." It's like, "Teach me about this knob on, on my like digital audio workstation." It's breaking down. It can go very deep into every concept in all of like music production.
[14:14] speaker_0: Yeah, yeah.
[14:15] speaker_1: And a lot of th- And I always thought that like, what was holding me back from learning programming? What was holding me back from learning music production? Because I don't have someone better than me sitting next to me-
[14:23] speaker_0: Mm-hmm.
[14:23] speaker_1: ... that can just, I can just ask dumb questions to. And it's like if you use it for your job only, you're missing all of the things you could be using it for that are like more important to you. I'm not gonna tell you what to do-
[14:33] speaker_0: Yeah.
[14:33] speaker_1: ... other than you should aim this at things that you're very passionate about because things that you didn't think you had time for. So I was like, "Oh, what if I wanted to make a game? Well, I gotta go get a degree." No, that's definitely not gonna happen.
[14:45] speaker_0: (laughs) No.
[14:45] speaker_1: Like, like start a different career? Like no, now you get to reach into all these skill sets that would have otherwise taken other lifetimes.
[14:51] speaker_0: Yeah.
[14:52] speaker_1: Yeah.
[14:52] speaker_0: But don't you think that for somebody to take your advice and do that-
[14:56] speaker_1: Yeah.
[14:56] speaker_0: ... they're clocking out of their nine-to-five.
[14:58] speaker_1: Yeah.
[14:58] speaker_0: They're exhausted. Now they gotta take this LLM thing and point it at something that they're deeply passionate about.
[15:02] speaker_1: Yeah.
[15:03] speaker_0: How many people have you found are actually that curious and that excited about learning that they take that advice and do it? Because this is what, this is what I mean when I say you're an anomaly, and I mean it in nothing but the best of ways.
[15:15] speaker_1: Yeah. Yeah.
[15:15] speaker_0: Is you follow your curiosity off of a cliff.
[15:18] speaker_1: Yeah.
[15:19] speaker_0: Literally. I mean, we're talking about you being like, "I'm gonna watch my net worth go to zero if that's what it takes for me to like learn everything I possibly can using this tool." Most people don't have that risk appetite, but to your point, they could still learn a little bit along the way. Do you think that people are doing that? Like, and, and don't you... Would you agree that it's not just enough to say, "This is what you should do"? It also requires that inherent ingredient of the insatiable curiosity from the person on the other side of that conversation.
[15:47] speaker_1: Yeah. You're right. It's not most people. Um, like the reason I say play with it is because when you're, uh, w- when you do something that's like you're intrinsically motivated by, no one has to tell you to do it.
[15:58] speaker_0: Mm-hmm.
[15:59] speaker_1: And like, if, you don't have to drag yourself out of bed to do it. You actually, like you get to a pl- like if it's your, if it's your passion, if, then, you know, you're, it feels fun, right? Than sitting in front of a computer for 14 hours a day, which, and also is probably not recommended for everyone, but it doesn't feel like a slog.
[16:15] speaker_0: Yeah.
[16:15] speaker_1: It's more like, "Oh, I don't actually have enough time in the day." (laughs)
[16:18] speaker_0: Yeah.
[16:18] speaker_1: "For everything that I now want to do."
[16:20] speaker_0: Yeah.
[16:20] speaker_1: "Because now I have superpowers."
[16:22] speaker_0: That's just it for me. It's superpowers. I was thinking about this the other day. When I was in high school, I was a language nerd.
[16:27] speaker_1: Yeah.
[16:27] speaker_0: I took Spanish, French, Chinese, and I was teaching myself Portuguese on the side. I just loved it. And I remember at one point talking about it with my parents, who only speak English. And they, I was, explained it to them as like, "I feel like when I learn a new language, I unlock a superpower."
[16:40] speaker_1: Yeah.
[16:40] speaker_0: "I literally, not, this is not a metaphor. I literally feel like someone who can fly because I now can have conversations with, build relationships with, fall in love with 300 million additional people-"
[16:52] speaker_1: (laughs) Yup.
[16:52] speaker_0: "... that I couldn't have done it with before."
[16:53] speaker_1: Yeah, and you like understand their culture way more richly.
[16:55] speaker_0: Dude.
[16:56] speaker_1: Yeah.
[16:56] speaker_0: It's, it's insane. And then with AI, this is, since falling in love with learning languages, this is the only other time I felt that similar, like-
[17:04] speaker_1: Yup.
[17:04] speaker_0: ... "I feel like I have f-f'ing superpowers."
[17:08] speaker_1: Yeah, yeah, exac- exactly. And that's how I felt when I saw this thing writing better code than me. I was like, "Wow, now I'm not limited by everything that I've ever worked on. I'm limited by, I'm limited by the question I ask."
[17:17] speaker_0: Yeah, you're not limited by your intelligence.
[17:19] speaker_1: Yeah.
[17:19] speaker_0: You're limited by the AI's intelligence. (laughs)
[17:21] speaker_1: Yeah, which is like the collective intelligence.
[17:23] speaker_0: Right.
[17:23] speaker_1: Right?
[17:23] speaker_0: Right.
[17:24] speaker_1: And you just have to wield it, bring it out of the AI into the real world.
[17:27] speaker_0: Right.
[17:27] speaker_1: And that, that's, that's superpowers. It's cognitive superpowers.
[17:30] speaker_0: You work for Reid Hoffman.
[17:32] speaker_1: Yeah.
[17:32] speaker_0: Who's the founder of LinkedIn.
[17:33] speaker_1: Yup.
[17:34] speaker_0: Um, how did you meet Reid and why did Reid hire you?
[17:37] speaker_1: This is like a year into me exploring AI independently. You know, I met Ben, uh, Ben Reilles, who works for Reid. And we were working out of a beach house in, uh, Playa, Playa del Rey. And when cust- custom GPTs came out, he kee- he goes over to me. He's like, "Hey, do you know how to like, you know, maybe build one on top of a personality of someone that has a lot of writ- writing and thought work, uh, like thought leadership?" And I was like, "Oh, yeah." And like, so I built a custom GPT on top of Reid's, um, corpus of work, scraped all his books. Uh, no, I got actual access to his books. (laughs)
[18:11] speaker_0: (laughs) Nice save.
[18:11] speaker_1: Scraped the podcast though. Uh...
[18:13] speaker_0: (laughs)
[18:14] speaker_1: And I created this, like, chat bot that could conversationally retrieve parts of his, uh, his corpus of thinking and all of his, like, writing over the last 20 years. And he's, like, a legend within Silicon Valley, so that ends up being a hu- you know, very interesting chat bot.
[18:27] speaker_0: Mm.
[18:27] speaker_1: When you talk to the chat bot and it's like, "Pretend to be Reid. Also, here's a bunch of things that he said-
[18:31] speaker_0: Right.
[18:31] speaker_1: ... over the last 20 years-
[18:33] speaker_0: Right, right, right.
[18:33] speaker_1: ... about technology, entrepreneurship." And then, um, one day, I, uh, I was invited, like, the Ben was like, "Oh, you should come up to the Bay Area." And I went up to the Bay Area, and I got to meet Reid. And, uh, and he was like, "You should come back tomorrow. We'll, we'll go deep on this topic." And so I went back the next day and we talked for five hours-
[18:52] speaker_0: Mm.
[18:52] speaker_1: ... about AI, uh, language models, basically this conversation right here, but, like, drawn out for five hours. No, actually, pretty much this conversation, but for, for five hours, and exploring like, okay, here's where, where I'm seeing... And this is, this is a year and a half, like this is a year and two, two months ago.
[19:08] speaker_0: Okay.
[19:08] speaker_1: So a lot of the things I was kind of saying are playing out actually played out in, in the last year. And, and he's like, he's an early investor in OpenAI. So for him to, like, he was probably the first conversation I had where I was like, "Oh, I'm not insane." Like, this is actually, like, like, this is actually, uh, what's happening to, to, like, the world of computing.
[19:27] speaker_0: How did that feel? I mean, you're, you're sitting in the presence of a Silicon Valley legend-
[19:32] speaker_1: Yeah.
[19:33] speaker_0: ... who is making you feel maybe like maybe for the first time that you're not insane. And not only does he want to chat with you, he wants to, he's talking, he's, he's picking your brain for five hours.
[19:43] speaker_1: Yeah.
[19:44] speaker_0: What did you feel like in that moment?
[19:46] speaker_1: I mean, it was a super fun conversation about, like, you know, if you can do this, then you can do that. Like, here's what it, like, what does that mean for, like, this entire, like, business vertical? What does it mean for that, like, entertainment?
[20:00] speaker_0: Yeah.
[20:00] speaker_1: What does it mean for, like... And, and so the, the questions, all of a sudden, it's like, yeah, I was like, "Oh, these are the right questions."
[20:06] speaker_0: Mm-hmm.
[20:06] speaker_1: Like, finally, like, we can have this, like, conversation about the implications of, of, like, this on knowledge work, which are huge.
[20:13] speaker_0: Yeah.
[20:14] speaker_1: Um, I mean, it was, it was very s- I mean, I, I think, like, it was after, after we started, we finished talking, I was like, like, it clicked to me. I was like, "Man, like, I've totally listened to Masters," like, "Masters of Scale." Like, I listened to his podcast, Masters of Scale, growing up, and it's how I got into startup.
[20:27] speaker_0: Yeah.
[20:27] speaker_1: So, you know, I've been following his work since I was probably, like, 15.
[20:31] speaker_0: Hm.
[20:31] speaker_1: Like, like, 12. Silicon Valley, it's like he's... Yeah, PayPal. It's like even before OpenAI, before-
[20:36] speaker_0: PayPal mafia.
[20:36] speaker_1: Yeah, it was PayPal mafia, right? So it was crazy. It, uh, and then he was like, "We should work together." You know, I said yes. And the last year and two months, I've been working with Reid. A lot of different special projects. Um, mostly, like, getting to do the same thing, but, like, with, without being completely alone is pretty key.
[20:53] speaker_0: Hm. This one's a, a curveball. You don't have to answer it if you don't want to, but I'm curious.
[20:56] speaker_1: Yeah. Yeah.
[20:57] speaker_0: Um, you've a very specific image of where we're headed in the future.
[21:00] speaker_1: Yeah, yeah.
[21:01] speaker_0: Um, how does that affect dating?
[21:06] speaker_1: Do I gotta not talk about AI?
[21:09] speaker_0: Yeah. (laughs)
[21:10] speaker_1: Yeah, it, it's, it's very hard. It's very hard. And then I'll be on a date and she'll be like, and she'll be like, "Uh, can you just, like, not talk about this?"
[21:17] speaker_0: No way.
[21:18] speaker_1: And then-
[21:18] speaker_0: She'll ask you to shut up?
[21:19] speaker_1: Dude, yeah, well, s- Well, a girl once told me that, like, literally last month, and I was like, the whole date in my mind, it was, it's like, okay, I'm not saying anything, but my mind is thinking about it. Then I just didn't speak for, like, 20 minutes. I'm like, "This is..."
[21:30] speaker_0: (laughs)
[21:30] speaker_1: It's like, "Wow, did I really become, like, that?"
[21:33] speaker_0: (laughs)
[21:33] speaker_1: Oh my God. I mean, it's a, it is a wake-up call for sure.
[21:36] speaker_0: That's hilarious. Uh-huh.
[21:36] speaker_1: And I'm just like, "I can't date this girl."
[21:40] speaker_0: Yeah.
[21:40] speaker_1: Like, was what I realized.
[21:41] speaker_0: Yeah.
[21:41] speaker_1: Just 'cause, like, 'cause it's like, "Okay, I gotta turn my brain off." It's like, mm, I don't have to. I mean, I might meet someone that actually, like, she, she, she might actually be on my, like... Like, I can f- It just, it just means that I'm not, like, I'm not interested in dating every girl. That doesn't make any sense. But, like, the right one, I'll meet her. It's a wake, it's a wake-up call.
[21:59] speaker_0: No, but I think it's cool. I mean, if you have something that you're that passionate about-
[22:02] speaker_1: Yeah.
[22:02] speaker_0: ... that it's like every conversation you wanna be like... 'Cause, 'cause your light starts to shine, right?
[22:07] speaker_1: Yeah.
[22:07] speaker_0: When you get to, like, talk about the thing that really lights you up, literally.
[22:11] speaker_1: Yeah, yeah.
[22:12] speaker_0: As you know, in the tech world, it's a very common question to ask someone, "Oh, so are you technical?" Especially if, let's say, you're meeting, like, um, someone who's founding a company and you want to get a sense if they're a technical-
[22:23] speaker_1: Yeah.
[22:23] speaker_0: ... or a non-technical co-founder.
[22:24] speaker_1: Yeah.
[22:24] speaker_0: Um, and historically, that question has essentially meant, "Do you write code?"
[22:29] speaker_1: Yeah.
[22:29] speaker_0: Right? Now, in 2025, what do you think qualifies someone to answer that question, "Are you technical," with, "Yes"?
[22:40] speaker_1: I don't even think that's the right question.
[22:42] speaker_0: Okay.
[22:42] speaker_1: Like, like, "Are you technical?" I mean, I know plenty of technical people that have zero skills in prompt engineering, and I worry for them because, like, like, the person that understands how to ask the right question of the model... Like, in my case, I, you know, I'm self-taught. Even data science, I was self-taught. I learned SQL and Python myself over a couple weekends when I was on the job and I needed to. Then I self-taught language models, at least the application. I can't build you a language model from scratch. I'm not an ML researcher. But if you have GPT-4 as an API, I can reimagine software through the lens of, like, these capabilities of language models. I can create chat bots, I can create chat bots that chain together tools. So, but that's all selflearned. And actually, there wasn't... So, so, and a lot of it was vibe coded. I mean, th- Vibe coded was, vibe coding was, like, coined by Andrej Karpathy this year.
[23:30] speaker_0: Yeah.
[23:30] speaker_1: But a lot of my friends are like, "You were the first vibe coder, like, two years ag-"
[23:33] speaker_0: (laughs)
[23:33] speaker_1: "... before it even worked."
[23:34] speaker_0: It's true.
[23:35] speaker_1: Right? And, and if you just scroll my Instagram, you can actually see the old videos. It, it's just me, like, making some, like, weird trippy design, asking for it, and it's like, "GPT taught me how to make this." Also, it wrote the whole website. It also just did most of the work.
[23:48] speaker_0: Yeah.
[23:49] speaker_1: I asked for it, and then here's what you get. So now, people... So a lot of times I'll get, I'll get emails and it'll be, like, headhunters trying to hire me for CTO roles. And imagine, you know, two years ag- Imagine how much-
[24:00] speaker_0: Yeah.
[24:00] speaker_1: ... I had imposter syndrome two years ago, 'cause I was like, "Dude, I don't know the programming language. All I know is I can now make this thing."
[24:07] speaker_0: I cannot imagine you with imposter syndrome.
[24:09] speaker_1: It's, it's very high, actually. Like, I have very high imposter syndrome.
[24:12] speaker_0: Really?
[24:12] speaker_1: Yeah.
[24:12] speaker_0: About what, though?
[24:14] speaker_1: Um...
[24:14] speaker_0: 'Cause you're, you're...Do you see how that's in conflict with the fact that you're like, "F it all, I'm just gonna do my own thing. I have so much confidence and belief that this is the next thing, like, I'm gonna do every, I'm gonna put every..." You literally put every single chip on the table.
[24:25] speaker_1: Yeah. Yeah. But, I mean, there are no experts, right? So that's another mantra. You gotta come up with these, like, mantras that, like, you can just tell yourself in the mirror to, like, fight the imposter syndrome.
[24:37] speaker_0: So what are your mantras?
[24:38] speaker_1: Um, you know, there's no experts. Like, the only p- like, the, the only experts right now are the people that are, like, actually playing with the tech. And I know this. A lot of my friends work at the frontier labs, and they learned from me how to use the models.
[24:48] speaker_0: Hmm.
[24:49] speaker_1: Right? So the person that uses the tech is the one that's learning the most. Um, then there's... You see a lot of talking heads, and then they have the worst pr- uh, perspectives, because they're not playing with it. When I see you, like, asking AI to make you an app, that's, that's the kind of, like, experimentation you need. Like, I once asked AI to, like, GPT-4 early on, I asked, like, it was, like, a coding copilot I was working on. I was like, "Make a list of 100 possible projects we could build in, in Python." And it made a list. And some of the ideas were like, it's like, "Write a haiku." I'm like, "A program that writes haikus? Okay." And then I was like, "Why don't you make the first version of each of these?" And then I went to sit by the pool. (laughs) And-
[25:27] speaker_0: You have a pool in your building?
[25:27] speaker_1: Yeah.
[25:28] speaker_0: Hell, yeah.
[25:28] speaker_1: Yeah, yeah. I mean, that's LA, dude. (laughs)
[25:31] speaker_0: I don't have a pool in my building, but... (laughs)
[25:33] speaker_1: Uh, yeah. And then I'm, like, sitting there, I'm like, listening to podcasts, right? And then I got this program on my computer that's just, like, cranking out code. So then, uh, and then it finishes, and then, uh, I go back and I get the code. And then I go back to the, uh, yeah, I go back to the pool, and I'm just reading the code it's generated. And I was like, "Wow, I didn't know you could do that. I didn't know you could build a presentation in Python. Like, PowerPoint and Excel are Python tools." Then I'm like, "Wait a minute, can you just replace all of McKinsey?"
[25:56] speaker_0: (laughs)
[25:58] speaker_1: Like, these are the kind of realizations I'm having.
[26:01] speaker_0: You're sitting by the pool reviewing computer-written AI code going, "Wait a second."
[26:05] speaker_1: Wait, wait, wait.
[26:06] speaker_0: "Kill McKinsey. Sell."
[26:07] speaker_1: Yeah. (laughs)
[26:07] speaker_0: "Sell, sell, sell." (laughs)
[26:08] speaker_1: So, so, like, yeah, okay, I have no i- th- there's some ideas that don't need to make money. But then you have an idea like that, and you're like, "Wait a minute. W- those guys are charging a million dollars a month for, like, a PowerPoint presentation. Unless I misunderestimate, like, I underestimate their role on the project." But you know what I mean.
[26:22] speaker_0: Yeah, walk that back real quick.
[26:23] speaker_1: Yeah.
[26:23] speaker_0: (laughs)
[26:24] speaker_1: Well, it's like, GPT-4 can write code and can do analysis and can build presentations, and it doesn't charge me a million dollars a month. There were another project, uh, one of the projects they've tried to build was, like, uh, the haiku. And I'm l- this was a crazy one. It's like, a program that writes haikus. And I'm thinking, "How can you possibly write that as a program?" The answer, in my mind, to this program is a language model.
[26:45] speaker_0: Right.
[26:45] speaker_1: Right? It's just like the, just make sure you put GPT in the program-
[26:47] speaker_0: Yeah.
[26:47] speaker_1: ... and be like, tell it to write a haiku.
[26:49] speaker_0: Right.
[26:49] speaker_1: And then it's only 12 lines of code. So it wrote a bunch of code, didn't work.
[26:53] speaker_0: Did it use GPT?
[26:54] speaker_1: No. But then I was, but that's when I was like... (clicks tongue)
[26:57] speaker_0: Oh, it didn't know it could.
[26:58] speaker_1: Yeah.
[26:58] speaker_0: Mm-hmm.
[26:59] speaker_1: I was like, "What? This thing..." I was like, "The answer to this is GPT." And it's like, "Yeah, you're right." And I'm like, "You attacked all these problems and ignored th- the fact that the language model exists."
[27:09] speaker_0: (gasps) Okay. Wow.
[27:10] speaker_1: Di- And that's, like, mind-blowing, 'cause I was like, "Okay, even, even I am not using this yet at the level I need to, because I need to tell it what it doesn't know."
[27:16] speaker_0: Hmm.
[27:17] speaker_1: So this is like the answer to the question of like, if I c- if AI can do everything, then like, what's the point? But I was like, even if it could do everything, it's not clear to me that it will do everything.
[27:28] speaker_0: Hmm.
[27:28] speaker_1: You still have to tell it what to do. And in this case, I was like, "The answer to the question of, like, write a program that can write haikus is to implement a language model."
[27:35] speaker_0: Yeah.
[27:35] speaker_1: And n- it not having that in the solution space told me that, "Oh, wow, anytime I ask it to do something, I might actually know what the right answer is and it might be completely ignoring that. And so I need to provide that context upfront. Otherwise, it's completely ignoring the most powerful tech that's ever come out."
[27:52] speaker_0: Right.
[27:52] speaker_1: Right?
[27:52] speaker_0: Which is a repeat problem, and you and I talked about this-
[27:54] speaker_1: Yeah.
[27:54] speaker_0: ... last night about how it's like, the, the issue, uh, the, of the fact that, like, models are only trained up until a certain date-
[28:01] speaker_1: Yup.
[28:01] speaker_0: ... that training stops, and so anything, any new frontier technology that's come out since then, it will not know how to implement unless you rag it, you, you give it additional context-
[28:10] speaker_1: Exactly.
[28:10] speaker_0: ... you fine-tune it, whatever.
[28:11] speaker_1: Exactly.
[28:11] speaker_0: Um-
[28:12] speaker_1: That, that's like the, the, like, vibe coding has great superpowers. Like, code generation is super powers.
[28:16] speaker_0: Yeah.
[28:17] speaker_1: But also it's like, a huge hole in its memory for the most useful, most recent technology.
[28:21] speaker_0: Yeah, yeah.
[28:22] speaker_1: And, mm, agents will solve that, like RAG techniques-
[28:25] speaker_0: Okay.
[28:25] speaker_1: ... like, uh, you know, just scrape the web-
[28:27] speaker_0: Yeah.
[28:27] speaker_1: ... check the docs, index the docs.
[28:28] speaker_0: Right.
[28:29] speaker_1: But, um, it's not completely solved yet. And s- and there's, there's more, like, I think there's more imperfection in that approach to problem-solving that shows me that this is why we need to e- why, this is why we exist.
[28:40] speaker_0: Right.
[28:40] speaker_1: Is like, to bring them to the real world, ground them in the real world more.
[28:44] speaker_0: What question would you ask instead? What is the new right question?
[28:47] speaker_1: Okay. How good are you at prompting the models?
[28:50] speaker_0: Hm.
[28:50] speaker_1: And, it's, I see this, like, people are freaking out, especially for coding interviews, they're like, "Ah, like, the people are cheating. And like, they're just taking AI that c- already knows the answers and then they're just, like, regurgitating it." There's like, a lot of the, people are making, like, ai- interview hacking AI.
[29:03] speaker_0: Yeah.
[29:04] speaker_1: How you think about the problem is way more important than did you memorize where to put that curly bracket.
[29:09] speaker_0: Right.
[29:09] speaker_1: And the language model now obsolete, makes that, that, like, the syntax obsolete. It was never about do, did you memorize how to write the code? It was about can you think critically about the problem, create an, create a solution that, like, brings the right code together to solve the problem?
[29:24] speaker_0: Yeah.
[29:24] speaker_1: And so, my opinion is that, you know, you got people banning AI in coding interviews, and I'm like, "No, actually, I wanna see your prompts." Like, "Show me. How do you, how did you even make this?" Like, I wanna see, like, sh- li- they sh- y-
[29:36] speaker_0: Yeah.
[29:36] speaker_1: You know, like, maybe you're transcribing them. Like, that's a huge green flag for me.
[29:40] speaker_0: Hmm.
[29:40] speaker_1: Like, you're speaking to the computer, because now I know that you know that, like, speaking is more, it's higher bandwidth. You get more across when you speak. You might arti- uh, or, like, do you screenshot your app and then show it to the AI and then ask for feedback?
[29:52] speaker_0: Hmm.
[29:52] speaker_1: 'Cause turns out, a f- a picture's worth a thousand words. Similarly, when you're trying to debug your app, it would help if the AI could just see your app.
[29:59] speaker_0: Right.
[30:00] speaker_1: But then that shows me that you understand multimodal prompting. I tell y- I'll tell you, like, good prompt engineers are like, it's like spellcasting.Right? They're like, it's like, "Leviosa," not, "Levio-sah."
[30:11] speaker_0: Yeah.
[30:11] speaker_1: Like, and it's like-
[30:12] speaker_0: (laughs).
[30:12] speaker_1: It's like, do you have the right combination of words?
[30:14] speaker_0: Yeah.
[30:14] speaker_1: And like, uh, do you understand how to bring the right context into the context window, whether that's images, whether that's, like, audio, whether that's video? Um, does, do you know what to give the model to give it a best shot at creating what you want?
[30:26] speaker_0: Famously, I'm sure you're aware of this, all the group partners at Y Combinator, um, one of the most well-known, um, accelerators in Silicon Valley, tout that despite advancements in AI coding and that there are predictions that, like, 95% of the code written in 2025 is gonna be written by AI, it's still very important for people to learn how to code. Do you agree?
[30:46] speaker_1: Kind of. Um, so it's like, why do we code? And then it's like, we co, we're, we're trying to build things and we're trying... And coding allows us to buil- it's like code, uh, software is like these ideas. They're not even, it's not even like atoms, right? It's like, it's like ideas, um, represented in digital space that execute some kind of job. And code is kind of like the recipe or, like, the instructions we give to the computer. I think traditionally code was like you had to memorize, it's like SQL, Python, uh, whatever programming language, C++, C#, Java, um, or, like, front-end stuff. So you had to know the language in order to get a certain thing done.
[31:22] speaker_0: Mm-hmm.
[31:22] speaker_1: SQL would allow you to do analysis. Python would let you do data science and machine learning. Um, HTML, CSS, JavaScript allows you to build web. But each code, each, each programming language has its strong suit and its weak points and has its use cases. And then the language model is actually, like, pretty good at all of 'em. And that means that you can just in, you know, you can just tell the language model in whatever language you speak what you want, and then it'll make the first version of that. And then you can ask for revisions. So actually, you are, like, it's not that you're, you're not, uh, in the past you has- used to have to memorize syntax, memorize blocks of code and what they did. Now, you do need to know what they do if you wanna debug the machine. But I think it's like look how many people drive cars that don't know how the engine works. So it is possible that, like, the a- that gets abstracted away from most people. And it doesn't mean that we won't have programmers. Actually, we'll have more programmers than ever before.
[32:16] speaker_0: Mm-hmm.
[32:17] speaker_1: It's just that they won't be coding as much as they'll be speaking in their own language to the computer and, like, looking at the output and being, like, giving it feedback. And if the computer's so smart, it should be able to translate your human instructions back into the right code.
[32:31] speaker_0: Yeah.
[32:32] speaker_1: And then, you know, someone that understands the code more on a, on a, can call the system out when it's, like, hallucinating-
[32:38] speaker_0: Yeah.
[32:39] speaker_1: ... when it's, uh, making mistakes, or can guide it towards the right answer more quickly. That's where expertise comes in, and it's not unique to coding. I think if you're an expert in something, you probably have a very rich vocabulary to articulate what you want and what good looks like.
[32:52] speaker_0: Yeah.
[32:53] speaker_1: Which means that when the AI, um, you know, in the same case, like music, a five-year-old asking for a song is like, "Oh, make me a song that has a trumpet," or, like, maybe-
[33:01] speaker_0: Yeah.
[33:01] speaker_1: ... doesn't even know the word trumpet.
[33:02] speaker_0: Yeah, yeah, yeah.
[33:03] speaker_1: And, and then they'll be able to articulate their vision. But then someone that's like, uh, you know, John Williams, that his vocabulary, he's got an orchestra-
[33:11] speaker_0: Mm-hmm.
[33:11] speaker_1: ... his expectations are so much higher because he has expertise, and he can articulate that with fine-grain.
[33:16] speaker_0: I feel like we're seeing a similar moment with, like, marketing professionals-
[33:19] speaker_1: Yeah.
[33:19] speaker_0: ... or people who work in, um, anything that involves the visual arts right now because with the yesterday's drop of-
[33:24] speaker_1: Yeah.
[33:25] speaker_0: ... uh, 4.0 image generation from OpenAI, it's like there's a bit of a melting moment in the marketing world where they're like, "Oh, shit." And it's like, well, welcome to the game that programmers have been dealing with with LLMs-
[33:35] speaker_1: Yeah.
[33:35] speaker_0: ... for the last, like, year and a half basically.
[33:37] speaker_1: Yeah, yeah.
[33:38] speaker_0: It's like now your expertise is gonna be the only thing that matters 'cause no one gives a shit if you know what all those buttons on Photoshop do.
[33:43] speaker_1: That's right. Exactly. And it's gonna apply, uh, I can't wait to connect these language models to my digital audio workstation so I don't have to click around-
[33:49] speaker_0: (laughs).
[33:49] speaker_1: ... the computer to make music.
[33:51] speaker_0: You're gonna, like, burn your keyboard and mouse in a fire and just, like, hook up a mic and start chatting with your computer full-time.
[33:56] speaker_1: Yes. I'm at that point actually.
[33:58] speaker_0: I believe it. Excuse my French.
[33:59] speaker_1: But, like, transcription is huge.
[34:01] speaker_0: Yeah.
[34:01] speaker_1: And transcription's never been better than it is now. And you can get a lot more, you can say more than you can type.
[34:06] speaker_0: Yeah.
[34:06] speaker_1: I don't write as much. It's kind of, like, one of my downsides. But I speak a lot and I talk to a lot of people.
[34:11] speaker_0: Yeah.
[34:11] speaker_1: And-
[34:12] speaker_0: You're a great yapper.
[34:13] speaker_1: Yapper. Yeah. Well, it's a yapper meta game. Like... (laughs)
[34:15] speaker_0: What did you say?
[34:17] speaker_1: (laughs) It's a yapper meta game.
[34:19] speaker_0: (laughs) Oh, my God.
[34:20] speaker_1: Like yappers and ADHD. Tha- that's, like, the future.
[34:23] speaker_0: Killer combo.
[34:23] speaker_1: You know? Like, I mean, if you're gonna take Adderall to get more productive, you might as well pay for GPT-4.
[34:28] speaker_0: (laughs) Okay, that's the name of the video right there. (laughs)
[34:34] speaker_1: (laughs)
[34:34] speaker_0: Oh, killer. Okay, wait, wait, wait. Sorry. I, I don't wanna lose this thread. When I asked you at the top of this interview, like, "Oh, so what do you do?" You're like, "I hack and I build things using AI."
[34:42] speaker_1: Yeah.
[34:43] speaker_0: Um, that is descriptive of how you spend your time.
[34:46] speaker_1: Yeah.
[34:46] speaker_0: But it's not, it doesn't fall neatly into a category that a lot of people can be like, "I have a picture in my head of what that looks like."
[34:52] speaker_1: Yeah.
[34:53] speaker_0: Um, tell me, like, when you sit down at your computer in the morning, um, or whenever you start your day, like, what are you literally doing on your computer every day? What programs, what applications are you opening?
[35:06] speaker_1: Um, I like to have, um, a coding copilot up all the time.
[35:09] speaker_0: Mm-hmm.
[35:10] speaker_1: So, uh, for me, it's, it's been Cursor primarily for the last, like, year and a half. I was, I was very early to Cursor. Um, but now, uh, vers- very into Cursor, Cursor and then Replit. Replit's great because their agent is really good. And both Replit or- and Cursor programming copilots, really sophisticated. And Replit agent makes it possible to deploy apps to the internet almost instantly.
[35:32] speaker_0: Yeah.
[35:33] speaker_1: So if you actually wanna share something with your friends or publish it, you can use Replit. I usually have, like, nine different messaging apps open.
[35:40] speaker_0: Okay.
[35:40] speaker_1: And by the time I wake up, I wake up around, like, noon, uh, by the time I wake up, people are already firing texts at me of, like, "Did you see this new model? Did you see this new model? Also check this out." And for me, uh, that's, that's been a huge thing because you're just not gonna keep up with everything yourself.
[35:55] speaker_0: Mm-hmm.
[35:56] speaker_1: And so for me it was, like, me practicing, me building with AI and sharing about it-... created a cluster of people around me, friends, a lot of people that are just, like, also kind of getting into the waters.
[36:07] speaker_0: Yeah. Yeah.
[36:07] speaker_1: And everyone's got their own thing. Like, I got friends that are way better at open source than me. I got friends that are way better at video models than me.
[36:13] speaker_0: Mm-hmm.
[36:14] speaker_1: Um, Don Allen is, like, my AR/VR guy.
[36:16] speaker_0: He's the best, yeah.
[36:17] speaker_1: He's, he's, he's, he's like the new Disney.
[36:19] speaker_0: (laughs)
[36:20] speaker_1: Right? Like, if you imagine him having real-time native multimodal models, like, two years from now, I don't... It's like, what's gonna stop him-
[36:26] speaker_0: Yeah.
[36:27] speaker_1: ... from telling, like, the story? So you have people that are better than you at ev- uh, at different things. And then every once in a while, you kind of like, you ping people. It's like, "Is this still, like, the b-" Like, people are like, "Oh, is that still the best code in CoPilot?" I'm like, "Yeah, I think so, unless..."
[36:38] speaker_0: Mm-hmm.
[36:38] speaker_1: But that network intelligence, um-
[36:41] speaker_0: Yeah.
[36:41] speaker_1: Yeah, I think the network is so powerful, right? Like, it allowed me to take the risk. But also now, it's just like, I wake up with a bunch of strong signals onto, as to, like, what's worth, like, exploring, what's worth, like, picking up. If three people are texting me about the same tool, I check that tool out.
[36:56] speaker_0: Yeah, fair enough.
[36:57] speaker_1: Like, on that day, you know.
[36:58] speaker_0: My thing is I refuse to touch X or Twitter with a 10-foot pole.
[37:02] speaker_1: Mm-hmm.
[37:02] speaker_0: So anything that's happening on X, like I've had friends send me screenshots and being like, "Hey, your thing's going viral on Twitter." And I'm like, "Cool. Thanks for the screenshot."
[37:09] speaker_1: Yeah, yeah.
[37:10] speaker_0: And it's so true. Like, when people ask me, "Oh, what's your most recommended reading source for AI?" I'm like, "I have my answers, but the real answer is, like, my friends just text me." (laughs)
[37:17] speaker_1: Yeah. Yeah. People are like, people are like, "Oh, what podcast?" And I'm like, "Dude, you think podcasts are..."
[37:21] speaker_0: (laughs)
[37:21] speaker_1: I mean, I get it. It's the podcast, you know, but moment.
[37:24] speaker_0: We're making a podcast right now. (laughs)
[37:25] speaker_1: This- (laughs)
[37:27] speaker_0: (laughs)
[37:27] speaker_1: No, but like, it's like, you know, like, uh, I'd rather just text a friend and be like, "Yo, is this good? Yeah. All right, cool."
[37:34] speaker_0: You know, we've talked about a lot today. And one of the things that I really love and respect about you is how many layers deep you're thinking about AI. Because there's kind of this first level of, like, tools and tips and tricks. And that's, I know, a lot of the questions that you and I both get-
[37:48] speaker_1: Yeah.
[37:48] speaker_0: ... from people about AI are that.
[37:50] speaker_1: Yeah.
[37:50] speaker_0: It's like, what LinkedIn certification should I get-
[37:52] speaker_1: Yeah, yeah, yeah.
[37:53] speaker_0: ... uh, to prove that I'm AI, um, literate? And then there's the second level of, like, how to shift your job/station/position in the world by adapting and learning these tools on a deep level.
[38:03] speaker_1: Mm-hmm.
[38:03] speaker_0: And then I feel like there's the third level of, like, uprooting your preconcept- preconceived notions about your identity and value and meaning and how all of that shifts-
[38:11] speaker_1: Yeah.
[38:11] speaker_0: ... based on the first two layers. Um, so with all that being said, when you think about it at that third layer of depth, um, what are you optimizing for in your life right now? Is it money, learning, freedom, something else?
[38:31] speaker_1: I think it's, like, this idea of agency.
[38:34] speaker_0: Hmm.
[38:36] speaker_1: Um, and the way a lot of people talk about it now, it's, like, kind of a buzzword in tech. But I think of it as, like, how do you end up in a place where you have more options tomorrow than you... Like, you feel like you have more options tomorrow than you did yesterday. Um, and I think there's a lot of scarcity around AI right now. Like, the, the idea that, "Oh, it's taking my job," or, "It's replacing me." And it will replace a lot of jobs, for sure. But on the other end of this, it's also the greatest expansion of, like, cognition of all time. Now you have people that... Like, it would be a shame if I thought of myself only as a data analyst now knowing what I can do, right? So I once took two days, and I just had Midjourney generate a bu-... I was like, "I wanna... I have a song I never put out. Um, but it would be cool to put it out as a video, like a music video. But I don't have a label. I don't have a manager or anything." It's like, well, then AI is like, "Well, you got Midjourney. You got CapCut. Like, you could just, like... And you got Runway, so you can animate the images. Also, I can help you create the world by just..." So then, so you spend two, two days, and Midjourney had generated, like, 800 pictures, turned 400 into videos. And I put my song on it, and then I put it out there, and I was kind of just like, "Wow." Like, I'm not a music producer, and, like, I've never been, like, recognized in that industry.
[39:45] speaker_0: Yeah.
[39:45] speaker_1: I have no audience. I have no listeners, really, outside of my friends. But I can make this thing now. And it previously would have taken a budget and a team.
[39:52] speaker_0: Mm-hmm.
[39:53] speaker_1: So I get to expand my own definition of self. So that's, I think this is, like, agency. It's like now you get to do that thing. You don't have to ask for permission.
[40:01] speaker_0: Right.
[40:01] speaker_1: You can just put the right tools together, ask AI for some inspiration. And even all these hard questions about what, you know... Like, I think, like, you can ask AI. And it's not gonna judge you. And you can be like, "Okay, given these are my strengths, these are my weaknesses, these are the tools I have, here's what I wish I could be-
[40:18] speaker_0: Yeah.
[40:19] speaker_1: ... how do we get there?" And you will be very... Uh, it's, it's like a new form of manifestation.
[40:24] speaker_0: Hmm.
[40:24] speaker_1: Right?
[40:25] speaker_0: Do you like motivating people in this way?
[40:28] speaker_1: Yeah.
[40:29] speaker_0: Why?
[40:30] speaker_1: I think, like, uh, when I see someone six months later or a year later-
[40:34] speaker_0: Mm-hmm.
[40:35] speaker_1: ... like, and I see how much... When... I like seeing them with increased agency.
[40:40] speaker_0: But, but here's my thing is, like... So I just gave a talk at a big company, right?
[40:45] speaker_1: Yeah.
[40:45] speaker_0: And they were like, "Come into our team, get everyone hyped about AI."
[40:48] speaker_1: Yeah.
[40:48] speaker_0: "And we're gonna pay you this bucket of money to do it." And I'm like-
[40:50] speaker_1: Yeah.
[40:51] speaker_0: "... yeah, sign me up. Let's do it."
[40:52] speaker_1: Nice.
[40:52] speaker_0: And I get in the room.
[40:53] speaker_1: Yapper metagame.
[40:54] speaker_0: Sorry?
[40:55] speaker_1: Yapper metagame. (laughs)
[40:55] speaker_0: Yapper metagame. (laughs)
[40:57] speaker_1: (laughs)
[40:57] speaker_0: Exactly. Slap another zero on there. (laughs)
[41:01] speaker_1: (laughs) Exactly.
[41:02] speaker_0: So I get in the room, and I'm giving this talk. And it's really interesting to me, actually, because in my talk, I hit on a lot of the themes that you did.
[41:11] speaker_1: Yeah.
[41:11] speaker_0: The whole thing about, like, aim AI at your passion.
[41:14] speaker_1: Yeah.
[41:14] speaker_0: Like, go all in on your interest. This is the time to, like, reimagine your identity, think big. Like, I also talked about, like, AI is your motivational thought partner.
[41:21] speaker_1: Yeah.
[41:21] speaker_0: Like, when you're feeling down, it can be also the thing that hypes you back up again, like, everything in between.
[41:26] speaker_1: Yep.
[41:27] speaker_0: And at the end of it, the first question that I got was, "What LinkedIn certification should I be going for for AI?"
[41:36] speaker_1: Yeah.
[41:36] speaker_0: And it's like, I want to feel excited to motivate people about AI and everything that we've talked about in this conversation in the future, but I find myself having a really hard time honestly getting excited to talk about this subject with anyone other than people who have...... a perspective similar to you.
[41:53] speaker_1: Yeah, I know, I-
[41:53] speaker_0: Or, or who are, like, complete newbies, but have, like, a curiosity. And like, I'm, I'm totally fine talking to people-
[42:00] speaker_1: Yeah.
[42:00] speaker_0: ... who are, like, non-technical or have no idea what's going on. But they're open and they're curious and they're, like, down for that. What, what drives me up a wall is when people are, like, so rigidly fixed in the current paradigm of, like, work and school and labor, and, like, so pessimistic about it too. And they can only see what's directly right in front of them. We talked about, like, the one, two, three levels.
[42:20] speaker_1: Yeah.
[42:20] speaker_0: They, like, they're max at level one. It's, I find it, like, demoralizing and frankly speaking, like, I don't wanna do it for no amount of money.
[42:28] speaker_1: Yeah.
[42:28] speaker_0: It's just like, it's a waste of my time.
[42:30] speaker_1: I feel the exact same. Um, like, it's like ... Uh, I've, I've been in a room full of pessimists and I'm like ... And then they're like, "What do you think?" I'm like, "If I'm the only optimist you're gonna invite-"
[42:42] speaker_0: Mm-hmm.
[42:42] speaker_1: "... like, I'm not gonna come."
[42:43] speaker_0: Yeah.
[42:44] speaker_1: And on the flip side, it's like students will ask me, like, "Oh, how am I supposed to compete with people that have seven, eight, 10 years of experience in machine learning?" I was like, "Don't even go for machine learning. Like, maybe go for this new thing, AI engineer, whatever it is."
[42:58] speaker_0: Yeah.
[42:59] speaker_1: Like, it's, it's a different ... It's like VibeCoder. Like, I see a, a ... Moon Valley had a job posting for VibeCoder the other day, and I was like-
[43:04] speaker_0: Shut up.
[43:05] speaker_1: I was like, "Yo, it's th-" (laughs)
[43:06] speaker_0: That's so valid.
[43:07] speaker_1: Yeah, yeah. And it's like that, prompt engineering, I was like, "These are the skills." And there's no textbook. It's that, it's like you want a certification, there's no certification yet. And maybe people are making them, but-
[43:16] speaker_0: They're, they're shit.
[43:17] speaker_1: ... the best certification is, like, share what you made.
[43:20] speaker_0: Yeah.
[43:21] speaker_1: You know? Like-
[43:21] speaker_0: That part.
[43:22] speaker_1: Someone'll test it, someone'll be like, "Oh ..." Someone'll be like, "How do I ... Like, I wanna, I wanna ch- edit all these videos of LeBron together, but I wanna have a song that's, like, more of a, like, an AI song, but this is the quality." I'm like, "Maybe try Riffusion and Suno or Udio." Next day, he posts the video and it gets a million views. And it's like, LeBron AI remix, like, whatever. And I'm like, "Wow, see?" Like, I gave him the right nudge.
[43:44] speaker_0: Yeah.
[43:44] speaker_1: But he knew how to assemble the thing in his vision.
[43:47] speaker_0: Yeah, yeah.
[43:47] speaker_1: And then the next day, now he's like, "Well, I'm just gonna post a video every day until they win."
[43:51] speaker_0: Yeah.
[43:51] speaker_1: And I'm like, "That's-"
[43:52] speaker_0: And that can't be taught. I'm sorry, but it can't.
[43:54] speaker_1: Yeah.
[43:54] speaker_0: People either got the hustle and grind or they don't.
[43:57] speaker_1: Yeah.
[43:57] speaker_0: And that's, I'm realizing the more and more I, like, you know, opportunities are coming up and I have the opportun- ... I have the chance to say yes and no to what I wanna work on. My biggest thing that I'm filtering for is, like, how much are you actually down for this?
[44:08] speaker_1: Yeah.
[44:08] speaker_0: Is this ... Am I just gonna be speaking to a wall or a- is this message ready to be received?
[44:13] speaker_1: I, I think, like, a lot of the corporations are, um, not gonna make it.
[44:18] speaker_0: Yeah.
[44:18] speaker_1: And it's because of this mindset. And a lot of times, it's not, it's not even the employee's fault. Like, they're not in an environment where they're allowed to even explore and experiment. And that's why, like, what your job is to get them to experiment outside of their work a little bit.
[44:29] speaker_0: Yeah.
[44:29] speaker_1: And not everyone has time, right? Some have family, kids, uh, you know. Like, there's so many reasons why you won't necessarily have time or energy. But you should ... What happens to me is, like, people will come to me because they're afraid their company isn't gonna make it.
[44:41] speaker_0: Yeah.
[44:42] speaker_1: And then they're like, "I can't bank on this team being my career. And Parth has already taken this, like, leap into the unknown. And maybe he has some answers as to, like, where the, like, the, where the world is actually going. And if I prepare for that, then it's okay if I get laid off." Like, so a lot of times, it's like you need that push. Like, in my case, I needed a l- I, I got laid off and then it pushed me into AI.
[45:02] speaker_0: Yeah.
[45:02] speaker_1: But I think that hap- that's happening to a lot of people. When people lose their jobs or they, they feel unsure, that usually is a moment for them to, like, ex- hopefully it's a moment for them to explore what's next.
[45:11] speaker_0: Yeah.
[45:12] speaker_1: And usually people realize what's next is much more fulfilling than what was before. The flip side is that I don't think you can just tell everyone to adapt.
[45:22] speaker_0: Mm-hmm.
[45:22] speaker_1: Um, like, I don't think everyone can. Like, it's n- n- you're not in the right, like, head space or, like, like, maybe you have, like, a family and kids and it's just you don't have that much time to put into the, um, the, like, the, like, learning curve yourself. But in that case, like, it's like every group chat should have one AI-obsessed person. And then that person is just gonna bring the rest of the group up into the future with them. And I think that's, like, it's, it's like you gotta lean on your network, um, and you're just not alone. And I think a lot of people think that they're alone, but actually you're not actually alone. And then be a little bit more intentional about, like, cultivating a network that's more resistant to this than, say, your company.
[45:58] speaker_0: Mm-hmm.
[45:59] speaker_1: Right? Like, if you're, if you're only using this for your job and then you're only using it to figure out how to make money for your employer, you will probably miss, like, all of the beautiful things that you could also be doing that could just help you.
[46:10] speaker_0: Yeah.
[46:10] speaker_1: You know?
[46:11] speaker_0: What's your dream?
[46:13] speaker_1: My dream?
[46:14] speaker_0: Yeah. Where do you, like ... If you kinda close your eyes and envision Parth in, like, the deep future.
[46:19] speaker_1: I used to dream about this moment minus the AI. And I feel like I got here pretty early in my life.
[46:25] speaker_0: What i- what is it about here that's just-
[46:26] speaker_1: Well, just, like, you know, living in LA, like, having a decent tech job, like, kinda just, like, exploring some of my side interests as well. But AI has accelerated and made a lot of that possible. Like, for example, a year ago, I was, I was at an event in, in Hollywood. They were talking about, like, AI film and they were talking about Sora, and it wasn't, like ... It was a very pessimistic conversation. And then they asked me, it was like, "How do you feel about this?" And I was like, "Am I the only optimist?" Like, I, I feel like this is the beginning of my ... It's like, yeah, maybe you're kind of ... You, it challenges you in how you make films, whatever. Sure, yes. Uh, s- I'm not trying to be disrespectful.
[47:02] speaker_0: (laughs) Yeah, not whatever. Sorry.
[47:03] speaker_1: But I am not, I am not even in the industry and now I get to experience it for the first time-
[47:08] speaker_0: Mm-hmm.
[47:08] speaker_1: ... as, like, a creator.
[47:09] speaker_0: Yeah.
[47:09] speaker_1: That was never in, in the cards for me.
[47:11] speaker_0: Yeah.
[47:11] speaker_1: And I'm like, "Oh, wow. I could, you know, I could probably make a short film with me in it, and then I could deep fake myself into the film, play myself in a bunch of different universes." Reality destabilizing, potentially. But also, now I get to actually make the movie, deep fake myself into it, give myself the Jeff Bridges in Tron treatment.
[47:27] speaker_0: Mm-hmm.
[47:27] speaker_1: And I was memeing that a year ago. Now it's real. So, a lot of that stuff that I was, like, I was like, "Oh, it'd be cool if one day I could do X." It happens, like, in eight months.
[47:36] speaker_0: Mm-hmm.
[47:36] speaker_1: And so it's like, you ask me what my dream is, and it's like y- you have to dream way more and way bigger.
[47:42] speaker_0: Yeah.
[47:42] speaker_1: Because a lot of this stuff, the only thing, it's just happening faster. And a lot of that stuff you thought was, it would be nice? No, it's actually already happening.
[47:51] speaker_0: Yeah.
[47:51] speaker_1: And that's why I say play with the tech, because instead of, like, putting off your dream, your side project, whatever, it's like, no, actually Replit might one shot half of it in eight seconds.
[47:59] speaker_0: (laughs)
[48:00] speaker_1: (laughs) Like ... (laughs)
[48:01] speaker_0: Yeah.
[48:01] speaker_1: You know?
[48:01] speaker_0: Actually, though.
[48:02] speaker_1: And you need to internalize that, because then you're gonna realize, oh, clearly I need to be dreaming bigger.
[48:06] speaker_0: Dude, I'm telling you, this is, people need to meditate-
[48:10] speaker_1: Yeah.
[48:10] speaker_0: ... on this technology. We have, we have officially crossed a barrier. But most people don't even realize that.
[48:15] speaker_1: Yeah.
[48:16] speaker_0: But we have already passed the barrier that separates technology from, like, spirituality.
[48:20] speaker_1: Yeah.
[48:21] speaker_0: It's gone. Those planes have merged.
[48:23] speaker_1: Yeah.
[48:23] speaker_0: We're already there, and it's just the people who have literally sat down and entered the prompt into ChatGPT of like, "This is my dream life that I wanna have. What do I need to do to get there?"
[48:30] speaker_1: Yeah.
[48:31] speaker_0: Who are, like, starting to chip away at that and experience it. I mean, I feel like I'm living it firsthand too.
[48:35] speaker_1: Yeah.
[48:35] speaker_0: Like, literally a year ago, less than, like (laughs) I was having a fucking career crisis. Like, rock bottom of rock bottom being like, "I don't know what I'm doing with my life. I don't know where I wanna go with this, like, what's happening?" Fast-forward, like, however many months from now and it's like half a million followers across platforms-
[48:51] speaker_1: Yeah.
[48:51] speaker_0: ... building ChatGPT, and it's like-
[48:53] speaker_1: Yeah.
[48:53] speaker_0: ... a lot of that is 'cause of AI.
[48:54] speaker_1: And you, exactly. I mean, you, and you've been a, you're poster child of this. I saw, I see this and I'm like, "Holy crap, like, this is ..." I mean, but, I mean, it's, it's going to happen for a lot of people. It, they'll take, they'll take their time. But it will happen to a lot of people over the next couple of years.
[49:08] speaker_0: Yeah.
[49:08] speaker_1: And I, I'm excited for it. We're early, we're very early.
[49:10] speaker_0: Hell yeah.
[49:11] speaker_1: Right.
[49:11] speaker_0: So Parth, where can people find you?
[49:12] speaker_1: I'm working on a website. Well, AI is working on the website, it's not quite finished.
[49:16] speaker_0: (laughs)
[49:16] speaker_1: Uh, parth.club. Also you can find me on LinkedIn, Parth fire emoji. And then, uh, Instagram, Substack, it's Parth Intelligence.
[49:24] speaker_0: Thank you so much for coming on. Really appreciate your time.
[49:26] speaker_1: Thanks, Cat.