Ravi Gupta——AI，或死——[Invest Like the Best，第411期]

My guest today is my good friend Ravi Gupta. Ravi is a partner at Sequoia Capital and also the host of Glue Guys, a podcast on the Colossus network that discusses both business and sports. I wanted to have him back on Invest Like The Best to discuss a recent essay of his that he called AI or Die. As someone with incredible operating and investing chops, Ravi believes we're entering an era where the constraints that historically limited small teams are dissolving, creating unprecedented opportunities for those willing to embrace change aggressively. We discuss why traditional metrics of corporate success like headcount and process adherence may become liabilities, what it means to be a world-class reactor versus a world-class predictor, and how magic per employee and organizational agility will emerge as crucial measures of value creation. Please enjoy my great conversation with Ravi Gupta.

So, Ravi, a couple of days ago you sent me a draft of this really interesting essay that you wrote called “AI or Die,” which is such an interesting framing for something that I know you've been thinking a lot about and have been thinking more and more about in a crescendoing way in the last few weeks and months. Can you start by describing the process of arriving at this essay and what precipitated it? Then we'll talk about all of its various implications.

1. The Pace Of Progress

Like everyone, you see these new model advances, and they've come out so quickly that you are using the models for whatever you're using them for one day, and it's hard to think of even what to ask them when the new ones come out.

A couple of things happened. One, I'm on the board of Sierra, Bret and Clay's company, and they started describing some of the things that were able to be done in o1 Pro that were not able to be done with 4o. I was surprised. The specific example that came up was when Clay was talking about something he was doing with it. I asked him to show me how he was using it, and when he did that, he asked the model a question: “Are my instructions clear? Is there anything else I can do to clarify what I'd like to have you do?”

That interaction was so different from anything I'd ever thought about with a model. I'd never thought to ask it a question. The model came back and said, “No, in fact, can you please clarify these 5 or 6 things?” I was pretty stunned by that interaction. That was point 1.

Point 2 was that a friend of mine who was at one of the model companies left to go to another one. I said, “Hey, what happened?” He said, “The pace of progress has changed so much in the last 3 months that I had to go to this one.”

In our business, take the Matt Kohler quote that you and I both love. Matt is a good friend of both of ours: “Our job is not to see the future; it's to see the present very clearly.” All of a sudden, it felt like the present was different from what I thought. I started trying to use it more myself, and I think I'm barely scratching the surface, but I was pretty stunned at what was possible from this. I'm somebody who lives in Silicon Valley and works as a venture capitalist, and it made me wonder whether other people were going to be as stunned as I was.

So maybe talk about what you mean by “AI or Die” in its most intense form.

2. AI Or Die Changes Everything

If you take that Dario quote from his essay from October, he describes what he believes powerful AI will be, and it's this idea of a country of geniuses in a data center. That's what's going to be available to anyone. Then you take what Sam has been saying. The quote he has is, “In a decade, every person will be more capable than any person is today.”

If you just take a moment and literally read the words again, or say them again in your brain, they are mind-bending. The point of “AI or Die” is that if they're at all right—if there's something approaching right—there's going to be dramatic change coming to any company and any person.

If we think about the most intense version of that, then you should be able to recreate what we have now with dramatically fewer people working on it because there's a country of geniuses available in a data center. I think the most intense version of that is envisioning a group of developers hanging out on a Friday night, sorting companies based on their market cap, their number of people, and their NPS in some calculation, and finding the biggest company with the most people and the lowest NPS, then competing over who can build it faster. It's just a crazy thought of what could happen, but I don't think it's crazy if the AIs turn out to be as powerful as we think.

The much less intense version that I think is very real is that anything that prevents you from embracing change right now is a huge limiter. It's a huge limiter. If it's a lot of people in your organization, adherence to quarterly earnings that you've committed to, or an unwillingness to make a change, anything that reduces your agility is a massive problem if this is changing as fast as I think it is—and as fast as the people who run these companies say it is.

Since you first had these couple of aha moments with your friends at those leading-edge companies, what have you personally done differently to try to respect this faster-than-you-thought pace of change?

3. Putting AI To Work

I love the word choice, “to respect it.” The first thing, to be very honest, was that I talked to Avni about what we were doing to educate our kids if this was true. My first thought was not, What does it mean for businesses? It was, What does it mean for our family? What does it mean for the kids? What is the right way to teach them in a world where the tools that are going to be available to them are going to be totally different?

We realized that there aren't actually skills that I can teach them. There are behaviors, feelings, and approaches. What we got to was some version of: We've got to make sure that they're ambitious, curious, resilient, adaptable, and high-agency.

I think the reason that that then dovetailed with what I've changed now is that I thought, That's the same thing you have to be if you're at a company. You actually need the exact same things. You and I have talked about this separately, and I think you've added a couple to that, which we should go through. Those same behaviors are going to have to exist in the company.

Specifically, what I've done personally is try to use it more for real work. I had dinner with somebody recently, and I hope he's not listening because of the setup I'm going to give you. It was with somebody who was on the board of a company that I didn't know that much about. I was not well prepared for the dinner, and it was somebody I wanted to make sure I had a productive conversation with. I had 20 minutes before I was going to the dinner, and I prompted the latest model, the $200-a-month one.

I know sometimes people just don't want to do that because $200 a month is a lot of money. It is, but it's not a lot of money when you think about it relative to seeing the future. That's literally what they're offering you right now.

I said, “I am a sophisticated investor”—that was a very generous term for myself—“and I have a dinner in 20 minutes. I am not prepared for this dinner, and I'd like to understand deeply what this company does in a way that I can understand within 10 or 15 minutes. I'd like to understand these 4 or 5 things, and I'd like to have some good ideas for what this person could do. Are my instructions clear?”

I used Deep Research, and it said, “No, can you tell me these 4 or 5 things?” I told it. After 4 minutes of thinking, it came back, and I read it on the way to the dinner. We had a great dinner. This person thought that I knew what I was talking about.

That is a pretty different use from, “Can you summarize this email for me?” I think the AI interaction most of us have is Apple's. Apple's an amazing company, but the Apple Intelligence summary of your text—

Terrible.

message is the most absurd. It's a terrible experience, right? It's truly laughable. It's used more for comedy than it is for actual value right now. And that's the interaction that most of us are having with this thing versus that “Wow” experience for me of, “Oh my gosh, I actually know a lot more about this company now, and I would have never known this in this time period.” That's real work.

So the biggest change is trying to have it help me with real work.

I want to dig in a little bit more to that imaginary spreadsheet you envisioned earlier, which you said had market cap, number of employees, and NPS. If your only thought exercise was trying to find a single target for a company to go after, you could have an activist help you with a hostile takeover of the company or something, and install yourself as the CEO with the goal of maximum value creation using this new tooling. Talk me through how you would process that search, trying to find that single company that could be most positively impacted if they embrace this—back to “AI or die”—and maybe will die if whoever's running it doesn't embrace this rapid pace of change.

4. Finding The Disruption

That is a good question and a very hard question. I think that when we say AI is much better than we think, practically speaking, it means that it's way better—and this is a stolen insight from a friend of mine—but it's better at figuring out hard-to-figure-out patterns than we would give it credit for. Do you see what I'm saying? It finds patterns in ways that wouldn't be obvious, and it finds them ridiculously quickly.

So you can be more ambitious about where there might be a pattern that's not obvious. Its ability to discern that is different. Maybe an example would be that if you give it some framework, it can give you good acquisition ideas for the company because it can identify ways to optimize for the question you have. “I want to improve my distribution, and I'm in this space.” It will find companies that are in a similar space, have tons of customers, and might be a nice tuck-in or transformational acquisition.

I guess one thing would be: Where can AI help me be more ambitious about thinking about what it could do? Practically speaking, Patrick, I wrote this in the blog: I think going through role by role and figuring out what that job actually does inside my company for a customer. If I replaced it with a really, really, really well-prompted AI, could it do it?

And I think that you have vertical, agent-oriented companies all over the place right now. You have them in customer service, in SDRs, and in marketing. There are companies popping up, frankly, on the marketing side right now. You go through each one and say, “Okay, which one of those could be replaced or augmented by AI today?” Then—and this is a really important one—what about in 6 months?

Paul Graham had this tweet that you and I both saw, which was that the people who are really bleeding-edge in AI should be building things that don't quite work right now with the models and are way too expensive. Because what do you know? You know that the model is going to get much better in the next 3 or 6 months, and you know it's going to get way cheaper, at 1/100th the price or whatever.

So I think maybe role by role: Can AI do it now? Can AI do it in 6 months? That's probably one exercise that I try to go through to think about the disruptability of my company by somebody else.

The second thing that I'd probably be thinking about is the ability to be the AI strategy for another company. Let's assume most companies are not going to be able to do this themselves. Can I be their answer for the future? That, to me, is an absolutely gold insight if your company has that possibility.

If you can promise your customers, “The world is changing really quickly. We are your way to deal with that,” that is incredibly valuable, I think, because all of a sudden you have infinite ability to move out of one spot into more because the AI will continue to get better and better.

The third thing—and this is not exactly answering your question—but I do think this is a big deal related to companies that are going to struggle. I think anything that reduces your agility in any way, anything that prevents you from adapting or changing, is 100x more painful than it's ever been before.

This is where I think all of us will start to think of companies and say, “Oh my gosh, everything is a sunk cost.” The example that comes to mind here is Satya Nadella in Q4 2022. Our mutual friend, Modest Proposal, has talked about this. If you go back and look at the earnings report, in Q4 2022 Satya Nadella is talking about the enterprise metaverse. That is discussed on the Microsoft earnings call 2 years ago.

Now look at what's discussed on the Microsoft earnings call. It's, “We're going to spend $100 billion on CapEx. We're doing everything related to AI.” He doesn't care what he said on that earnings call. All he cares about is what is happening going forward.

I think that so many companies are held hostage by what they said to their team, what they said on an earnings call, or what they said was going to happen. This willingness or desire to be consistent with their past selves is preventing them from embracing the biggest technological change that we've probably seen in our lifetimes.

Can you riff a little bit on how all of this change and this frame of thinking makes you think about individual employees inside of an organization and what they actually cost? The simple thing would be that they cost $150,000 in salary and benefits or something. But I think you're thinking about this in some new ways, which go beyond just hard, direct costs and are more holistic from the company's perspective, related to the customer. Talk about the cost of an employee and how that's changing.

5. The Small Team Advantage

You and I have been talking about how there's this meme that's starting to arise around small companies—

Small-team meme.

Small-team meme. Yeah, small-team meme. I think that Dharmesh Shah from HubSpot talked about this: SMB doesn't stand for small and medium businesses anymore; it stands for small and mighty businesses.

If you start with this premise, maybe small is even better than we thought. Maybe small is even more of an advantage than we thought. The question is, why?

I think that there are a lot of hidden costs to having a lot of people. We've all talked about this in the past: coordinating amongst them. That seems like a hidden cost. Then there are all the other things that we've talked about in the past. If you have a lot of people, you spend a lot of time interviewing and hiring. You also spend a lot of time putting people on PIPs. You spend a lot of time discussing all those very specific things.

I think the bigger-picture, zoomed-out question is: How much of your time is spent dealing with employees in a way that doesn't actually help your customer? What percentage of your day, if you're a CEO, founder, or leader, is spent on things that your customer cares about? I think that number is vanishingly and astonishingly small for a lot of people.

This is why I think sometimes you'll hear CEOs say, “I don't even like my job,” because they don't actually spend that much time doing stuff for the customers. I do think that risk exponentially goes up the more people you have.

I have a friend who runs an organization that's around 400 people, and he was telling me that he had to let go of around 20 people last year. He's like, “It's taken up the craziest amount of time figuring out how to do the layoff, figuring out how to talk to each of those people, what the severance package is for each of those people, how to talk to the people who stay, how to describe to them what's going on and why it's going to be durable and why it's okay.”

There are just all these hidden costs that go in, and they all reduce the percentage of time you are focused on delivering a resolution to your customer. If we envision that there's this incredible advantage right now that's pronounced—and even more pronounced than it's probably ever been before—it's the advantage of people who spend time thinking about what their customer wants and how to deliver it to them in the best, fastest, cheapest way.

I think the moments when you get to free up your brain to think about that are the moments that are going to deliver value for your company. Going back to your question of how I would think about the biggest value-creation lever for a company if you were dropped in as the leader, it would be: What percentage of time does the senior leadership team think about how to deliver value to the customer in the best, fastest, and cheapest way in 6 months, 12 months, 18 months, and 24 months?

The lower that percentage is, the more opportunity there is. There's a massive hidden cost right now.

Are there companies that you know well enough that you would say are exemplars of doing a great job of the senior team spending lots of their time on what's good for the customer?

I'll give you a couple of examples that come to mind. I'll give you one that's inside our portfolio, one that I work with closely, and one that I don't.

Bill McDermott, famously, when he took over as CEO of ServiceNow, talked about the number of customer meetings he had in his first 90 days, and it was some number that effectively equated to 10 per day. Ten customer meetings per day, including weekends. I think he said he met 1,000 customers in the first quarter.

That is baffling. ServiceNow is a huge company. He has all these things that he has to do. If what he's saying is true—which I presume it is, because he's Bill McDermott and he's an amazing CEO—that would make me pretty darn bullish about ServiceNow realizing what its customers want and how to go build it for them, because the CEO is so incredibly externally focused.

I love this idea, and every company that I am lucky enough to talk to, I tell them, “Please be externally focused.” First, it’s so much more fun, but it’s also so much more impactful to go and solve customer problems than it is to deal with internal nonsense.

I do think Sierra is doing a really good job on this dimension, too. I think that team is focused on what customers want, how they can deliver it to them, and how they can embrace AI to the maximum ability. I think that means don’t do one-off things like have AI help you with it. Anytime you figure something out that can be helped with AI, do it. Tell the whole team.

Practically speaking, Patrick, I think it’s almost easier for us to think about the companies that probably aren’t doing this, and they probably make us have a little bit of a pit in our stomach: “Oh gosh, how is that going to work? What are they going to do then? Are they going to go and change?” I actually think right now being private is a huge advantage, as a general matter, because you really have to be courageous to not let something that you’ve said in the past to a public-market investor stop you from doing things.

You know how big of a fan I am of Fiji. Fidji will do whatever is right for Instacart. It doesn’t matter if they’re private or public. But there are a lot of people who feel really obligated to their past selves and not to doing the right thing for the long term.

I also think it’s not a coincidence that the people who are going to do this well are people who understand customer problems, but then they also understand the technology of what it can do. I think the anecdotal experience of AI not changing your life is really dangerous right now. If you want to be contrarian right now and be like, “God, the models, they’re not that great,” that’s fun to say, probably at a cocktail party, and it fits with people’s anecdotal experiences. But that is dangerous as hell. Go sit with a bunch of people devoting their lives to this and have them tell you about the future. I really believe that that is incredibly valuable.

What do you make of this inversion of that old trope that the race was between startups getting distribution before the incumbents got innovation? Going back to the small-team meme, everyone’s posting about Cursor and companies like it that are growing from $0 to $100 million in revenue with 10 to 30 people in a year. You could debate how much better Cursor is than a big company’s Copilot or something, but they’re getting distribution way faster than we’ve ever seen before. What do you make of that, and what does that tell us about how the future might unfold?

6. Distribution Still Beats Incumbency

I actually think that trope that you mentioned was pretty brilliant for a very long time. I suspect it still applies. Can a startup build distribution faster than a big company can build a product or build innovation? I actually think it’s still the right question.

I think the thing that’s probably changed is that the speed at which you can build distribution with a great product has probably gone up. The willingness for somebody to adopt something that feels like magic—maybe that’s the point. I think that AI can make your product feel like magic, and I think people are really willing to adopt things that feel like magic. So the speed at which a really well-executing startup that produces magic can get distribution is probably higher than it’s ever been, would be my guess. I’m not enough of a student of this to know that for sure, but I do think they are producing magic.

The thing that’s actually interesting, though, if you’re a startup, just to make it real, is that if somebody already has distribution, thoughtfully deploys AI, actually goes and disrupts themselves, and puts it into their product, they’re even harder to compete with right now. I think Microsoft is way harder to compete with now than it’s ever been before because they have this ridiculous distribution—insane distribution—and they are really pushing to get AI into their products in a way that feels like magic.

I think that no one doubts that Satya Nadella gets this. Nobody wonders whether they get it. The only thing that limits their speed is their organizational size. It’s not intestinal fortitude from Satya. I think there are other big companies that you really worry will not actually be able to get the product in because they’re going to let organizational stuff get in their way.

I actually think it’s still the same question. I just think that if you’re a big company, you better be awfully paranoid about a startup because its ability to build distribution is way faster than it’s ever been. If you’re a startup, you better be really clear about who you’re going up against. There is a massive difference in speed among the big companies in how quickly they are leveraging their distribution advantage.

It’s becoming popular to say, “Be an AI-first or AI-native company,” for startups. It’s kind of obvious what that means. What does that mean for a 10,000-person public company that sells software, provides a service, or is a white-collar-ish company or something like that? What would you recommend those CEOs literally do? Not a mindset or a first-100-days presidential plan where they say, “We’re going to become AI-native.” What does that mean?

You and I have talked in the past about this idea of enthusiastically rehiring somebody as the standard: If you could do it again, would you hire this person to do this job? I think you have to enthusiastically re-underwrite your business. What I mean by that specifically is that you need to go and say, “What is the purpose of your company for these customers? What is the best, fastest, and cheapest way for us to go and deliver that? Is it what we’re doing now or not?” You can hold nothing sacred.

Let’s take pricing. Today, a lot of companies have seat-based pricing models. I don’t believe that will be the future of the way that people buy software. I think there is something we’ve heard from different people: Maybe they’re going to pay for the job to be done. Maybe they’re just going to say, “I want this piece of work done, and I want to pay just for that piece of work. I don’t want to pay for each person who uses your software. I want to pay only when something gets done.”

That is a massive and dramatic shift if you’re a big company, to move away from the pricing model that got you there. But understanding that that might be the future of your pricing is a big realization that has a bunch of knock-on effects as to what you go do then. Maybe it means you’ll have 60% less revenue now if you move to that, but in the future you have the potential for 5× more revenue because you have a bigger customer base that you can go after, or you can do more jobs for somebody.

I think one thing about holding nothing dear is that the pricing model has to be evaluated, all under this idea of enthusiastically re-underwriting the way that you’re delivering the purpose of your company to its customers. That’s one. Two, I think that you do have to go role by role and figure out what the AI-superpower version of each of these roles is. Is it that it’s replaced? Is it that it’s augmented?

I think that is a big deal—the specific, role-by-role conversation. I know people are like, “Well, we have 5,000 people. I don’t want to do that.” I’m like, “Can you tell me what’s more important than that? What is more important than figuring out whether you have the right people doing the right jobs for your company?”

The third thing, Patrick, is this idea of what percentage of our time we spend thinking about our customers. I really think that’s valuable. You know this thing that Shopify does? Shopify will just kill meetings and end recurring meetings randomly. I think doing a calendar audit of what percentage of our time as a company—as a leadership team, first as a CEO, then as a leadership team, then as a company—is spent on things that move the needle for our customers is valuable. What percentage of time do our engineers spend coding? All these things I would do, and it’s easy for me to say because I don’t have the responsibility of doing this. You asked what I would suggest to somebody. It would be that.

The last one, which is maybe the most optimistic, is: What have we dreamed about providing for our customers that was impossible before? Let’s all use this idea of what is barely possible now and too expensive. Oh my God, that’s going to be possible in 6 or 12 months. Let’s go build that.

One of the things that you and I have been talking about separately is how much we should care about margins right now. It actually means you should care a lot about margins in certain areas and not a lot about other areas. If something’s an AI-first product, you probably shouldn’t care that much about its margins, because it’s going to get so much cheaper to deliver. Whereas if something’s not an AI-first product, you should care a lot about the margins because that is the durable future.

Those are a few things that I think I would do and suggest to people. I also think anything that prevents me from being agile—anything that prevents me from making a change—I would try to write down and figure out how to eliminate. Maybe that means you don’t provide guidance anymore. Maybe it means that you tell people, “There’s going to be a negative, and then in the long term there’s going to be a positive.” Maybe it means that you tell your team that you’re in the midst of a big change and some things are not going to be as predictable.

You have to be able to make the changes you need in the company, and it's hard to predict those changes right now. I think you need to give yourself the ability to be very, very, very nimble.

What would you say to a leader who said, “Okay, I get it. This is cool. I love using ChatGPT on the weekends. I understand why this is interesting, but I'm scared that I'm going to pivot my entire battleship of a company in this new direction, and it's a wild-goose chase. It takes longer, or things change too much, and it doesn't actually work in production use cases. I'm going to waste my next 2 years chasing this thing, as I might have chased the metaverse. There's no there there, and I show up and it's just like it was a waste of my time, and I screwed up the company as a result.”

“I'm willing to eat years to wait and see where these things get and where the equilibrium settles out, and wherever that third Geoffrey Moore category is—not early adopter, not early majority, but the next one. I'll be one of those, and I'll be fine.”

7. Conviction Cannot Be Outsourced

It is so easy to say that when you're not in the seat. I want to recognize that. My response to that is: you can't outsource your conviction, which means don't do it because somebody tells you. I do think, before you come to that conclusion, you owe it to yourself to spend weeks or a month really trying to convince yourself that you're wrong, because I think this is the point that I feel most strongly about.

Go figure out for yourself what's actually happening, with a lens of trying to break your prior thought on it. It's convenient to think that you can just gradually get into this. It is really convenient to think, “I'll just figure it out later, and I'll do it in a gradual way,” because that's just way easier for us to deal with as people. I think any time you have a convenient belief about something important, you should really examine it. In this case, examine that convenient belief, and if, after a month of being super open-minded and really trying to break it, you come up with that, great. Do it.

But do it affirmatively. Don't do it because it's the convenient path that allows you to not have to change very much. You and I are both NBA fans. It is really hard for an NBA team to be really good for a long time and never bottom out before they get to be good again. It's really hard to just keep it going and reloading and reloading and reloading. It usually has a really painful period afterward.

I'm a Bulls fan. Look at the Bulls—geez. It's very convenient to think, “Maybe I'll just be able to do it both ways. I'll just be able to stay good, and then I'll just figure it out and get the second track, and we'll make it work.” Maybe, but I would affirmatively really try to figure it out by trying to break it in your own mind.

One of my favorite points that you wrote about and that you and I have talked about is whether or not you're treating these things like you would a genius who just joined your team, with the capabilities it has today—forget 6 months from now. If that person showed up at your office and you're paying them, how would you treat them? Are you treating these models and their capabilities in the same way? I love that thought experiment. Can you flesh that out?

Yeah. I think we have a standard for these models that is so insane right now, which is effectively: if I just ask you to do something, and I'm not clear on what I'm asking you and you don't know anything about my company, do you just make it happen? If you don't get it perfectly the first time, I basically am like, “Meh. This thing doesn't work.”

The thought experiment you and I are talking about is: what if you just knew the person was brilliant, and you knew that they would work 24 hours a day for you, and you just knew that they had this crazy potential? The first time they messed up, you wouldn't just be like, “Oh, they suck.” You'd be like, “I have to do something differently in order to get the most out of this person.” That would just totally be the reaction.

Instead, I think we are treating the models like they are someone's nephew that we had to hire, and it's like, “No, I gave him the chance, and I knew it. He sucked.” The mental model should be: no, this is the person that you fought like crazy to hire, and you were so lucky to get, and you didn't think they were going to come, and the first project wasn't perfect. It's like, “No, I'm going to make this work.” That mental model is so different and so much better, and frankly will have so much positivity for your company.

I really like that mental model. If something doesn't work right with the model, it's your fault, not the model's. It's not a question of the model's capability. It's a question of your prompting and your understanding and your ability to give it context. Context is going to grow even more important here. I think most of what's missing from the models in order to be magic for you is context.

Part of the reason I think that Cursor and Cognition, along with some of these other companies, have done so well is that they actually have shared context of the codebase, and then they get better and better with the more work that they do alongside you. I think what's missing a lot of the time with what people try to get out of the model is giving it enough context for it to be great.

The relationship idea here would be that the relationship in which most of us have the most shared context is the one with our spouse. With your spouse, you can flicker your eyes at a party, and they know it's time to go. You just have so much shared context that the smallest movement of the way your eyes look, or the length of a blink, and all of a sudden they know everything.

Avni can perfectly tell what I will think about something without me even being there. She can perfectly imagine how I will react to something. That's because of a lot of shared context. I think my point on envisioning these models as a collaborator is: now take that to work. There are people you've worked with for a long time where they just know what you're going to want. These models, I think, can and will be able to do that as collaborators, but we have to give them the context.

I think the only way to really lead through this is to lead through optimism. The only thing I didn't like about the title of this recent blog is that it sounds negative: “AI or Die.” Remember, yes, if you don't embrace this, your company has a real risk of fading into irrelevance. But you can live in a way that you never thought possible. You can make magic. You are here to make magic for these customers, and you actually now have the tools to do it.

That's the part that I really believe has changed in my thinking on it. If you approach this with optimism, dude, what a time to be alive. What a time to be alive.

I love Tyler Cowen's idea that most of his writing now is for AIs, not for humans—that he is generating context that will make them better for him in the future. I hope that there are more and more ways that make it easy to do that. I think lots of people who are optimistic about this would very quickly opt into something that generated more context in an easy, low-friction way for these things to work with. I love that idea.

Think about context—how much you already have and how much you're building as part of preparing to use these things really well. I also really like your idea of the difference between a world-class predictor and a world-class reactor. Can you talk about that new dichotomy?

It's such a perfect question for me because it gets to merge together things that I love that may not seem like they have anything to do with each other. The first person I heard say that was Coach K at Duke. He said, “I am not a world-class predictor, but I am a world-class reactor.” He was referring to when college basketball changed from people who stay for 4 years to people who play for 1 and then go to the NBA. He was like, “I couldn't have predicted that that's the way the rules would go. I didn't know. But you know what? Once that was the game on the field, I played it extremely well.”

That actually was very helpful for me, Patrick, along with Kohler's quote, because it's very intimidating to think about predicting the future. There are really smart people in Silicon Valley. There are really smart people all over the world. I don't know how to predict the future. I do not have the IQ points for that. There are people who are much brighter than me who can do that.

I think the thing that I do have in my control is the ability to respond quickly to new information. I can control that. I can control being open-minded to seeing new information and then reacting quickly. I can control my aversion to change. That feels empowering to me, and it's something that's in my hands, and it's in the team's hands too. I can help the team be like, “Guys, let's embrace this change. Let's embrace this reality, and let's go react to it.”

I think right now the world is totally awesome for people who can be world-class reactors rather than world-class predictors. If you look at the people who created these labs, they were world-class predictors. That was an amazing insight however many years ago: look at what the scaling laws are going to do; look at these exponentials. They themselves, funny enough, are humble about it. They're like, “It wasn't that hard. If you just believe that the scaling laws continue, you could bet that this was going to happen.”

But it was hard. Looking at what a model can do today and thinking, “That's going to be cheaper in the future,” isn't being a world-class predictor. That's just looking at the facts. What is hard is choosing optimism and going to react to that, but it's totally within our control. It's hard the way going to the gym is hard.

It's not hard the way creating a new mathematical theorem is hard. It takes discipline, and it takes fortitude, and it takes a choice. But I love those kinds of things because those are in our control, and they don't require a new brain. They require more discipline. That's something all of us can do.

What do you think you would have done if you zoom back, you pick the moment, to the most intense operational trenches that you were in at Instacart if you had this set of tools back then? What do you think that would have been like? How would you have used them?

If we take our logic of work from the customer backwards, one thing I think that we were good at at Instacart, and Instacart's still good at, is the percentage of time they spend thinking about the customer. So I don't think there's a huge optimization on that. I do think that figuring out the most abstract thing—what does a customer want? A customer wants to save the trip to the store. They don't want to go, and they want to be able to be at home when they get it.

So when I was thinking, to your point, about having these tools—the better, faster, cheaper version—I would be trying to make the order use as much context as I could to make their ordering process be as fast as possible and to make sure that their order is delivered as accurately as possible. Those are the main things. When I say make the order as fast as possible, I would be trying to do stuff where their last grocery order is something that I can ingest in some way, even if it wasn't on Instacart. Therefore, I can get them through to where they can check out as quickly as possible.

And then, on the as-accurately-as-possible side, I'd be trying to be better at figuring out what's in the store right now so that I never show them something that they're not going to get. And then, on the delivery side of it, I am not technical enough to tell you the optimal batching strategy that would be used using an LLM. But whatever I could do to cut minutes off that delivery, I'd be trying to figure out: What is the AI way of doing that?

Maybe the other thing I'd be trying to do is, is there something we can do over the phone that's not economic today but will be economic in the future for people who don't want to do it online but want to just call? Older people maybe aren't as familiar with the app. What can I do to make it feel like magic for them? They say, "I want these groceries." Okay, cool. Then they show up.

What you're saying, though, has something very embedded in it that's very real, which is that the more your product is physical, the more the bottleneck for using AI is the physical world rather than the software world. And I do think it will be slower because there are constraints that are physical in nature. The AI is not fixing traffic today. It's not making a human being run faster through the store.

I think figuring out whether your constraints are physical or whether they're software-related, digital—I think the more digital-only you are, the faster this will come and create magic. It's sort of like this: Either you're going to create magic, or someone else is going to create magic with a lot fewer people than you. And the customer is going to expect magic. This Bezos idea that customers are divinely discontent means they are just going to expect, "We already do this." We are going to expect magic from every experience. And either you're going to deliver it, or someone else is going to deliver it, but that's the new standard.

We've talked a lot about the role of status before in all things. Apply your ideas about status and our status-seeking nature and behavior to this new world.

This has been observed in the past, but I think sometimes there is historical status in business tied to the size of the team that you manage. Sometimes still today, "How many people do you have?" is a question that's asked of a startup founder as some indication of the progress or the scale of their company. How much money have you raised? How many people do you have?

I think if you take this idea that you talked about—the small-team meme—or Dharmesh's idea of small and mighty businesses, it seems possible that the new status is having a small team. The ratio of magic to people—and I think magic, sure, it can be measured in revenue if you want to use a very blunt instrument—but maybe this idea of magic per employee implies being like, "Oh my God, you did that with 10 people? Woo!" That kind of thing.

Can't you feel that already? That feels like that shift has happened already.

I think so, and I think that if it's happening, that's great. It's not because that means people will lose their job. That's not the point. The point is that each person will deliver more magic. They might leave a bigger thing and start a smaller thing or join a smaller thing and have more impact, have more purpose.

Nobody likes being a cog in the wheel. Nobody likes feeling like if they left, nothing would change. And I think that the more we can get to people getting more out of their potential, that would be freaking awesome. It does seem like it's starting to go that way, and I think that is an awesome thing: small teams feeling like they can do anything. Oh man, there's nothing that stops us from doing that. I think that would be an awesome thing for the ecosystem.

And frankly, I think that this is conflating 2 things, but most companies are probably way too big relative to what they need to be as public companies. There's no way that most of these companies are run efficiently right now. Even without AI, you have this thing where these companies are bloated relative to what they should be. I think now you factor in AI, you actually have a real reason to change that and to make yourself lean, customer-oriented, and seeking of magic. But I do think that this idea of the new status might be a small team. I think that's very possible.

How are you changing your investing as a result of all this? Whether that's the way you look for companies, whether that's the way you underwrite a company, whether that's attributes of a founder that are changing. Some of these things are the small-team-meme companies. If you look at the price of the equity, the valuations that they've raised at are high nominally. They're growing insanely fast. It's hard to know what they should be priced at.

Are you changing your valuation expectations? What are all the ways in which you think you're actually approaching things differently than you did 2 years ago or something?

8. Investing In The Slope

One of my favorite concepts that I've heard—I don't know who the first person was to say this. The first person I heard it from was Pat Grady, my partner—which was, "We invest in slope, not intercept." I guess if I really take that as the logic, I think that's still very true. I think the thing that is possible is that, for the best companies, the slope can be way different, and the slope can be way steeper. And I think if you take that to its extreme, that means that for the best things, sure, you should be willing to pay more.

What I think it really puts a premium on, though, is being earlier into those companies. Figure out what those things are going to be. And I think that what that then comes back to is, okay, what are the attributes of the people who are building them? I think that comes back to some of the stuff that you've talked about and we've talked about, but I mentioned ambition, I mentioned curiosity, I mentioned resilience, I mentioned adaptability. You mentioned imagination in there. I think that's a hugely important word for thinking about this.

So figuring out people who are close to what this can do, optimistic about what it can do, have some imagination, and then are also adaptable as hell, because it keeps changing. When it changes, the only thing that you keep dear is the obligation to deliver magic to the customer, and everything else can change.

Probably the bigger thing I would add to my list, Patrick, is that in the past, I think resilience and grit—those are things that have always been there for any founder that you really want. I think the thing that I do believe is newly important—it's always been important, but more important for me—is agility and adaptability. It's the world-class reactor point. It is the willingness to change on a dime with new information, and the desire to go and figure out that new information, and to never hold a convenient belief and to examine it.

That's probably new, and that combined with more slope for the best ones, I think what's going to happen is the best companies are going to be worth more and the mediocre ones are going to be worth less. I just think the gap is going to widen, and that is exciting and intimidating, because you better get it right.

You get to sit literally at ground zero of the community of people that are investing in these companies. It could probably be well argued that any company raising is going to Sequoia first, or among a handful of firms that they're going to first, and would love to take their capital if they could. You get this complete tip-of-the-spear view into this world.

If you draw a radius around your office, and the class of investors in Silicon Valley within that radius is wise to this, does that mean that maybe the better way to make money as an investor is to be like a private equity investor in New York that just appreciates this more than your peers do who aren't in the center of things? They don't watch Cursor happening.

I don't think it's better. If you believe this is true, I think there are multiple ways to express it, so maybe a few different ways. One, I think if you are an investor in early-stage companies and you pick the right founders building the right things, I think the upside is better than it's ever been. That, I think, is an amazing place to sit, and it's about picking the right people and being lucky enough to have them pick you.

So that, I think, is still an incredible privilege and a lucky place to sit. I genuinely believe that. I think that if you are someone who doesn't sit here, there are different ways that are really good to make money. Figure out the public companies that are actually going to embrace this.

I think that Mark Casey at Capital Group tells this great story from a long time ago about how, during the beginning of the internet era, he was really impressed with how Home Depot was thinking about this. His example was, “We went through item by item that we sell, and we figured out the price of the item relative to the cost of shipping. We specifically went through, and we are tracking that over time, but the ones that are most relevant for internet sales are the ones where this ratio makes sense for a customer. We will continue to update our opinions as we figure out that shipping costs go down, and we might choose different products if customers want to buy them.”

I thought that suggested they were ready for a world where people were going to shop differently than they did in retail, and I suspect Mark did really well on an investment in that company a long time ago. By the way, one of my favorite episodes, of course, was with Ken Langone. It was amazing, and the Frank Blake one, too. But figuring out what public companies are going to embrace this, and what kinds of people are going to embrace this, is a big deal.

Then I think on the private equity side, my instinct is to encourage those CEOs to get closer to what this can do and have that come back and impact their business. The reason I laugh—I was laughing a little bit when you asked—is because the shared delusion of the “how we’re using AI” slides in a board deck that everyone sees as the board member and produces as an operator is probably one of the funnier things out there right now. It’s some 20 to 30 percent reduction in some sliver of your business, just to be like, “Hey, board, don’t worry, we’re using AI,” and the board’s like, “Okay, cool. I’m glad you’re using AI.”

Nobody really thinks that’s enough, but everyone’s doing it for each other. I would say that one way to make money is that the more you’re not doing that, the more the companies you work with are either saying, “Hey, we’ve looked at it. It doesn’t work for us. Here’s why,” or, “This is a huge deal for us.”

What should board members do? This seems like potentially a really big problem. If you’re a well-paid public-company board member, maybe on some prestigious company, a lot of these roles are not hyperactive in influencing the business. They’re governance bodies that sometimes don’t do much. What is the risk and opportunity set right now for board members of these companies?

I’m a feel person on a lot of stuff. If you think about the only real job of a board beyond some very highly legal corporate-governance stuff, it’s: Do you have the right CEO? If you’re on a board right now and you just think about how you feel at each company you’re on the board of—do you have the right CEO?—you feel amazing right now if you have the right CEO. Truly, this is a place where I feel very, very, very lucky. If you have the right CEO, you’re like, “Okay, cool. They’re going to ride this wave. They’re going to figure this out. They’re going to embrace this.”

If you’re a public-company board member and you don’t have that feeling right now, I think you’ve got to examine why you don’t have that feeling and what you’re going to do about it. Ideally, what you want is for that person to be somebody who actually does want to embrace this but is nervous about something board-related. What you want to do then is appease their fears: “No, we want you to go and play. We want you to play free. We want you to go figure this out.”

I think that if you don’t have somebody who wants to embrace this—not because of board fear, but because of their capabilities or because they don’t want to play in this next chapter of the game—you’ve got to figure that out. I definitely think, as a board member, remembering that you don’t get to help in a lot of ways on a public-company board, but one or two ways you can help are by asking a good question every once in a while that makes a big difference. This is a good time to ask one of those questions and see how you feel.

In a world of accelerating change, this is something Doug Leone would always say. He would talk about accelerating change. I will add the second part: In a world of accelerating change, how do you feel about the person leading your business? You probably either feel amazing or you feel nervous.

Are there any other conversations, as you’ve processed this change, that most stand out as interesting or notable? What other conversations have you had that most impacted you in the last, say, 3 months?

There’s been a bunch. Maybe I’ll describe it more as the spirit of the conversation. This is probably the thing that I really hope comes through the most. The conversations that have been the most impactful to me all have this element of insane humility in them, which is, “Wow, there is a lot happening. I don’t totally understand it. I need to go and figure out more.”

All of them are not assertions as much as they are curiosity. They are so much like, “Oh, man, there is a lot going on. I’ve got to get closer to the metal. I’ve got to do more truth-seeking.” The conversations that have been the most interesting are all like that. They are people who are really trying to listen. They are people who are really trying to get closer, and they are all people who are sort of—I don’t know if confused is the right word, but certainly like, “Man, there is a lot going on right now.”

That humility is probably the thing that I’m most attracted to in a conversation right now about these topics: “Here’s what I know. Cool. Now I just want to figure out a bunch of stuff.” That’s why I come back to: Don’t outsource your conviction. Don’t do something because I think I believe something. Do something because you believe it, but go try to figure it out.

I think that the people that I respect the most—I heard one time that the Collisons, I don’t remember who said this, said they listen with predatory curiosity. Think about that language: “I really need to know what you’re saying and why you’re saying it.” I think right now the people who are going to do great things are really willing to change, and are very curious and very humble.

I love this story Shane Battier told you about his career training when he was young and this idea of the ghost. Can you relay that? Because I do think it’s, as an analogy, an interesting way to think about what’s happening here.

Shane Battier, a former professional basketball player and a dear friend, is very different from most people in that, when he was growing up, he knew he wanted to be an NBA player and was working accordingly from a very young age. A lot of kids say that, but they’re not serious about it. I think Shane describes himself like this: “I was a professional basketball player from the time I was 10 years old. I went and did these things. I was working. I was working with a goal in mind.”

So he tells this story that there was a poster in his room. The poster in his room basically said, “Somewhere, somebody is out there working while you’re not. Somebody somewhere is out there practicing while you’re not.” Shane would talk about how that haunted him. There was this ghost that was out there, and it would make him work harder because it was like, “Gosh, if I take a break, there’s somebody out there who’s doing it.”

He tells this story, and of course it ends well because he plays in the NBA, ends up being amazing, and has this 13-year career. It’s about the value of hard work. But the thing that’s so interesting is that Shane met the ghost. When he played in the NBA, Kobe Bryant was that ghost.

Kobe Bryant was working all the time. Kobe Bryant was born a month after Shane. It’s crazy. He grew up in Italy. He wasn’t around the corner in Michigan. There literally was someone on the other side of the world working all the time, and then Shane comes face-to-face with this guy, who was out-of-this-world good. That was how Shane ultimately had to measure himself.

As we think about this moment and what’s happening, there is someone out there working right now to produce the magic that you promise to your customers. There is someone out there working on it, and they are working on it a lot. You’re going to come face-to-face with them because your customer is going to decide at some point whether they want to work with you or with them.

You’ve got to approach that with joy. You’ve got to approach it like, “I’m going to come face-to-face with them. I’m going to go earn the right to play against this person, and I’m going to be amazing, and I’m going to go and take this opportunity because I’m given this chance.” But that reminded me so much of this moment for a CEO or a leader or anything. The ghost is out there and you get a chance to play against them, but you’ve got to embrace it, and you’ve got to go do it because the ghost is good.

I think it’s helpful to think about what’s happening from both directions, through both the lens of opportunity and the lens of risk. If you think about the title of your post, “AI or Die,” I think one represents opportunity and the other represents risk, even the story about Shane. Two sides of an interesting coin.

Maybe in closing, talk about the optimistic side of this. I know that this is your primary takeaway: that this is a tremendous opportunity for the right kind of people, and that it will lead to lots of fascinating things, products, companies, and careers. Maybe just try to put a bow on this for us. You’ve thought about this a lot now. You’re talking to these founders every day.

You're working with the investors that get this the most. What's your summary of this entire exploration?

The post starts with this quote from Ayrton Senna, which is that this is a moment where you can actually pass a bunch of cars in front of you. I think the reason that creates optimism for me is that you are only limited at this moment by the scale of your ambition. This is something Roelof asks a lot of founders: What is the scale of your ambition? At this moment, you can go and pass 15 cars in front of you.

Remember, you get to decide what race you're in. If your company's worth $10 billion, this is a chance for you to be worth $100 billion. If your company's worth $3 trillion, it's a chance for you to be worth $30 trillion. If your company's just getting started, you have a chance to beat people that you never thought you'd have a chance to beat. You get to choose the race you're in.

The reason I think there's such an incredible optimistic moment here is that there's going to be a bunch of change. None of us want to live in stasis. None of us want to be stuck where we are. All of us want to have upward movement that comes from doing something great and earning it.

This is a moment where, if you earn it—if you're willing to learn this stuff and be committed to being excellent—the world is available to you in a way that it's never been before. You don't even need to hire a lot of people. Think about all the constraints that existed before. If it's true that a 20-person company can do something that a 20-person company could never do before, it's so much easier to find 20 awesome people than it is to run a 10,000-person organization that's so far away. You can do it faster.

The thing for me is that, for high-agency, ambitious people, you've been given tools to enter a race you never knew was possible and fucking win it. That is so awesome. That is what I hope people get. Sure, if you choose stasis, you might go into irrelevance. But if you choose optimism, forward motion, and embracing this, there is no limit to what you can go and do.

That is the “what a time to be alive” thinking. That is the thing that makes you excited about your kids. Oh my gosh, go back to what Sam said: Every person in a decade will be more capable than every person now. I mean, that's awesome. Now let's go get it.

Our mutual friend Sam Teller, one of my favorite people, has this line that has always stuck with me: If you just see excellence in people at the highest level, it's a huge competitive advantage because you have this bar that you can walk around with, compare things to, make better decisions, and work with better people. It's all about having seen the highest level of excellence that you can.

I feel like my takeaway from your post is that we've actually democratized this. We've all had a ChatGPT aha-moment experience or several, a Midjourney experience, whatever. The imperative that I take from your post is that we actually should not rest until we've created one of those in our business.

We should have a moment that wows us in our own product in the same way we've been wowed by some of these incredible original products. Until then, we shouldn't rest, and then we should keep going. I just think your writing is such a cool call to action for an exciting wave of the market, the market economy, and companies. So thank you for writing it, and thanks for outlining it with us today.

Absolutely. Thank you for having me.