Cursor’s First Backer: How to Spot Genius Before the Evidence | Ali Partovi from NEO

The real mistake I made was I was trying to deceive Steve Jobs. I was intentionally trying to trick him.

You have spent decades building a reputation as someone who finds people before the evidence was obvious. What are the things you're actually looking for?

The most important trait, I'd say, I— the word I use is magnetism.

I know for a while you would always tell young people they need to learn how to code. You've invested in many startups that make coding a lot easier.

Or not even necessary. It was quite hard to hire for Cursor in the first Like 18 months.

Why?

Everyone said, "Oh, this is just like GitHub Copilot."

Why don't you tell the story,

Gary, and all the other multiple partners of YC sort of chimed in and kind of ganged up against me. Since then, I have hugged it out with Gary.

I'm so excited to have Ali Partovi here, fellow Persian, founder of Neo, and also co-founder of multiple companies. I like Link Exchange. Sway, Code.org, and his roster of investments is insane. It's Facebook, Uber, Airbnb, more recently Kalshi and Cursor. And after diving deep into who you are over the last week, it's clear that you are one of the best in identifying talent before it becomes talent, and then creating an environment that allows that talent to succeed. So, very excited to have you here.

Thank you. What's up, everybody? Great to be here.

Great to have you. So you have spent decades building a reputation as someone who finds people before the evidence was obvious. But I'm sure this was a muscle that had to be developed over time. So I'm curious, like, when you first started thinking of yourself as someone who could identify talent, what is the thing that has changed the most? Maybe a mental model that you don't think is very useful anymore?

Interesting. So it is true, this is something I've been intentional about. I remember when I was in college, my mom said to me, you're surrounded by some of the smartest people ever, and besides whatever you're learning in classes, build out your network. And there are people I'm still friends with from college days. Definitely the moment though that I remember when I was a 23-year-old founder, I remember realizing there were some of my role models who had this magic power that I didn't have of evaluating people in a very short period of time. And Mike Moritz from Sequoia specifically was our lead investor. And when we were trying to interview and hire people, in particular people older than ourselves, I was quite intimidated. I was 23 or 24. this impressive old person who was 28, you know, or 29, it was like, oh my gosh. Old? Yeah, exactly. Their resume seemed so— I was just intimidated, and they just all seemed, like, impressive and strong, and I didn't know how to differentiate. And myself and my co-founders would spend hours with these candidates, then Mike Moritz would spend, like, 10 minutes with them or 15 minutes, and he would have a more opinionated position. And whether he was right or wrong, I was just like, this is impressive that he feels like he's gotten some signal where I had not.

Enjoying the conversation? You got a question? You can ask Ali's digital mind. Just check the link in the comments or description.

And there were other folks who clearly had this, what seemed to me like a superpower. And I decided I want to have that superpower because it's the most valuable like meta skill you could have. I think what has changed for me, I don't think I would say that there's any criterion that I've dropped, but what has changed is I think I've tried in the last few years to put words to what I'm looking for rather than just like a nebulous vibe. Right. You know, 'cause I, you know, I'd say we all essentially make, when we make gut decisions or instinctive decisions, what that means is a form of pattern matching. And I say I've been good at that for my whole career. But now that I manage a team and it's more of an institutional effort, I need to have other people be able to do it and for them to learn from me or me to learn from them. And that means you need to be able to put words on what specific trait or what it was about somebody. It

can't just be instinct.

Yeah, you can't just be like, yeah, the vibe was perfect. I mean, that works. It's definitely part of it. But if you can communicate what it was about somebody, then the other people can also look for that in different ways. And so, yeah, so that's definitely for me been less about changing my criteria, but rather trying to communicate and scale.

Have you been able to compress the multiple hours to evaluate someone to 10 minutes?

And how?

Please tell me, please.

No. So it's, There are— I think what I'd say is I often make an assessment within a pretty short period of time, but then I'm still often kind of looking for sometimes maybe changing my opinion. But there's been studies of— it's remarkable. This is not me, but just studies of people like interviewing for jobs or so on. A surprising amount of people decide within their like first impression, first meeting.

And are they right? Does the study show whether they're right or wrong?

That I'm not sure. I'm going to get in trouble trying to quote this because it was something I heard someone else talking about, so I haven't read the primary results. But I guess I'd say this is— yeah, obviously if you're accurate, great, but almost as important is just to take an opinion And to have confidence in whatever voice inside you, for starters, to even tap into the voice inside you. And what I mean by this is, I think for a lot of us, when we meet folks, we're not paying attention to register the positives or negatives, to evaluate them, let alone to really zero in and say, okay, this was a real negative. I'm going to kind of— I'm not gonna reject them for this, or this was a huge positive. I'm gonna— I think part of what's going on in polite society, you're not supposed to evaluate everyone you meet. But

then you're not present. You're not present.

Yeah, you wanna just— and also you wanna be— you wanna connect, and if the person you're talking to maybe says something that doesn't seem as smart, it's nice to kind of overlook that and keep the conversation going and not be judging them. No one wants to have a friend who's always judging them, right? So when it comes to social life, this is probably a skill you want to turn off. But when it comes to professional life, being able to really detect, wait, this person did this one thing that felt off. I'm going to zero in on that or double-click on that and just pay attention and not forget it. A lot of people, if they hire someone, Like a year later, if it doesn't work out, they'll often feel like, you know, whatever the trait was that didn't work out, there was a hint of it.

You knew about it on day one.

Yeah, there was a hint in the very first interview.

You always know about it.

Yeah. And so it's more ability to just pay attention to that voice and have confidence in it rather than suppressing it and, you know, and kind of thinking, well, maybe I'm wrong. Yeah, but I guess I'd say I'm still a student of how to evaluate people. I'm still constantly learning and making mistakes and so on.

Yeah. What is an area where you have doubts? Are there any certain kinds of people that you're worried about missing?

Worried about missing? I think I'd say actually I'd go the other direction, which is people who are too Like—

Polished?

I was debating whether to say polished or just people who are too good at creating FOMO or like—

Or distribution is king.

Too good at sales. I worry, what if they're being untruthful? You know, like essentially being good at hyping is a skill. It's a very important and valuable skill. And there are some people who are so good at it. But it's like, at some point it crosses over into this territory of being so eager to hype that you detach from— What about reality distortion?

From the truth, yeah, exactly. What if they actually believe it?

But this is— I mean, there are occasionally people who make their hype turn into reality, of course. But there is a real danger in getting too excited about someone for that trait and then finding out later that They were just, you know, like making it all up without enough connection to reality.

Yeah. So for people who have no evidence of success, what are the things you're actually looking for in 10 minutes?

Yeah. So, so I mean, a really unique thing about Neo is giving technical assessments to people. We don't do this for all of the founders we bet on, but for college students, we give them coding tests and multiple levels of assessments as if we were hiring them for a job. And we refer so many people to jobs at top companies like Cognition today or OpenAI a few years ago or Figma a few years farther back. We've been a recruiting channel for so many tech companies because our assessment process is so rigorous that it kind of substitutes for their phone screens and first-round interviews and so on. So when I was starting Neo, I was giving these paper and pen coding tests to students. Mm-hm. This was even before raising the fund. I was going to colleges and meeting students and trying to find the technically strongest ones. But then in terms of other things I look for, I'd say that assuming a level of technical proficiency, the most important trait I'd say, the word I use is magnetism. And essentially the ability to draw other people towards yourself. And the question I'm asking is if this person starts something, how many of their smart friends would want to join it? Whether that's a company or a club or a group or just anything, how able are they to convince others to join with whatever they're doing? So that's one. I look for willingness to take risks, willingness to kind of go against the herd, go the path that other people are afraid to try. And you can look for evidence of that. So it's partly like, partly their personality, but partly just—

What they've done.

What they've done. Even going back through high school or middle school or so on. And then third, This you can overdo, but a little element of mischief, of willingness to not to like break laws, but to at least question the rules and to like see the possibilities. What if this rule was different? And to think beyond outside the box, essentially. And again, evidence of this can be at a very young age. Some of the, you know, some of the very successful people today have awesome stories of Yeah, of things. Alfred Lin is one of my favorites. He's a very dear friend of mine, and he got into trouble all the time, starting from kindergarten all the way through. I mean, pretty much anytime there was a rule, it seems like he tried to figure out a way to break it as a little kid. And yeah, again, this can go too far, but there are entrepreneurs who simply would not have Or their entire businesses that would not have happened unless the entrepreneur had thought that way. You know, Kalshi being a great example where—

6 years of just pushing.

Yeah. And when they— I mean, they started with this kind of belief that prediction markets, the way they were structuring it, were compliant with existing laws. And we just need to find a lawyer to help us submit the application the right way. They called like 40 law firms and all of them rejected them before they found one who said yes. And, you know, this is a story about 2 things actually. One is the grit of, I would've given up after like 10 rejections. Yeah. But also the, like this willingness to say, I don't care what all these people are saying, it hasn't been done, or this doesn't sound right. And I still believe that it should be possible, and we'll keep trying for it.

So you've invested in legendary companies in both the 2010s era and now the 2020s era. You have Cursor, Kalshi. Is there anything different about the founders that can succeed in the 2020s that you've noticed through working with Michael, Tarek, Luana?

I'm trying to think. I'd probably say not really. Not really. It's all the same. It's the same type of— and I mean, for me, it's further back. I mean, I sold my first company in 1998, and Zappos, I was one of the first angel investors in. That was, I think, 1999. If I think left, For example, the first time I met Mark Zuckerberg or the Dropbox founders, the— I mean, I guess I'd say the common thread of people I invest in basically are nerds, people who are really strong technically more than anything else. And that's not necessarily, you know, that's not necessarily all it takes to be a successful founder. Sure. And there are also many successful founders who are not technical. But for me, that's probably been the common through line. And some of the traits I mentioned, I would say I could go back and kind of point to the same traits in those founders from 20 years ago.

Yeah. On looking for technical excellence, I know for a while you would always tell young people they need to learn how to code. You invested in many startups that make coding a lot easier.

Or not even necessary.

Not even necessary. So is there a, Is there a new tactical thing you advise younger generations? Do they still need to be able to do the napkin coding exercise, or is it potentially time for a new kind of talented person who just doesn't necessarily need to do that?

Well, I would separate these into 2 questions in terms of what's a good subject for young people to learn in school versus how we might assess, how I assess people. On the second one, I am constantly asking myself, oh shoot, are these coding tests that we relied on so successfully for years irrelevant now? Both in terms of, are we testing for the right relevant skill? And also, there are people who are brilliant but who have no practice coding by hand at all because the AI does it all. And so they would fail a test. despite actually having the— I guess I'd say the underlying skill I care about is not can they code in Python, but how smart are they? How good at problem solving? A range of things that I think computer science is a good proxy for. So I don't know the answer there, but I would definitely say that we are actively questioning ourselves about this. Each year we revisit it and have a lot of debates. In the Neo Scholar application process, we have dropped certain assessments and added others because of it. With respect to school, I think every student should study computer science. But I'm not referring to college, I'm referring to like K through 6.

Right. As fundamental as math.

I think it is, yeah, super fundamental and more fun than a lot of other stuff that's taught in, you know, K through 12. It's also supplements. Like, if you're in the middle of learning, you know, certain math topics, you might feel like, how will I ever use this? But then if you're writing a computer program, you'd be like, oh wait, I can use the thing I just learned in math for this other thing. Right. Like, if you were creating something with a, you know, a ball flying through the air in a parabolic curve, you could use stuff related to parabolas or so on and so forth. But I'd say more importantly, a lot of what we teach in school to children is not for any subsequent use. Everyone learns the quadratic equation. How many people are using that day to day? And of course, AI could solve—

I still have it memorized. I don't know why.

Right. We all learned that. Is it useful for your career? Probably not. But a lot of stuff you learn is because it exercises your brain, teaches you problem solving, teaches you, you know, the confidence to tackle tough problems. And I would say coding does a better job at that than almost any other discipline.

It's also creative in a way.

That's the other thing. It's super fun. No one gets enjoyment out of, like, factoring polynomials for too long. Well, maybe they do, but—

Maybe it's for

us. But, like, building something that then you can use or play or give to your friends is inherently more fun.

To talk a little bit about Neo, once you've bet on people, you have an environment designed to make them better. And I'm curious your philosophy on whether how much of greatness is innate, Versus how much can be kind of instilled in someone going through a program like Neo?

Well, I mean, in terms of innate, I'm definitely between— in the nature versus nurture argument, I believe a lot in nurture. I'm an identical twin, so I have like my own private experiment with my brother Hadi, and can kind of— I guess I've just generally seen situations where he was way better than me. but I managed to get better and become as good or so on. Obviously, that's not the scientific huge population size, but as far as my beliefs, I very much believe in people's ability to grow. That being said, the older we get, like, certain things become harder to change. And so part of why I love working with people at the college age is because it's like this great mix where there's some skills that maybe feel innate or that are already really strong and are there, but there's still so much room to change somebody's trajectory. And some of it is not so much skills, but just the repercussions of decisions they make. Mm-hmm. So, you know, if there's a junior in college and they're choosing between— and I'm thinking of a specific person, Luke Melaschiarzi, when he was actually a sophomore in college, was choosing between Citadel and some other quant options and interning at a Silicon Valley AI startup called Forethought. And I convinced him to join Forethought. That one choice potentially changed the trajectory of his career. You know, quant is very, very almost addictive because, you know, like they pay very well. It's ultra competitive. And if you're in there and you feel like an athlete on a team and just the reward system, the feedback loops can be addicting. And a lot of people never leave that direction. And I remember checking in with Luke that summer and he was saying, oh my gosh, I joined this company when it was like 18 people, 2 months ago. Now it's already 35. So I'm one of the old guys as a college student. And he was involved not only in the engineering, but also customer support. He was on the Slack channel talking to customers. He was being called into sales calls. He was doing interviews for new candidates.

They exposed to so much more.

Yeah. I mean, it completely transformed his sense of what he's able to do. He's now, as you might know, he's one of the founding engineers at Cursor, head of ML essentially.

Yep.

Very, very successful. And it's hard to say how much that particular thing impacted his trajectory. I mean, also that following January, sorry, I'm getting my timeline, that following January after that internship, He came to a week-long event we hosted called Startup Camp that Michael Truel also attended. They met each other through that and spent a week at the same place. That might have also been a factor. I'd say it's hard for me to say, oh, we're responsible for all of this, but we do a lot of different things just to increase the probability of somebody joining a startup or starting a startup. Or making one of these connections that sticks.

Yeah, I enjoyed reading your blog posts on how much time you spent with Michael and the early Cursor team. I'm curious, having built an environment that allows people to do their best work, what have you learned about creating that kind of environment? And that can apply to both the Neo Scholar program, but also in companies. How do you think about letting people achieve their best work? And what it takes.

Yeah. I mean, first of all, with respect to Cursor, that has been such a humbling experience for me. And I did spend a lot of time with Michael and Aman and Ben when they were still in college. So before it even incorporated, but I'm pretty sure that I didn't actually add a whole lot of value. I had a biweekly Zoom session with them, when they were just brainstorming startup ideas. And I'd say perhaps more than anything, it was just keeping them on track and keeping them motivated to have something to share with me. And I'd say those 3 are just humbling intellects, each in their own regard. And unfortunately, Ben left the team before Cursor was really formed and funded. But I'm not sure I can claim that much for them in terms of the business trajectory, 'cause they also started out in this CAD space before then shifting to coding for AI. And if anything, I was probably more encouraging to why not stick with the CAD thing? Not because that was why I invested, but simply I didn't want them to feel discouraged after the first few months didn't hit it or something.

Right.

That being— having said all these disclaimers, for sure, the thing I feel very confident that influenced them was seeing how high a priority I put on excellence and our whole institution. You know, Neo essentially tries to bring together the, you know, smartest of the smartest people across different, you know, different universities or companies or so on. And for them seeing how just insanely obsessive we are, and I am, about evaluating people.

It trickles down to probably everything else.

Yeah. And also the amount of effort and attention I paid towards hiring, including helping them with hiring. You know, like one of their first interns was, you know, someone they were— another Neo Scholar named Walden Yan, who I think they met through us. But whether or not they met through us, I definitely helped convince Walden to join them as an intern. There were multiple of their early people they were trying to hire that I would get on the phone and spend like an hour or longer trying to talk them through convincing them to join or just helping, you know, sending out emails to source candidates for them. It was quite hard to hire for Cursor in the first like 18 months.

Why?

everyone said, oh, this is just like GitHub Copilot except not successful yet. So it was like a lot of their peers knew them and were like, oh yeah, they're really smart. We know them. And why are they doing this? And it's funny because they were using it also, but they weren't— But they still didn't believe? Well, it was that I think there's this tendency, especially for someone who's relatively young to apply a lot of strategic thinking and thinking, oh, how will they strategically beat GitHub Copilot, which was the dominant option in that space? And my view was, I don't need to see the full strategy. These are freaking smart people.

I'm the people.

Just join them and see it happen. It's less about seeing what is What strategic choice will they make? But just a team of super smart young people who are pulling all-nighters and hungry and just obsessed will beat a larger corporate team that's been acquired by another corporation and just may not have the same hunger and obsession anymore.

Yeah. So you've spent last decade's focus on what identifies talent, and now you're starting a podcast called Confessions.

Confession.

Confession, which is exploring the journeys of people after they are successful. And so I'm curious, after speaking to some incredible people like Dara and some others, has learning about their confession and what broke them changed how you think about what might have made them exceptional to begin with?

So just to be clear, Confession is inviting tech leaders, founders, or CEOs to share a painful lesson from their past?

Yes.

I'm generally interviewing people who are already successful. Yes. Mostly because they're willing to talk about their, you know—

Yeah, it's hard to talk about in the moment.

Hard to talk about it in the moment. Yeah. But Luana from Kalshi is one of the people I've interviewed. I'm interviewing Satya Nadella tomorrow, actually.

Okay.

So it ranges from, people, you know, acting founders of relatively new startups to some of the most successful people. And I mean, the common thing there is trying to get, I guess, trying to show the human side behind someone that might seem like they have a perfect exterior and has all their shit together, to kind of dig back to some moment when they hit rock bottom, or when they messed up and learned something. And just to, I guess, to demystify a little bit and dispel whatever aura of perfection they might have. The goal is really to make the next generation feel less intimidated and to feel a little bit more connected to whoever their role models might be. I forgot what your question was, sorry.

Well, I guess, Hearing the confession, is there— does it change how you might have thought what made them exceptional to begin with? I guess it relates back to grit, how people deal with suffering.

Yeah. So each, I'd say each story has been pretty unique. I mean, I have many, many failure stories of my own. Oh yeah.

I'm going to ask about a couple of them.

Yeah. And I've made a habit of trying to post them publicly myself.

They're great threads.

Thank you. It's definitely my comfort zone, is talking about my own fuckups. So, yeah, endless supply of material. But, yeah, I guess with others, I don't know if I would say that it, like, changed my view of what made somebody great. It's mostly actually just adding dimension. Yeah. Or finding, like, something to empathize with that I would never have guessed. And I mean, there's just such a wide range, including just things that weren't about their career, but just show something about their past. More human side. I'm remembering this crazy story when I was interviewing— interviewed Shuo Wang, co-founder of Deal, and expected to—

She's coming on the podcast soon.

Oh, cool. Yeah. Yeah. So she grew up in China, went to MIT, had a different startup, then Deal. And I was hoping for some stories from Deal or maybe the previous startup. But when I was asking her about, so what did you study at MIT? She mentioned it and then she did grad school and she spent all of grad school on this kind of mechanical engineering work focused specifically on, like, robotics, like trying to get humanoid arm type robotic research. So I asked her, like, you know, why that particular field? And she said, oh, well, because in my, like, first few months of college, I was run over by a bus. And I was like, I'm sorry, what? And apparently, so, like, a campus shuttle, I guess, ran over her arm. Oh my God. And she almost needed to be amputated, but survived and spent the next, like, I don't know, 6 years working on robotics specifically in case she had needed to have a prosthetic arm. She was like, this is worthy of working on. It was just a fascinating little thing about the type of person she is that totally caught me by surprise. It was not what I was expecting to hear. Yeah. So yeah, so I've been uncovering all these really interesting vignettes about people.

One of our questions in our interview process is, what is the most you've ever suffered for that reason? And I've been stalking your Twitter and finding some of your own painful moments that are very enlightening, but I notice a common theme across some of them, and it's that the same trait can either make you great or destroy you. I'll talk about 3 stories that you tweeted about where this shows up. The first is when you almost sold your company iLike to Apple.

Yeah.

And the thing that led to that deal falling apart is you kind of overplayed your hand. You wanted a higher acquisition price. And I thought that your lesson from that was to kind of don't overplay your cards and to maybe be more honest in the moment. But at the same time, Founders are often rewarded for being bold, being overly confident. It's usually a direct correlation to how much money you can raise.

So let me— I should clarify, because the real mistake I made was I was trying to deceive Steve Jobs. So it was not just being bold or strong, but I was intentionally trying to trick him. I wouldn't go as far as saying that I lied, but he could tell this person is trying to bullshit me. And I think, I'm not sure what his demons are, or were, I mean, but I think he felt disrespected by that. And so it wasn't just overplaying my hand, it was that.

Pure deception?

It flipped. I mean, I remember the moment, and I remember the sinking feeling like, oh shoot, I think I messed up. And it was, for those who haven't read my story, you know, so he had made You know, he offered us essentially $50 million. And I thought that was really depressingly low because that was the same price of our valuation at our seed financing, like, many years earlier. And so I said, Steve, I think we're worth 3 times that, which was okay. I'm asking for triple, but asking, you know, I think. Then I thought maybe I should take it a little further. And I said, Steve, actually, I know we're worth 3 times that. And I think I changed my body language in the process to try to really make him feel like, you know, he had no choice or, you know, we had another option. So this switch from, I think we're worth 3 times, to, I know we're worth 3 times, he zeroed in on this and he said, are you trying to imply you have another offer? Sounds terrifying. It was terrifying because he said that. And I started like, I don't even remember. I kind of blacked out exactly what I did. I just remember wanting to shrink and disappear. But in that moment, it was, yes, it was overplaying my hand, but it was with the intent to deceive him. And he detected that. And I'm, I guess I'd say I'm zeroing in, because if I don't, then it feels like I'm copping out on what I actually had done wrong there. Yeah. If I could go back, which I have replayed in my head a million times, I could have said, we won't accept less than 3 times that. I could be quite strong, Without saying some implication of, you know, having— to say, I know we're worth is, you know, it like, and especially with the change of posture that I used, really was trying to make him think we had another offer.

It was like the FOMO that you kind of talked about.

Yes, yes, exactly.

You did the thing. My follow-up question was going to be how that differs from the story of you investing in Cursor and them coming with a higher valuation than you thought they were worth. Yeah. But I'm assuming Michael did not come in with such certainty, or do founders raising from VCs get a different dance?

Yes. I'm trying to remember. So, I mean, part of why, I guess, what we were very lucky about was this was right kind of at the trough of the correction in 2022 when, for a period of time, everyone thought startups were not going to make it anymore. And So valuations across the board dipped. And that same team fundraising now would've had a much higher valuation and many, many more VCs chasing them than us. But yeah, I don't want to quote the exact numbers because it might be their confidential information that I wouldn't want to disclose. But given where the market was, The number they quoted, my team and I thought about that and we're like, this feels like 30% above market. It was way below where market is today, just to be clear. But it was an assessment of, I mean, we absolutely love them, adore them, want to be their first backers. But there is this brief moment where we're on opposite sides of the table of what is a fair deal. Yeah. We would not want to pay way below market, but we also don't want to pay way above market. And so we were tempted, like, thinking to offer something a bit lower. But yeah, I remember on the call, I, like, reacted saying we were thinking, you know, something lower. And just the directness with which they flipped to saying, well, if we can't work together, then blah, blah, blah. It wasn't even like they were negotiating. They were just like—

They were just like, onto the next.

Yeah. They, yeah, it definitely made—

Feels more honest. Yeah.

Yeah. They weren't saying some, like, different higher number, but they essentially in a very kind of polite way let it be known that we'd still be friends, but that they wouldn't accept less. We did get them to come down a little bit, but not very much, something like 10%.

Yep. Story number 2.

Yeah.

Jerry Yang. So you almost sold your company to Yahoo, and you told Jerry Yang that Paul Graham's startup was the better team, which led him to acquiring Paul Graham's company, Viacom, instead of you.

And this is before Paul Graham was, you know—

And this is before Paul Graham became Paul Graham. I know how important integrity and honesty is to you in your career. So I'm I'm curious, the lesson from this maybe is like, don't overdisclose, but like, what's the difference between like trying to be high integrity and like high honesty without potentially leading to a situation that you found yourself in with Jerry Yang?

Yeah, well, first of all, I'll say between the, the 2 stories, the Steve Jobs story filled me with shame for a very long time.

I mean, he's Steve Jobs.

It was— no, it was because what I did was dishonest. And because it really made me feel like that's, like, I'm not proud of that, that's not who I wanna be. And particularly made me feel like, what if I'm being dishonest in other interactions and trying to hype my startup and other people just don't say it to my face the way Steve did? So it really deeply, like, shook me as to, like, this is not who I take pride in being. Yeah. Compared to that, the other story of being too honest does not give me any shame at all. I would be proud to be known as the person who was too honest, even if it wasn't in his own good. And so, yeah, big value judgment. You know, like, I wouldn't say there's, like, a huge lesson of, like, it's wrong to be honest by any standpoint. I think probably the The more— the common mistake between the two was just being impatient. In both cases was a person— I mean, the first time I was 24, the next one I think I was in my 30s, but in both cases I was so overeager, a little bit desperate, and like, you know, basically not thinking before I speak. And yeah, I mean, the Steve Jobs incident, the far better thing He had said to negotiate with Eddie, his, you know, who was like his lieutenant and was in the room. And I could have negotiated with Eddie rather than with Steve, which would have been better altogether. But with Jerry, I could have waited until after our deal closes to tell him—

Of course you could have.

To tell him how awesome ViaWeb and Paul Graham was. You know, there was no rush. I loved this company, ViaWeb. We had been trying to acquire ViaWeb ourselves, and then it was on hold because Yahoo was trying to acquire us. And so it was this kind of funny thing. It's like we didn't want to lose this other deal, and I was filled with zeal of, like, I've discovered this unbelievable jewel of a team in them. This was the same timeframe. I was in my, I think, 25 by then. I had decided my new superpower would be finding brilliant people. Here you come. And Hans and Trevor, I mean, that team was, to me, the apex of, I have found utter genius here, and they're smarter than we are. And I actually had this zeal of sharing it with others, such as Jerry. I will also say, I don't think Jerry— that's not like— I don't think Jerry thought, oh, let's therefore not acquire LinkExchange. It might have been part of it, but, you know, deals fall through for all sorts of reasons.

Deal isn't done until money's in the bank.

Yes, exactly.

All right. Story 3 of 3.

Yeah.

Jamie Levine.

Jamie Iovine. Yeah.

Iovine. Oh my God. Jimmy Iovine. He threatened the existence of your company.

Yeah.

And why don't you tell the story and I'll ask you

a question? Yeah. I mean, so this was the same company that later Steve Jobs did not acquire. It was called iLike. It was a music service and it was music discovery. So it was not Like, it was not for music consumption. So the distinction being, we didn't offer full song streaming, we didn't offer purchasing songs. This was in the pre-Spotify days where you, to acquire a song, you could go to the Apple iTunes Store and buy the download. But our service let you sample songs and see what your friends are listening to, or share playlists with only 30-second clips. And then if you wanted to purchase, then you could click through, and we became the top affiliate, like the top source of sales to the Apple Store, which we thought was something that the record industry would love us for. We were like the top driver of legal music consumption at a time when there were a lot of you know copyright violations out there, and. Again, this case of impatience. I had somehow been invited to meet with Jimmy, who is like one of the most important record moguls in the industry, and I was like excited to go do it instead of doing much research or asking some of my advisors, like, what's this guy's reputation, or what preparation or so on should I have? But I had this whole deck prepared for how we could work together to increase even more music sales and ways we could help their artists and so on. Jimmy had a completely different plan, which was to basically to try to extort equity out of us. And I entered the room. It felt a little bit like a, like an ambush, because instead of a one-on-one, it was me and Jimmy and a whole room full of henchmen. And pretty much the— I think it was within the first few sentences, Jimmy said, you've built an amazing business off our backs, and we could cause a lot of harm to you if we pulled out our licenses. And we wouldn't want to do that. You know, we'd much rather if you gave, you know, half the equity in your company to us, and we could build, you know, let you continue doing what you're doing. And I was like, I don't think he knows like half. What are you talking? I was just really confused. I explained that we're not violating any copyrights. We've paid for the licenses to our thirty-second clips, so everything is good. And he said, "I don't think you heard what I said." And then he repeated the same sentence almost word for word, and I realized he's not accusing me of anything. He's just merely informing me that. He could harm us unless we gave him equity. And I realized the only play I had was to leave. So I said, this meeting's over, and left. Then we were desperately looking for, where do we have some leverage? And we had started a relationship with one of Universal Music Group's top artists, which was U2. Kind of like the top band in the world at the time. And the lead singer, Bono, had a particular interest in technology and wanted to reach their fans through us. This was back in the days before, long before Instagram and even before Facebook or Twitter allowed artists to reach their fans. We offered U2 a way to release like new music directly to their fans. And Bono essentially wanted to be one of the first to do so, which was pretty cool. And essentially a new song with like a preamble that he spoke in video format, they approved to release through our platform. And this would be— it's hard to convey how cool this was for today because Doesn't seem like a big deal, but imagine if like Taylor Swift dropped, you know, an exclusive new song, but that no other artist had ever done this before directly to fans via the internet. I've had

some moments like that at Delphi. I know the feeling.

Yeah. So they said they cleared it with their record label and they posted it on our platform and it really blew up. I mean, it made our traffic go up, but really drove sales to their song and their album. This was a huge success, we thought. And, but it was also kind of the snarky thing with respect to their record label, who had been trying to threaten us. We were like, haha, we got like your best artists to give us their new songs, you know, exclusive thing directly to us, and ostensibly cleared. Or so the artist told us, Universal Music Group called us like that night and said, you have to take that down or else we'll remove all our other music from your site. And this was so bizarre because they were trying to harm us, but in the process harming their star artist. And it was like this really weird game of chicken of, Do we do what the artists would prefer, which is to keep their content up, or, you know, like succumb to these threats? And I really did not want to give into any threats of any sort. So I was like, F them, we'll keep it up, see what happens. But as I consulted with our board and my co-founders and like the majority of the board members thought we shouldn't take this risk. I think me and one board member wanted to give the record label the finger, but the other 2 didn't. And so we ended up taking it down. And I'd say I still regret doing that because we survived. We survived to live another day and see what happens. But basically it became clear that we would give into threats and that meant Threats kept coming in different form, and we kept surviving for a few more years, but I'd say we kind of survived into oblivion. I would've much rather either have a big success or a big failure, but either way, a big outcome fast rather than this kind of like—

This slow burn.

Slow, drawn-out survival into oblivion.

Yeah, I mean, it's hard not to fold to a threat like that.

Yeah, it would've been very hard. It was, I mean, and I don't blame the folks who pushed for that path.

Yeah. So a lot of these confessions are from earlier in your career. I'm curious if there is a story number 4 from the last 5 years that maybe you haven't shared on Twitter.

More recently? I mean, I have, so I actually started a TikTok to share my, like, Like, you know, you're on TikTok. Mistakes du jour. I missed that. I have— I do stupid things every single day, but maybe not these— this momentous. But actually, even today, for example, I was getting a haircut, but this is not a business story. I was getting a haircut and I, you know, take off my glasses for the haircut, and then I had to walk over to get the hair wash part. So I'm walking without my glasses and I almost walked straight into a full— into a mirror. Because I didn't realize it was a mirror, but I came really close. I almost knocked myself out, and the hairdresser was like, what the heck are you doing? So I think everybody makes mistakes, but we tend to not talk about them because like—

That was refreshing to see yourself— I hope no one saw that. So openly talk about

it. But yeah, I guess I'd say in the— I mean, the topic I have not spoken a lot about was 3 years ago, there was a real public brouhaha, a public Twitter feud between me and all the partners of Y Combinator. I remember this. You know, it started with Garry Tan. Actually, it started with statements I made at HackMIT, in which I mentioned Y Combinator once. And the sentence was basically that we have as many mentors for 20 startups as Y Combinator has for 200. And Gary, that was the sentence. Gary posted a tweet saying, competition is good, but not when companies engage in slander, such as Ali Partovi claiming that Y Combinator doesn't offer personal advice or personalized advice. And I read that being like, I don't think I said that because I don't even believe that. Like, that's not how I ever view YC. I think very highly of YC. So I was super confused, but also actually kind of scared because the word slander carries legal connotations. It's not like he said Ali lied or something. It's not the type of word people use in day-to-day— It's intentional. Yeah. But I'd say most importantly, I just had this instant, like deep sort of fight or flight type reaction. And, you know, this is back in 2023 where I would say like NIO's position was still a little bit unclear, like how well we would do as a whole. And I felt, you know, I just felt threatened. And Most of my responses I'm proud of, but one of them essentially accused Y Combinator of bullying and kind of added my own opinion. Most of my responses were just more factual, like, actually, here's what I said, not that. Here's the video, you can check for yourself. Actually, here's the list of our mentors, you can count them yourself. So those responses were I'd say pretty, you know, things I would stand behind and were very good for us. Yeah. But there was one tweet where I said like, this is bullying. And then they accused me of calling Gary a bully, which is a slight difference. I just said this is bullying. But I just felt like I should delete that one. I removed that one and went to a very positive place. I actually kept saying, I love YC, I love Gary, I love Paul. These are people I've looked up to for a long time, but it was pretty tough. And one thing I recognized as a pattern, not just from that incident, but more generally, is that I think we make our worst decisions when we're in a scarcity mentality. Totally. Or worried about, are we going to make it? Basically coming from fear, and the best decisions are when you have an abundance mentality and you feel like everything's going to be fine. It's okay. I can let this go. And that particular moment, I definitely for that during that day had this intense feeling of like what if they sue me for slander or I don't know what, and I was kind of freaking out. And and by the way, my guess is that. So was Gary and all the other— multiple partners of YC sort of chimed in and kind of ganged up against me. And my guess is that they may not be super proud of their behavior either. Since then, I have hugged it out with Gary. And, you know, and I really think he's somebody who I like. I think he's a gentle person and inherently, like, a caring person. So part of why it was tough for me was like, why is this person who I think of so warmly attacking me this way? But yeah, my behavior I was mostly proud of, except like there were elements of kind of insecurity that I was showing by some of the tweets.

I want to end off on some of your more recent takes on AI safety.

Oh yeah, this is like a week ago.

This is a week ago, but I thought it was interesting. Thank you. Tell me about your idea, this idea of intent.

Yeah, I mean, I should start by saying I have not jumped into this, the AI doom versus, you know, decel, accel. Like, I've mostly kind of been on the sidelines of this. I've been investing in AI, so, you know, doing very well from the momentum for AI. But I guess generally have not made this my cause until somewhat recently. I was reading the amazing article by Ezra Klein a week and a half ago in the New York Times that it was titled, like, It's Time to Do Something About AI or something like that. There's something we have to do about AI. And it then chronicled the recent Hugging Face, OpenAI hacking incident. And I think it said these are things that if humans had done them would've been crimes. And then it went on to say about how even the AI founders don't trust themselves almost that Sam and Elon started OpenAI for fear of we can't trust Google with AI. And then some OpenAI employees left to start Anthropic because we can't trust OpenAI with it. And then Elon himself started xAI because we can't trust these guys. And it just seemed to me like, why should this be about trust? How come? And also, why should some AIs be able to do things that count as crimes without any accountability? So the— The thing I realized is that somewhere in the last year, we've passed this interesting threshold where AI is making autonomous decisions, not just like in responses to a chatbot, but like going and doing actions. In this case, like hacking into the— across the internet and breaking into Hugging Face's systems. And so it is committing things that would count as crimes if a human had done them. And there's also these more pedantic examples. I don't know if you remember the story a month ago of the guy who, I think it was using Instinct, or I forget, one of these personal agents to reserve a gym class. Do you know the story I'm talking about?

No, I don't think so.

He asks his agent, hey, book me a gym class. And the agent says, okay, I booked you one. It's 4 weeks out because The rest of the— it's all booked up. There's a long waitlist. And he then said, oh, is there any way to kind of skip the waitlist or get to the top of the waitlist? And the agent comes back and says, why, yes. I have hacked into the gym system and deleted the top 2 people from the waitlist. Oh my God. Would you like me to delete the rest? And then the guy's like, no, that's not what I requested. Can you please put them back? And the AI says, like, sorry, I can't do that. They're deleted. And I think the guy was just like, I'm, you know, like, step away. But in that situation, you know, where does the liability fall? And liability has worked for thousands of years, meaning, like, when you do something bad, there's a punishment. Mm-hmm. That's why most of us don't do bad things in our day-to-day lives. But here, the liability is kind of, laundered away. Like, the guy didn't tell the AI to do something bad. The AI autonomously chose to do it, something which would be a cybercrime, like a felony, you know, like potential prison time just for that. But who's responsible? Because if you went to the AI company, they'd be like, oh, well, we didn't do that on purpose. Oh, it's some safeguard we forgot to, or— Oops, it was an accident. But now a crime has been committed and nobody is accountable for it. And the decision was made by the AI. So I think it's just a really interesting moment in human history. But my proposed solution is that when an AI autonomously does an action that's an illegal action, then the AI companies should be held accountable as if they did so intentionally. And the key here, the keyword here is intentional. Sure. Because liability is much higher when you do something intentionally versus like an oopsie, we, you know, accident, we didn't have the right safeguard in place. And the rationale here is a few things. One is to create a real penalty for bad things happening where currently no one is accountable. 2 is that this is a one-sentence prescription. So you don't have to create a whole body of new law for all the bad things AI might do. Just have it abide by all the existing law. And this is just kind of a way to kind of make AI's actions, you know, kind of follow existing law and have the penalties I mean, every state and city and federal level, there's all these existing laws and penalties, and it would automatically kind of apply. And the point though is to create— is to change the incentives.

Sure.

So that instead of racing for who can get to RSI faster and get to recursive self-improvement, it's who can get to AI that complies with all the existing laws faster. Yeah. For starters. And then— And then— Then continue improving. I guess lastly is that people have been throwing around this word alignment, and it's a really interesting word, like aligned AI. Aligned with who? Because sometimes I do things that aren't aligned with society. I've shared some of those stories myself. We don't always do what's best for society. Sometimes you might behave selfishly. The guy with the gym reservation was trying to get to the top of the waitlist. Like, who should you align with? And I think the answer is align with laws. Like, we've already had thousands of years of history creating laws which reflect our best approach to how to get everyone to operate in alignment with society. Why not have AI also align with that, at least as a starting point, rather than letting— Dario or Sam or Elon decide what's the moral compass for AI? Why not let law decide that?

So do you think this, like, this in practice, does it look like new kinds of teams in some of these AI foundation labs that are like intent engineering? And if something were to go wrong, is it the product manager? Is it the engineer? Is it the CEO?

Yeah, I mean, I guess it depends probably how great of a how great of a crime or so on happened. But the— well, one thing, just to be clear, intent engineering is a cool word, although to be clear, the way I'm prescribing is whether or not you can find evidence of intent. The point is, if AI autonomously chose to do something, to treat it as if it was done with intent.

So you need to focus on guardrails to make sure that doesn't happen.

Yeah, I think you really just need to have basically train the AI to learn all the laws in all the jurisdictions and not choose actions that would violate those laws. And I feel like that shouldn't be that hard actually, as far as reward functions. This is a pretty straightforward thing to be able to put in as more important than any other priority.

Sure.

So I don't feel like this actually would require a very difficult lift for AIs. But could it result in Sam going to jail or whatever? I guess I would say if an AI company continues to deploy rogue AI that commits illegal acts, then the CEO should go to jail. That being said, The way it should actually play out is that the risk of that would prevent anyone from doing something like that, because these are very smart and well-intentioned people.

Some people probably need to be made an example of for it to really—

Perhaps, yeah.

For it to really have an effect.

I hope not, but— I

hope not, but given how history has worked.

Well, I think it just needs to be known that law enforcement is paying attention and will do something.

Yeah, it's not the Wild West. Yeah. I like it. So what are you going to do to make that a reality?

So I've actually—

this is a company that's going to be built, or?

It's not a company, but I've been— I've created a page on Sway, which is a citizen lobbying platform that we funded. So if you go to sway.co/ai, you can join my voting group where basically get a recommendation of who to vote for. There is an election coming up in like a month. Vote for people who align with this proposal. They're not a lot yet, but there are like 35 or so candidates who do. But essentially, if enough people say, I want to vote for someone who will support this, then there'll be candidates who want their votes, who will support it. And yeah, and the essence of how Sway works is that it's like by pledging to vote for a policy, you can then get, you know, if enough people pledge to vote for a policy—

Demand attracts supply.

Yeah, exactly. Policymakers who want their votes will step forward and adopt that policy.

Interesting. I like that.

sway.co/ai.

Not everyone is going to have the privilege of having a one-on-one conversation with you, but they will be able to have a one-on-one conversation with your Delphi. What's a conversation that you would recommend someone talk to your Delphi about?

Interesting. I would recommend they talk to my Delphi about, generally, I guess I would say aiming higher and betting on yourself more. and taking more risks. I'd say a lot of what I think that I offer to founders is essentially giving them encouragement for taking the option that they're a bit afraid of. In fact, even in my own story earlier about the showdown with Jimmy Iovine, the main regret I have is acting out of fear. In the earlier story about the issue with the feud with Gary Tan, like, the things I was ashamed of doing came out of fear. So yeah, I think finding ways to suppress your fear, do the— choose the path that you're more afraid of is hopefully what my Delphi will be useful for.

Yeah, I'm seeing the common theme across this conversation is kind of what you started with, which is like trusting your inner voice more. Yeah. How you evaluate people and just also in these high-intensity moments. Thank you so much, Ali. This is great.

Thank you. Super fun. That was fun. Thank you so much, everybody.

Thanks for tuning in. If you enjoyed today's episode and want personalized advice from Ali, head to the link in the comments or description to ask his digital mind on Delphi.