他打造了 Google、Facebook 和 Square 的收入引擎

The one thing I think that's going to be truly future-proof is judgment. Why? Because you have the big challenge of AI slop. Every product leader I've talked to is extremely worried that, because you have these engines running rampant, they're just going to produce lots of code.

1. The Changing Nature of Product Development

In an era when you can do everything, the question is which of these things matter and you should truly do. I thought an interesting place to start would be the changing nature of how people are building products.

The biggest story by far in technology seems to be Claude, or Claude Cowork as well. The ease with which both technical and non-technical people are able to build something that they can imagine seems to have been a complete explosion in their ability to do so.

You've built a million things. You've invested in 700 companies, watching people build things. You're about as prolific as they come as a product person. Maybe just give us your state of the union of how the world feels to you in terms of technologists building products and how fast that's changing.

What's interesting about product development is that 10 years ago, or even 5 years ago, there were very clearly defined roles. Product managers articulated what to build, designers designed it, and engineers built it.

Over the last few weeks and months, I've been talking to many companies, but over the last 2 months in particular—December 2025 and January 2026—it's become very clear that something has fundamentally changed. What has changed is the notion of a long-horizon, long-running agent.

I've experienced it myself. About 6 months ago, I tried to use Claude Code in the early days to build something. I called it a video-transcription tool. I tried to build it, but it kept failing, and then I had to go in and try to debug it. Ultimately, I gave up.

Two weeks ago, while watching some episode of some TV show, in 1 hour I was able to basically prompt my way to a good video-transcription tool because these agents are now resilient to failure. You don't have to be very technical to use them. This changes the expectations of product teams.

After I did that, I started talking to 3 kinds of companies: portfolio CEOs of companies I've invested in, the large AI labs, and a bunch of young, AI-native companies, to see what the similarities are between them.

A few things emerged. First, product development as we know it is changing because the models and capabilities are growing so fast that, if you try to be very strict and stringent about exactly describing or prescribing what you're going to build, it's not going to work.

Almost everybody has gone to a bottoms-up approach, where it's not driven by product management anymore. The only thing product managers do now is articulate what the customer needs are at the highest level, and then they are the guardians of the why.

The actual product is built bottoms-up by engineers, researchers, product managers, and designers, all working together on the code itself. Capabilities and models are changing very fast. If whatever you thought 6 months ago is what you continue thinking about, you've fallen behind.

It's very important for product managers to understand what these models are capable of and to be hands-on. They sit with the engineers and researchers and write code, do prototypes, and do anything and everything needed in a hands-on way.

The first thing we're seeing now is that PMs are starting to check in code with either Codex, Claude Code, or whatever into the actual production repository. Right now, engineers have to review the code, but you're soon going to see Claude Code, Codex, and other tools actually review the code itself before engineers commit.

All the companies are struggling with how to evaluate these people. Earlier, there was nothing called the prototyping interview. Now there's an explicit interview in the interview loop called prototyping that literally forces product managers to be hands-on.

2. The Merger of Product and Design

Second, the product manager and designer roles are merging increasingly. The designer role is an interesting one in particular. A lot of companies are going through headcount allocation this year, and I'm hearing from many teams that, when given the choice between an extra designer and an extra engineer, they're saying, "You know what? The design systems are already laid out."

Now that we have the design system already laid out, we can use AI to do work around these design systems. We need maybe a small number of designers at the company level to manage the design systems and the design language, but AI can leverage the design language to do designs, so please give us an extra engineer.

The ratio of designers and product managers to engineers is growing in product. It used to be 1 to 3 or 1 to 10. It's going to 1 to 20 now.

3. Managing Non-Deterministic Software

The other very important thing that's happened, which is fundamentally different, is that when I was growing up, products were deterministic. There was a workflow: you knew if X happened and a user did X, Y happened. It was very clear—when you did X, Y happened.

Today, you could do X and Y happens, but if you do a slight variation of X, something completely different happens. It's non-deterministic software. What that means is that, on the other side, you have to have an evaluation, or what's called evals in AI, and someone has to evaluate whether or not what the software is producing is reasonable across various use cases.

Obviously, they can be human evals, AI evals, and so on. But who owns the evals? It's the PMs. It's the PMs and the researchers. The PM's job is to be very clear at a high level about what the user needs are and then have a very clear sense of whether this product is good to ship or not by evaluating it.

To evaluate it, many times you've got to write AI yourself to evaluate the results of AI, because humans can't. PMs are very good at coming up with evaluation techniques.

So it's the non-determinism of software, the speed at which things are going, and the overall notion that the capability frontier is being pushed out every 2 months. It makes it an incredibly challenging yet incredibly exciting time for product development.

If you think about it, my friend Zach has this great way of thinking about AI: we had the Industrial Revolution for goods, and that basically kicks off an Industrial Revolution for services.

This is an interesting opportunity to ask about what your philosophy of product is. You're such a product-centric person and builder—that's what you've done, and that's what you've invested in.

As we face down this Industrial Revolution for services, what's your broadest possible philosophy of product as we enter this era?

Very simple. A product person, or product manager if you call them that, has a job to balance customer needs and business needs.

There has to be somebody at the company who's the keeper of the why. Why are we building it? What customer need are we solving? Why is this a pain point? How intense is it? How deep is it?

Second, how does this add value to the company? If you build this thing, solving this customer need, how does that add value to the company? Balancing those 2 is a very delicate act.

You can build something amazing that adds a tremendous amount of value to the customer but doesn't add any value to the business. You can do something that's awesome for the business, like raising prices, but is value-detracting for the customer.

Balancing customer needs and business needs at the highest level is what I think of as product. What it comes down to, in my opinion, over the last 10 or 15 years, is this notion of outcomes.

Outcomes, I think, are what define the best product people, and outcomes have to be defined in the form of customer behavior. I strongly believe that because customer behaviors are leading indicators for every business outcome.

If you think about it, the simplest thing that a product does is make somebody go from a not-a-customer state to becoming a customer, and from becoming a customer to becoming a loyal customer, and then maybe from becoming a loyal customer to becoming a paying customer.

There are all these different product states. If you do a poor job, they can go from becoming a loyal customer to becoming a churned customer. These are all behaviors. Everything you do or build should be attuned to the goal of what customer state change it leads to and what customer behavior change it leads to.

I tell every CEO I meet who's trying to hire their first PM or doing their first product review that they need to ask why. The only question you need to ask is why. You should not let any feature go out if there's not a clear hypothesis behind it.

The hypothesis has to be articulated in the form of a customer behavior change. We believe that, by launching this thing, customers will go from doing X to doing Y, or from spending X minutes a month doing this to Y minutes a month doing this.

You have to have a hypothesis grounded in some data, something you know about the customer, or some secret about the customer.

You mentioned at the start the difference between the video-transcription tool 6 months ago versus more recently and how quickly that changed. It's just such a hard future to reason about, given the pace of change.

How do you reason about it? Is there anything that can be truly future-proof?

4. Judgment: The Future-Proof Human Skill

Yes. The one thing I think that's going to be truly future-proof is judgment. Why? Because what is the biggest challenge you have when you have 1,000 AI engineers writing code? You have the big challenge of AI slop.

Every product leader I've talked to is extremely worried that, because you have these engines running rampant, they're just going to produce lots of code.

Which of this code is even valuable? Which of these things are even valuable? In an era when you can do everything, the question is which of these things matter and which you should truly do. On the product side, that is judgment around what needs to be built and evaluating the output. On the engineering side, it is evaluating the code, because if you don't understand what the code says, I think you can have AI engineers writing beautiful code that could be wrong, could have bugs, or could be vulnerable.

Someone needs to review it and make sure—you have to have human review at some point, especially critical code that is in the core of your system. Similarly, in design, you have to have judgment around, does this make sense? Does it make sense in the broader design system? I think judgment is the number one thing that humans are going to bring in an era of infinite productivity. The question is, what are the things to be productive on, and are we building the right things?

5. Building Durable AI Applications

As you evaluate companies today, build things yourself, and just think about this problem and the trajectory of these tools, maybe walk through how someone should think about building an AI application. If so many people are excited about it, it feels like a gold rush with this new technology. There are so many things that we can do that we couldn't do before, or things that specific people couldn't do because they weren't technical that they can now do. How should people think about attacking building something new—an application using AI—starting today?

First and foremost, it has to be a deep and compelling problem. The good news is there's a tremendous number of deep and compelling problems today in every vertical, in every industry. Why? Because until recently, software was used more as a tool by people, by humans. We finally have software that is agentic in nature, which means it can do the job of people.

So the question you have to ask is: In what industry are there roles of people that are highly paid, doing somewhat of a repetitive job, that can be done by software? Every 3 months, the answer gets deeper and deeper. You couldn't have told me that a designer's job could be automated by AI 6 or 9 months ago. You couldn't have told me that an architect's job could be automated by AI, or a lawyer's job. It turns out increasingly, in every vertical, these capabilities are getting better and better. So you want to start with, first and foremost, what industry do you want to be in and what kind of job do you want to do?

Second, you want to target a high-value workflow. You want to target a workflow, a way of working, that is deep, that is complex, and that requires custom data. I met with the CIO of a Fortune 500 company a few weeks ago. I think one of the challenges with this whole space is that the models are becoming so good that if you try to build a company that is light, that is not a hard problem, the foundation-model companies are going to eat you.

This CIO that I met at this company said—I was asking him about a few startups I had invested in and worked with—he said, "Look, I don't know why I would use any of these startups. Gemini has an Agent Builder product, and I also use ChatGPT Enterprise, and they also have an Agent Builder product, and I have 1,000 IT engineers who work for me."

They all want to be retrained as AI engineers.

So I'm just going to put them, using these horizontal tools, to build my AI agents. Why do you need any startups?

That's the kind of thing you're going to face: If the CIO of a company that is your target customer can build what you're building with these agent-building tools, you're not going to be successful. So you've got to really go one step ahead of what can be built, multiple steps ahead, and you've got to extrapolate to where the capabilities of these agent-building products can go. You've got to do something very, very different.

What that means is you've got to have durability, because ultimately, as venture capitalists or even as an entrepreneur, your time horizon can't be to build something that lasts for 1 year. That's the biggest challenge. It's not building an application; it's building an application that's durable, that basically will stand the test of time.

I think there are a few things around durability. One, you need to have ownership of a scarce asset. A scarce asset could be a license of some kind. It could be a regulation of some kind where you have unique insight into it. Second, you might need to own a control point. A control point is a thing that controls how people interact with money or with data, so you want to own that.

Third, you might want to have hardware, which is hard to replace. Fourth, maybe you want to be part of an essential workflow. Fifth, you want to have network effects. You want to think about those things and figure out how, after you take on that workflow, you can make it more durable.

Finally, I think your ambition has to be to replace the entire system. In other words, increasingly, what is going to happen—and I'm seeing this more and more—is every vertical has either a legacy or somewhat new system of record, which is a system where most of the data is stored for that system. For example, in legal, there's a company called Filevine or another company called Clio. There are a few of these companies. In sales, there's Salesforce. In healthcare, it's Epic.

Now, for many years, these companies all had APIs that, if you entered that industry, you could build an agent company on top of these APIs. In 2025, things changed. These companies started seeing that these agent companies, AI companies that are being built, are starting to take on the functionality of these companies and are treating them like a dumb database.

So you started seeing last year that these companies are cutting off access to APIs. Slack has done it most publicly. Slack is owned by Salesforce. They cut off access to Glean, where Glean can no longer access Slack data. The reason is they don't want Glean to build on top of them and then slowly suck out the value that Slack has.

I'm hearing from other verticals that they're doing one of 3 things. They're blocking access to APIs, they're offering their own agents for free bundled with the product, or they're charging these AI agent companies to access the data. Just to access data—the API was free. They're saying now it's like $2 an API call or something like that. So they're basically trying to make the model of these agent companies unviable.

I think it's going to be very hard for an end customer to use multiple companies where you have a system of record and then you have this agent that sometimes doesn't work with it properly. So the agent companies have no option but to also start building and offering a system of record. Every company I know is now trying to figure out, how do I build the entire platform and not just a system that does some workflows?

I think last year everyone was like, "Oh, we can do workflows. We can build what is called the system of action and live on top of the system of record." I don't think that's an option anymore.

The Slack example is a good one of a last-generation software company that was very big and very successful. One of the most interesting investor questions—and I'm curious for your answer from the perspective of a builder and a technologist—is the degree to which these horizontal model companies are going to destroy or be very bad for old software companies.

6. The Risk to Legacy Software Companies

Over time, it will be trivial to spin up your own Slack that has features that you want for your company and is very reliable in all the same ways that Slack is, and therefore Slack's in a lot of trouble. How do you think about that question? Obviously, public markets seem to think software is in a lot of trouble. The multiples are really, really low. How much would you be worried if you ran a good, solid but older software company today?

There are 2 or 3 kinds of software companies. I think the software companies that should be the most worried right now are those where they are pricing the product based on utility. Zendesk is a good example, where literally Zendesk prices seats and each seat comes with utility. In other words, each seat corresponds to a customer service agent that tackles a certain number of customer tickets.

That company should be worried. Why? Because I can have an AI agent sit right next to Zendesk, and you can slowly siphon off the utility. Instead of paying for 50 Zendesk seats, you can pay for 20, and I can have 30 AI agents sitting next to Zendesk, and that siphoning can happen over time. You don't have to have an all-in-one decision; it can be a two-way-door decision. Those are the most endangered companies, in my opinion.

You need to change your pricing model to be based on outcome, and you need to actually build the product to be based on outcome. It's easier said than done, because literally you're going from $20 or $30 per seat to maybe charging a buck or 50 cents or 20 cents per ticket result, and you don't know how that's going to turn out. So you've got to change your pricing model, and I think that's a very challenging thing. That's why I think many of them probably need to go private, because they have to make this business-model transformation in private. I think it's going to be hard for them to stay public.

The companies that are less exposed are ones where the utility is not based on seats but on data that has been collected and captured over a period of time. The more timeless the data is, the more protected they are. Slack, for example, I would say might be in a little bit more precarious a state, because the half-life of the data in Slack is very short.

That's a great way of putting it.

But if you have an ERP, NetSuite is a great example. Somebody uses NetSuite as an ERP. Now, I don't know how NetSuite actually charges, but it doesn't matter. However many seats you buy, the reality is it runs your whole business, and there is no compelling reason for someone to put their career at stake by ripping out NetSuite.

I know there have been a lot of AI-enabled ERP businesses over the last year, but there's no compelling reason to take NetSuite and say, "I'm going to rip it out," because it is career-limiting to suddenly take NetSuite out when you're a company running on NetSuite.

So I think those companies are much more insulated. You could argue that NetSuite has more time to build AI agents on top of it because they have the data, and they can train the AI agent on top of it and bundle it.

I think the software public markets do not distinguish between these 2 types of companies: companies where the half-life of data is low, and where you can literally take half of the value of the company and put it onto an AI company that sits next to it; and something like an ERP system, or even Salesforce for sales data and records, where those are real customer records. It's going to be hard.

So what are AI-native companies doing? The first thing you've got to do if you ever have to compete against them is spend 1 or 2 years building a system that literally migrates your Salesforce instance to your own company's platform. One of my companies is an AI-native company. They literally hired engineers in an Eastern European country for 2 years to build this migration, this transition tool.

You have to build the migration tool, because who's going to migrate it? You can just present your spanking-new system, but the data is still there. Even for Square, for a small business, I remember they had a point-of-sale system. They wouldn't move to us even though it was cheaper because they had gift cards, customer data, loyalty data, payments data, all of that—even credit cards.

So we had to build scripts, and that took us months or years to build for a simple POS. For something like Salesforce, you can't just say, “Well, here I am. I'm a much better CRM because I connect.”

There is this thesis, which I completely agree with: If you look at CRM, what does a CRM contain? It contains your customer record. Your customer support system contains what your customers are complaining about, and Jira or Atlassian contains what your product development team is building.

All of these things should be linked, right? Because there is no linkage, you should be building—the biggest—you should be addressing the biggest complaints of your customers, which are in Zendesk. For those Zendesk customers, you should know where they came from, who bought them, who sold them, and what the AM is.

7. Sources of Stickiness in the Age of AI

All 3 of these systems should be linked together, but they're all 3 different companies. There are companies that are trying to unify these things, and it's a great value proposition. But guess what? None of your customers is ever going to move unless you build a simple, seamless way to take the Salesforce data and move it to your instance, the data from Jira to your instance, and the Zendesk data to your instance. So it's literally a 2-year effort to build migration. Otherwise, you've got to get Accenture.

How do you think about stickiness in this era, just as a general concept? When the friction for creators to build something net new is so easy, is so low—you can do whatever you want really fast—how's anyone going to use anything for a long period of time?

The age of AI stickiness, I think, comes from a few sources. I think you need to have network effects. DoorDash is sticky not just because it has this beautiful app, but because it's a network of restaurants, dashers, and consumers. So you can't just attack one.

You can't vibe-code your way to those 2.

Exactly. And so, network effects.

The second example of stickiness is when you have financial services or money moving through you. I think many of the systems of record—for example, Toast—have payments going through them, and I think that really is interesting because you can't just start building the point of sale. You also have to have money flowing through it.

If you look at banks, they're a good example. Once you have something like Mercury as a business bank, it is hard to switch because you have money flowing through it, and you have regulations and other stuff embedded. I like things that are a combination of financial services and software because of that.

The third stickiness is from hardware. You can actually have hardware. Toast is a good example: Toast gives you hardware for free, but if you try to return the hardware, you have to pay them. But in either case, the hardware is there, and somebody can't just build software. They also have to take the hardware, put it into the thing, and rip out the Toast hardware.

The fourth one is access to a unique asset. I was thinking about a good example, and I came up with the example of Sierra, where I think the unique asset is Brett Taylor. They have full control of Bret, who's one of the best salespeople and the chairman of OpenAI. He can make a call to any company, any country, and they'll take his call. You can't really outsell Bret, and so I think there's alpha in that.

I think you need one of these 4 or 5 things, which are basically indicators of durability. The half-life of software today is so short that you need to be one of these things to make it durable. Harrison Helmer has this thing called 7 Powers, and so you've got to have a few of those 7 Powers embedded in the business model from day 1.

8. Leadership Lessons from Google

You've been so lucky to work for some of the most well-known CEOs and founders of this modern era. I'd love the chance to ask you a little bit about each of them and what you learned from them, and then, more generally, just things you've learned about what great leaders do to run companies.

Maybe going all the way back to Google and starting with Larry—Larry and Sergey—what did you learn from watching them operate and lead?

Yeah. One of the most interesting things about all the leaders that I've worked with, who I think have built generational companies, is that they have a superpower that is very aligned with what the company needs to succeed. The company was really shaped in their image: the company, the culture, the early hires, the products.

When I joined Google in January 2003, the first product I got exposed to, which I didn't know about, was a product called Caribou. Caribou was an internal code name for a product that was launched on April 1, 2003. Publicly, it was called Gmail.

I didn't believe that this product existed because, in the internal alpha, it said this gave you 1 GB of storage. Back then, Yahoo Mail was the dominant product, and it gave 10 MB of storage. So this thing had 100 times more storage.

This really epitomized Larry and Sergey's philosophy, which was basically: Build the best technology on the planet. They were deeply technical, and every product was held to a standard of technology and scale.

I'll never forget that AdSense was the fastest-growing product in Google's history, and we went into reviews and Larry would be disappointed in us. We asked why. He said, “What percentage of all ads on the internet are you serving?” We'd say, “Less than 1%.”

His goal was not—again, he didn't care about the revenue. He cared that Google was involved in serving every single ad on the planet, versus making a business of whatever—$1 billion, $2 billion, or $10 billion.

The focus on scale and the focus on technological superiority, and that investment—Google Street View, TPUs, Waymo—all of these, I think, show 10-plus years of investment in an uncertain future, but knowing that if you invest in technology, good things are going to happen. And good things happened, but it took a decade. That's investing in technology capabilities.

Before we leave Google, you had this interesting idea about communication and Eric Schmidt, obviously another key Google person. Can you tell the story about him presenting the company strategy using nothing but images? This is an interesting example of communication.

Yeah. Eric was incredible. One of the interesting things I've seen is that almost every great founder or founding team needs an Eric figure. If you look at it, Mark Zuckerberg had Cheryl Sandberg, Jack Dorsey had Keith Rabois, and Tony Xu at DoorDash had Christopher Payne.

Everyone had somebody who was complementary to them and really helped. They were amazing at, say, technology and scale. Eric was amazing at bringing a team together and leading, and I think Larry and Sergey learned a lot from him. Larry, of course, became CEO after Eric stepped down, but Eric was incredible.

Eric would get a product leader who would become seconded to him for the weekly strategy or the annual strategy-planning session. I did it, I think, in 2007. My job was to go to Eric and say, “Eric, how do you want to present the strategy of the company?”

He's like, “Well, it's very simple. I want you to go and interview each of the different leaders of the different teams. There's only 1 constraint I have.”

I'm like, “What is that?”

“You can't use any words to describe what they're doing.”

I'm like, “What do you mean? You have to use words.”

“Nope. You've got to use only images.”

I'm like, “Why is that?”

He's like, “People don't remember words. They remember how things made them feel. You can put words in the speaker notes I'll use, but I want you to come up with the most compelling image that exists for what they're describing.”

9. Learning from Mark Zuckerberg

It was a crazy thing because I never thought of doing a presentation that way. I went to each of the businesses—AdWords, Search, YouTube, AdSense—and then had to come up with a compelling image that was easily accessible to the whole company, yet represented what they did.

Do you remember a specific image? I'm so interested in this exercise.

It seems like a potentially productive exercise for anyone to try to jam what they're trying to say into only images. I'm trying to pin down an image and how you arrived at it, or—

I think for YouTube, it was a graph. It showed the number of videos being uploaded every second and how that had changed from the time Google bought them. So it wasn't even a graph; it was literally showing this incredible hockey stick that happened over the last 18 months. I think we couldn't even show the numbers. The thing had to be compelling enough that the line would just have to be like a U or something like that when it went like that, because we just showed it like this. You couldn't say something like “100x” or something; you couldn't say that.

So we had to show that. The one thing we wanted to show was that Google Search had gone from being used by small and midsize companies to being used by the largest companies on the planet. We showed the logo of—I think they had a very large Fortune 50 company that they had acquired.

What did you learn from Zuck?

Zuck was, and is, actually, I think, the greatest mind on building growth and engagement in consumer products broadly. I've seen him basically sit in a room and critique something a product team would have come in with—a very well-thought-out consumer product flow—and he would look at the flows and say, “That is not going to be compelling to users. That is not something that a user is going to engage with. Change it to this.” You'd say, “My God, why didn't I see that before?” So he's very, very good at thinking about how consumer products should be designed to maximize engagement and maximize growth. Both is probably the best way to put it.

The second thing he's amazing at is learning by following. When I joined, my task was to lead the ads product team, and Zuck at that point knew a little bit about ads because he had worked with Sheryl quite closely. Sheryl had worked on ads before. But then, within about a year, he shadowed us. He came to the ads team, he basically sat with us, and he came to many of our meetings. Within a year, he got to the point where he was generating ideas for the ads team.

10. Jack Dorsey and the Philosophy of Great Design

One of the most foundational ideas of Facebook ads came from what is called Custom Audiences. Custom Audiences is the foundation of most ad systems now. It is the idea that, as an advertiser, you want to reach people who are similar to your customers. So if you're a bank and you have, say, 100,000 customers, how can you give this set of customers to your ad platform and say, “Look, instead of describing these customers”—what did they do before? They would describe their customers: “I think they are 25- to 34-year-old women.” That's not good enough. Instead, if you can just tell us who your customers are and we can map them to our users, we can then find people similar to them.

So uploading that data into our system securely and doing it in a way that doesn't compromise any PII was the key thing, and it all came from Zuck. How? Because Mark Pincus was the CEO of Zynga. Zynga was the largest advertiser on Facebook. Zynga, like most gaming companies, was very focused on acquiring whales.

Because whales, for any gaming company—casinos, et cetera—make up 80% of all revenue. Probably for all the betting companies, 80% of all revenue comes from whales. So he was very frustrated with us. We would do these quarterly reviews with Zynga on the ad side because they were large spenders on ads, and they would constantly be yelling at us, saying, “We want to get more whales.” We were like, “Yeah, you're getting users, and it's your idea. You need to figure out how to get whales from your games. What do you want us to do? We can help you acquire users.”

So he once, I think, talked to Zuck, and Zuck came to us and said, “Why can't they just upload their whales into our system? We know who the whales are. Why can't we just find people similar to those whales?” We were like, “That's interesting, but we actually didn't know who the whales were.” So they needed to tag for us who the whales were, and basically we started doing it. Similarly, we started finding users similar to the whales they had, and it worked so well.

Then we said, “Why don't we take this approach and use it for other types of customers who we didn't have data on?” It became truly—it was a transformative thing for ads, and it was all Zuck's idea. He just has something about connecting, making connections between disparate domains, which is pretty amazing and unique.

Jack is—I mean, I think he's on par with Jony Ive and Steve Jobs in terms of his thinking about design. He understood what good design means. Good design doesn't mean visually pleasing. It means a product that is designed so well that you don't have to give your customers a manual on how to use it. They should be able to see the product and use it.

Think about your point of sale. Every point of sale except Square, and things that have copied Square, still requires you to train a barista for several days after they join on how to use the point of sale. Square is something you can download from the App Store and start using as a point of sale to run your business—a category where you have to train somebody for weeks. That's an example of good design.

He brought that to every part of the company, removing friction from what is traditionally—I mean, Square's whole premise was removing friction from small businesses applying for financial services. That extended to the product. It also extended to risk. One of the most interesting things that I didn't realize is that Square, at its core, is a risk company.

When you apply to a bank for payment processing—in fact, the company was founded because Jack's co-founder, Jim, was rejected many, many times when he tried to accept Amex by banks. He was a fairly successful glassblower in St. Louis, and he basically was selling $2,000–$3,000 glass sculptures to people who would send him checks. A woman called from Panama one day and said, “I want to buy this on his website.” He had this beautiful piece of glass. They agreed on the price, and she said, “Can you take my credit card number?” So he said, “I don't accept credit cards.” She said, “Sorry, I can't send you a traveler's check or check or whatever the case is.” So he lost the sale.

He went to his friend Jack Dorsey. They had never built hardware; they had never done any of that stuff. But they brainstormed and realized that the iPhone, which had just been released a couple of years before, had this thing called the audio jack that basically could be used to put a piece of hardware in and process cards.

I can't even imagine the leaps you have to make to get there.

But the number-one thing that they realized is that most small businesses are denied by banks when they apply. Square instead said, “We are going to accept 95%.” But what they did was put risk at the transaction level. So they accepted you as a person, as a business, but then once you started processing transactions, they would run machine-learning models, and on every transaction: “This transaction is risky; this is not.”

It shifted the level.

And so that kind of lazy but brilliant onboarding is something that characterizes a lot of good thinkers. Sergey was very similar. I've come to this conclusion: when we were going to launch AdSense in 2003—I’ll never forget this—May 2003 was when we were doing our final launch things. Sergey was our sponsor. He came and sat in the meetings and said, “What are you guys building here?”

11. The Product Manager as Editor

We were like, “Website publishers are going to apply from all across the world. It's a self-serve product. We have to review them and say whether we should approve them or not to run AdSense.” He's like, “Why do you need to approve them?” We were like, “What do you mean? Our ads are going to be running on these things—Google ads, or ads powered by Google. You don't want to be on a porn site or something else.” He's like, “Why not?”

We didn't really have a good answer for why not. I was like, “Well, standards or policies.” “Okay, but what if they lie?” He was right. What if they lie? We had so many people applying with Nike.com, for example. It's true. It was very hard to know who owned a domain, right? I could apply with your domain and basically get accepted. He was right in some ways. We were just doing it to cover our asses, it turns out.

And so he said, “Okay, kill all this.” We had literally spent half of our engineering team building this complex approval system with ops and so on. Ops were super excited. They had hired a lot of people, and now you're telling us not to do it, and instead to do it in real time for every page that loads because we had the JavaScript on it. We know what URL it is. Look at the content at that point.

We were like, “It's too slow. We won't be able to look at the content because it's billions of pages.” “That's fine. Let it load 100 times, and after 100 impressions, if any URL hits 100 impressions, then start reviewing it.”

You're not trying to put lots of checks up front.

But we're being intentional about where and why. Most things don't even get to the level where you care about, so only do stuff when you need to. The same thing happened with click fraud. Click fraud was one of the biggest challenges that we faced, where people click on their own ads and make money. How the hell do you solve that? The reality is, you don't.

You just wait, and you start understanding what click fraud is, and then you solve it. So be reactive and solve it when it needs to be solved at that point versus waiting. The Square thing was exactly moving risk from the business level to a transaction level. The same with AdSense: move risk from the publisher level. You're basically—you cannot gate, because getting somebody to come to you and sign up is one of the rarest things in history.

Someone is coming to you and expressing an interest, and you're saying you're going to put 10 different barriers in place. That's the opposite of self-serve.

So a pure self-serve product would never have any reviews of any kind. You're going to be immediately activated. Go on, and we'll do checks in real time based on what you're doing versus banning you or stopping you.

You gave me both these amazing examples, and then you also said that Jack would do this across the company, not just in the product. How would you sum up the process of great design that you've observed from the people who are the best at design? What is the thing—the method—that they're going through over and over again as they apply it to different parts of the company or product?

The number one thing I've seen is that they try to minimize the number of steps. Everything should be on 1 page, and you need to cut down things. In fact, Jack called the product manager role “product editor.” Why? Because he believed, rightly so, that the role of the product manager is not to add more features. Any of us can look at a product and say, “Here's 10 things you should build.”

The best designers, the best product people, edit down things. Similarly, if we have 100 features, what are the 2 things that really matter that will drive the customer outcome? The best designers really take 10 pages of design and say, “Cut out all the excess.” So I think it's the process of editing, and this goes to judgment.

I think in an AI age, humans with amazing judgment, which is really editorial capability, are the ones that are going to do well and thrive.

Apparently, Rick Rubin would say that he wasn't a producer; he was a reducer.

Great example: reducer. I like that.

I wonder how that applies also to communication. Maybe this is a fun opportunity to ask you about the format that you've landed on that a leader can send to his team on a weekly basis. I think it seems like this idea of reducing and simplifying can be applied in so many ways by great leaders. Talk about it in terms of communication from leadership to a team.

One of the things that people, especially founders of startups, don't realize is that initially most startups start with 2 or 3 people, and then they go to a point where they're all sitting in a room together. Everyone can hear what you're saying. But as soon as a company goes into what I think I call “2 rooms,” where they're not in the same room together, then you have to communicate. You have to let people know what's going on. You have to bring everyone together.

There are a few artifacts that companies need to start putting into place. One is the notion of an all-hands. An all-hands seems clichéd and unnecessary, but even with a 15- or 20-person company, just getting together once a week, maybe on a Friday, Monday, or Thursday, and basically sharing what people have built and been working on, and then having the leader—or one of the leaders—address everyone is a great way to get people together.

12. Three Pillars of a Successful Ads Business

The second thing is a weekly CEO email. I think this is a very powerful way for the CEO to get across to the team what is on their mind. The best way, I think, and what I've done myself, is during the course of the week, you start jotting down things that you think you want to communicate. Then you'd spend Saturday or Sunday, whatever the case may be, taking all of those things and reducing them to 2 or 3 things that matter that you want to get across.

Most businesses, I think, can be communicated along 3 dimensions: progress—product, business, and team. What's happening on the product side? How is it becoming more remarkable or serving our customers better? What's happening on the business side? How are we doing better as a business? And then what's happening on the team front? Who have we added or subtracted? What changes have we made?

Most importantly, don't be afraid of repetition. Don't be afraid of repetition, because repeating it once, twice, thrice, or 4 times is when people actually—when it seeps into their bones.

What's the literal format that you do? So you've got 3 sections in your email. What is the structure that you do personally?

The format I've used in the past, and what I recommend, and what I've seen at least 15 CEOs adopt to good effect, is 3 sections. One is called “top of mind.” So this is product, business, and team. What is top of mind on the product side, on the business side, and on the team side? It doesn't need to be all 3. What's top of mind for you? What's keeping you up at night?

I think this is the thing that literally everyone is hanging on to, because I remember seeing it from Jack, from Mark, and from Sheryl. I think just seeing it put on paper or put in an email is so powerful. That's one.

The second thing is a performance update. I think everyone wants to truly understand how the company is doing—how the company is doing on the dimensions, I think. And this is where, especially being a startup, most people are 1 dimension removed from how the company is doing. They all want to know that they're doing well, and I think this is the way.

And the third is miscellaneous. It's things like recognizing specific people, quotes from customers, or maybe an off-site announcement. But the most important section, where you should spend 60% or 70% of your time, is top of mind.

How transparent should one be in that? As a leader of a business, I could tell you what's top of mind, but a lot of it either might be sensitive, or I would worry about scaring people or worrying people about something that I'm thinking about or worrying about. What keeps me up at night might create stress in the business. Where should one draw the line in terms of how candid they are about that?

I personally think more candid is better than less. Why? If you're more candid, what you can do is actually get people—you can actually ask people to suggest ideas. And that's the thing: if you have good talent at the company, if you actually ask them, “What do you think I should do? What do you think we should do in this situation?” I think people will rise to the occasion, especially when the company is small.

We want people to have more input, and if there's a one-way-door decision that we're going to make, where making it takes us one way or the other, I think it'd be great to get feedback from more people.

I want to talk about ads and everything you've learned about building an incredible ads product. You've basically built the core business—the important core business engine—at multiple places across your career.

As a company, you either die or you live long enough to become an ads company. And so we're seeing it now with OpenAI. It's happening.

Now, how do you build an ads business? There are 3 fundamental ways to succeed in the ads business—3 and only 3. One, you need to own a very coveted group of users, and you need to have a surface on which those users interact. Google Search is a great example. It's a surface on which a very coveted set of users interact with. Obviously, they express high intent.

So Google is one of the most profitable ad businesses. Facebook is very similar. It took us a while to figure out what was common to both these users. Turns out what was coveted was the identity. We knew who these users were, and we could match them to customer and other data. So you could precisely target these people with the messages you wanted, and you could find people similar to them.

ChatGPT—their combination of intent and identity data is unparalleled. Google had intent data but not identity. Facebook had identity but not intent. Having both of these things together is the dream of any advertising person.

I don't know how many searches they see, but they're going to see more. And these are complex, multi-phase searches, right? That's the other beautiful thing. You search, and then you search again, and you're just building up searches. At Google, you typically search and then you lose the person because they go off and click, and you don't hear from them. These are natural-language queries, ripe for amazing targeting.

That's one way of making money. But you have to own a first-party product. You have to be the first party.

Second, you have to drive outcomes. That's another way of making money, where you don't own any inventory but you can drive outcomes for advertisers. The best example of this is a company called AppLovin. AppLovin is a 100-plus-billion-dollar company. They drive one outcome really well: mobile app installs.

And no one believed that people would need that many mobile app installs. Turns out everyone wants to get mobile app installs. It was initially restricted only to gaming.

But now it’s every mobile app where they sell one mobile app install. So AppLovin has built a massive infrastructure. Now they control the buy side, they control the sell side, and they even control the middleware.

You could argue that they kind of control the auction for most mobile apps in a way that almost mirrors how Google used to control—or people say they control—the web. But AppLovin has built an amazing engine to deliver mobile app installs at a certain cost.

So that’s the other way—the second way to do it. You deliver an outcome at a certain cost. The third way to do it is if you are the exclusive provider for a large advertiser or a large source of demand. A good example is a company called The Trade Desk, where Procter & Gamble, for example, goes to The Trade Desk and says, “I spend with Google, I spend with Facebook, and here’s all my other display budget. Trade Desk, here you go. You can figure out how to distribute it and how to run it.” And so those are the 3 ways, but you’ve got to be exclusive. Those are the 3 ways that you can make money.

What business ideas don’t work in advertising? What are the business models that are just doomed to fail?

Trying to be a middleman on top of these large platforms, from my understanding, doesn’t work. The Trade Desk, I know, doesn’t work on Google or Facebook at all. It doesn’t work with Google or Facebook as a first party, but AppLovin, I think, works only a little bit on Google and Facebook. Mostly, they do their stuff on the unwashed web, basically outside.

You’ve got to stay out of Google and Facebook’s ecosystems, because if you’re trying to build your business on top of Google and Facebook—or probably soon OpenAI—as an ad company, you’re going to get squeezed over time. Every time you build a new capability on top of Google, it turns out Google learns what you’re building, and Google has the best engineers on the planet. So does Facebook. They will take your capabilities and incorporate them into their platforms.

There is almost certainly going to be a cottage industry of companies that come and say, “I’m going to help you optimize ads in ChatGPT.” There are already companies that help you optimize placement in what are called answer engines, or AEO, instead of SEO. All of those are not going to create durable, enduring companies.

What would you be worried about if you were one of these fairly monopolistic owners of a massive ad network, like the ones we’ve discussed? There’s Uber and Amazon in the mix, DoorDash, Facebook, and Google. If you were running their ads businesses, what would scare you?

Consumer behavior change—where people don’t open up the apps anymore, but use agentic interfaces. They use AI interfaces that are not owned by my company, this company, to do their transactions. If you assume that a big percentage of things are repeat, then could you put those repeat things on autopilot through an agent? You never open the app, so you lose opportunities to advertise, and you lose the relationship with the customer over time because customers start trusting the AI agent.

13. Selecting North Star and Check Metrics

You can’t bury your head in the sand. You have to go and experiment. That’s why, when ChatGPT opened up its apps platform, all of the commerce platforms started experimenting. The thing I would look for very carefully is the early adopters using the app. Obviously, they’re going to connect their Uber account with their ChatGPT account. I’m going to look to see, among these people who are connected, how’s their behavior on my app? Are they going to my app or not?

Are they opening my app or not? Are they opening my app much less frequently? Because if that’s the case, then obviously this experience is so compelling that I would have a choice to make. How do I make this experience maybe not as compelling as my app experience, or how do I incentivize them here to open up my app?

There’s a new battle happening for that first category, which is a new interface to be owned. I’m curious if you think being the first mover matters to building a new ad network, because there’s Gemini, there’s Anthropic, and there are a bunch of people who have tons of users using this new interface. How do you think about the landscape of the new potential entrants to build the next dominant ad network? What advice would you give these various parties?

The good news is that being first doesn’t matter. Why? Because you control—especially if you’re in category 1, which we described—you control your first-party inventory. In fact, being second or third means you can learn from the iterations and mistakes that the first one makes. Your inventory is not going anywhere.

Some might have more urgency to monetize than others, but Gemini doesn’t need to monetize anytime soon, so they can just sit back. They have a lot of ad expertise and data from Google. They can sit back and wait until they need to monetize.

In fact, a good strategic move for them might be to say, “I am the zero-ad platform,” like Apple claims, or Google can claim that Gemini has no ads in it. There is a certain set of customers or consumers who care about that.

But the biggest thing, I think—and OpenAI has done a good job of articulating this—is that ads should not influence the content that is served to me or the recommendations that AI gives to me. I think they should be relevant, but they should not be influencing the recommendations.

Second, you have to keep a high bar for engagement and usefulness. Unfortunately, however relevant ads are, the reality is—and this has been proven—that once you start showing ads on a previously unmonetized, zero-ad surface, user engagement goes down over time.

It does, because some of the engagement gets siphoned off by ads and some of it gets siphoned off in different ways. Many holdout groups across many companies have proven this. So the question for any one of these companies is: How much engagement are we willing to take in exchange for monetization?

I think, first, you need to have a holdout group. I’m sure they’re having a holdout group of people who never, ever see any ads, because that’s your fresh group that never sees ads, and you need to understand their behavior. Then you need to always understand how people with ads are behaving, and figure out what the engagement hit is from each quantum of ads. You need to then give your ads team a certain engagement budget.

At Facebook, there was an engagement budget every year that, between the News Feed team and the ads team, we had to adhere to. In other words, yes, we wanted this much revenue, but the check metric on the revenue was: We can’t take more than X% dip in engagement overall for News Feed.

What are the attributes of a good North Star metric? What advice would you give someone who’s trying to pick the thing around which the company is going to optimize?

Yeah, the North Star metric is a metric that is an indicator of company growth and customer value. So it actually balances customer value and business value nicely. The North Star metric, in my opinion, should not be revenue. It should be something that is directly correlated with customer value.

For example, if customers are doing well, the North Star metric should go up and to the right, but it should also lead to the business doing well. For Square, the North Star metric was GPV, which is the volume of payments processed. It was not correlated to revenue—it was somewhat correlated to revenue—but most importantly, it showed that the amount of payments processed by the company was continuing to grow.

At Facebook, the North Star metric was MAUs. It was actually monthly active users. Then, over time, it went to daily active users because it was, in a sense, an indication of how engaged users were.

One of the most important things about an NSM is that it needs to be coupled with what we call check metrics. In other words, if NSMs are left alone, then, as you know, incentives drive behavior. So if you tell a team, “Go and optimize this NSM,” they will do what it takes to make it go up 100%.

But then many things that you don’t want to go down could go down. For example, in the DoorDash case, you could say, “I want to grow GMV,” which is gross merchandise value and the North Star metric. GMV is the total value of all the orders that go through the marketplace.

I could make it grow by setting the delivery fee to 0 and setting everything to 0. What happens then? The company’s revenue goes to 0. So you basically want a check metric around the health of the customer and a check metric around the health of the company. Those are the guardrails around this NSM.

In the case of DoorDash, it might be: I want to maintain a certain gross margin percentage, or I want to maintain a certain customer retention percentage, something like that. Again, margin is typically a good one to use because, in some ways, that is an indicator of the company’s health.

There are 2 ideas that we talked about when we first met. One was the need for the very best software companies to stand alone, in the sense that someone can just go use it without talking to a human, and it just works for their problem.

So, fully self-serve. I’d love to hear you talk about that. A related idea was on the builder side. On the investor side, you mentioned to me that all the great investments you’ve had—the companies that have really had explosive growth—have had a high number of one of 4 qualities, which I think were gross margins, low cost to acquire the customer, high retention, and a tight sales cycle, which maybe maps back onto the self-serve thing.

So, talk about the relationship between those 2 things.

The self-serve notion actually came from Google. Google was the first company I worked at that achieved massive scale. Within the ads team, we basically had a wide number of customers using us—millions of customers. There were a lot of small businesses, but there were also large companies.

What we ended up doing to serve the large companies was build a lot of tools. Large companies didn’t want to use the product themselves. They had agencies using it for them on their behalf, and they also had internal people at Google—support, sales, and operations people—using it.

On the product side, we built a lot of tools for our internal colleagues, for our sales and operations colleagues, to manage the system for our large customers. One day, I think we were at a Larry review, and we were showing these things we called ICS, or internal customer systems, to Larry. I think we weren’t meaning to show it, but to show him a demo, we somehow got into it. He was like, “What is that?”

We were like, “Well, it’s a system used by our internal teams.”

He was like, “Why’d you build it?”

We were like, “Well, we have to help our large customers.”

He said, “What do you mean? Our small customers don’t have access to it?”

We were like, “No.”

He said, “End it right now.”

We were like, “What do you mean, end it?”

14. Hiring Functional Experts for the AI Era

He said, “I want to make sure that everything you’re building for large customers is also available to small customers.”

So, we basically had to take everything we had built over the years for this enterprise system and make it available to customers. An interesting thing happened: the smaller customers adopted it much faster, because some of the things we were building had advanced knobs and so on that we didn’t think they would use.

It turns out the self-serve customers were the most sophisticated users. If you do something that’s interesting, there are all these small agencies, entrepreneurs, hustlers—all of these folks. If you can help them make more money, it’s a testament to human creativity and ability: they exploit the system in ways that you never even know, and you learn a lot from working with them.

I’ve seen in every case that when you open up your system to self-serve, you learn so much more about the capability of your product than if your sales team is doing it on their behalf. In fact, I’ll never forget that in AdSense, I think we had some of the largest publishers in the world sign up and start using us on a self-serve basis, and then we engaged with them after that.

I think companies like Atlassian, Square, and Nike signed up. I think we had Whole Foods. I think Nike started with a Square device and started using it in one of their stores.

Self-serve does 2 things. One, it makes your product better. It makes your product better because these folks use the product in ways that you don’t expect or anticipate, and it forces you to think about what self-serve actually means.

The definition of self-serve is that the customer can not just use, but onboard and use, the product without ever talking to or engaging with a single member of the employee base at the company. When you do that, you have to think about how they actually get set up with the product.

It puts a lot of effort on onboarding, because onboarding is one of those things where most people drop off if you don’t do a good job. Then you’ve got to get them to a moment of delight very quickly. All of those things, if you’re not building a sales product, you don’t even think about. In a seller product, you think about it every day. It’s like a consumer product or a business product.

Second, what it does for you is open up the aperture to your customers. With, say, 100 salespeople, you can reach maybe 10,000 customers. But with a self-serve product and the right word of mouth, you can reach millions of customers.

Look at Cursor, for example. It is used in every large company, I bet, but only maybe 1% of companies have a top-down motion. In 99.9% of companies, some engineer got it. Figma is another great example.

After I invested in Figma, I joined Square. One and a half years later, I tried to push Figma down from the top into the design team. Having learned design, I said, “You’ve got to use Figma.” The designers refused to use it. They were using a tool called Sketch, and they said, “We’re not going to use it. Sketch is much better.”

I felt, “Okay, it’s not my place to tell them what tools to use,” so I backed off. Two years later, a mid-level design manager came in, brought in Figma from their prior company, and basically got it used across the organization. It kicked out Sketch.

With self-serve, you can get into these situations where, even when there’s an incumbent, you can infiltrate and be an insurgent in a unique and powerful way that a direct-sales motion could never have produced.

One of the other dimensions that’s changing fast is careers. I’m curious what you think about the sorts of people that will thrive best in this new era. If you’re a person hiring someone, what are the sorts of things that you would place extra emphasis on now in the AI era?

The number-one thing, I think, is going to be a focus on doing and building. I think CEOs have gotten too comfortable over time, and I think this is changing hiring middle management very, very quickly and hiring C-level people instead.

15. Advice for Managing a Career

I think you’re going to see the rise of AI agents doing a lot of work, but then humans who manage the AI agents and are individual contributors. The number-one skill that is going to be relevant 2 years from now, probably even 1 year from now, is to become a functional expert who knows how to build AI agents to do that function and orchestrate an army of AI agents to do that function.

There was a great article the other day I read about a product manager at Meta who’s nontechnical, but who basically built a bunch of AI agents to do his job as a product manager so well that even his engineers were saying, “Teach me how to use AI agents well.”

That’s what you want. You want somebody who is essentially acting as a manager, but not of humans—of AI agents. Management has to be a full-time job. What I mean by that is, if you manage 3, 5, or 10 people, that’s not enough. You either need to be managing 50 humans, or you need to be an individual contributor.

There’s something called span of control, which means how many people you manage. In some ways, a span of control of less than 10 should not be allowed at any company at this point. Everyone should have a full-time job, because think about it: if you’re managing even 15 people, maybe you meet with them once a week. That’s 15 hours. What are you doing for the other 25, 30, or 40 hours?

You should be working.

On the company side, don’t hire managers for as long as possible. Hire doers. Hire builders.

How do you—what is your favorite way to assess whether or not someone is that, in interviewing them or learning about them?

The best way is to give them a work project. Engineering does a great job. Engineering has always done a great job. Every company I’ve been at has had engineering coding interviews, programming interviews.

Yeah. Do stuff.

16. Evaluating Founder Authenticity

Everywhere else, you can just bullshit your way through without doing stuff. You can just talk and talk, and talk is not enough. You’ve got to actually do stuff. Produce an artifact.

At Square, we established work projects. Even for corporate development, I remember our work project was, “Give me one company that Square should buy, analyze the company, tell us why we should buy it, and tell us what the synergies should be.” The best candidates had to do that.

Every function needs to have a work project. You need to put them in a room without AI and get them to do the project, get them to do the work that is ideally very similar to the work they’re going to do.

For product managers, we would almost give them a product we were thinking about and say, “Here’s a product we’re thinking about. Figure it out. Should we build it?” The first and most important thing you want for these kinds of things, especially for customer-facing roles, is that they need to take the voice of the customer. In other words, they need to justify the why.

The best product manager candidates rejected the premise completely, and they did it in a beautiful way. They went and talked to 10 customers on the street. It’s so brilliant. They said, “I talked to 10 customers. They were all Square users, which is so easy—Mint Plaza, you go there—and we found that none of them want this premium insights product. So, we don’t build it. We’re going to build this other thing.”

I said it was amazing. That’s what you want to see. You want agency. You don’t want people to just say, “Give me what to do and I’ll do it.” You want people to reject the premise or question the premise in the first place.

Square should not buy a company? That would be great. Why? Tell me why. That’s the kind of thinking you’re looking for.

What was Tony’s thing?

Tony’s thing was that he would give people either $10 or $20 and ask them to acquire 1,000 customers for DoorDash consumers. Some people would literally say, “I’m not going to take this challenge.”

"I'm not ready for it," or something. And it's great if you literally opt out of it. Some people would take it, and nobody even came close to acquiring 1,000, or even 100, I think, but the goal was to see how many different things they were able to try in the course of a few hours.

Someone went to the gym, printed flyers out, and gave them out. People tried all kinds of things. It was a brilliant way to just filter out people who didn't want to do stuff.

Is there any other advice that you would give the person building the career? We talked about evaluating, being a builder, and all these sorts of things. How should one think about managing a career in the AI era?

Stay at every job long enough to have impact. Over the last 18 to 24 months, I've been seeing this phenomenon of job hoppers, or job optimizers, I call them, who stay at a job for 12 to 18 months and then move to the next job, and then stay 12 to 18 months and move to the next job. I think that is one of the biggest red flags as a hiring manager that I see because I don't think you can achieve anything of value. You can't have any impact on a company in 12 to 18 months.

I think it takes a minimum of 3 to 4 years to have impact on a company. My top advice is: stay long enough to have an impact, build a network, have fun. From the moment you start a job, don't be thinking about what your next job is. Once in a while, maybe one job didn't work out amongst a series of jobs, and you left it after 18 months. But if I'm seeing 2 or 3 jobs back to back, that's an immediate red flag.

I posted this on X, and tons of managers wrote to me saying it's an immediate red flag. You do yourself a massive disservice, and you won't even know what the problem is. You'll get rejected, and you won't know what happened. People want people who stick around and build. Who's going to hire you if they see that's your behavior? I think it's very short-sighted, very short-term thinking. You've got to build something of value, and that comes with time.

17. Best Practices for Board Management

So much of the theme here has been identifying a superpower, having one in the first place, evaluating one, matching it to a problem with a leader, and so on. With your investor hat on and your new firm, Marathon, how do you assess the capacity or existence of a superpower in a person? How have you learned to do that?

The most important thing I look for is founder authenticity. If you think about it, 3 of the 4 companies I worked with—Google, Facebook, and DoorDash—all started in school, all started in college, and they all started as a toy problem almost that the founders were curious about. They started with an authentic curiosity: Can this be built? And then it got built and it started.

Similarly, with Jack and Jim, they started solving a real problem. My first question to every founder is, "Tell me your founding story. Why did you decide to start this company?" The founding story, in my opinion, expresses why they chose this problem, and ideally it should touch on what the superpower is and what compelled them to work on this problem.

I've had many people work with me or for me who have gone out to start companies with the only reason being, "Well, I have my buddy, and we both want to start a company together." I really advise them not to do that because just going out and starting a company because you want to start a company with your friend is the wrong reason. I want to understand whether there is an authentic lived experience that they've had in their life that compels them to work on this product.

Dylan from Figma, if you talk to him, is steeped in design. He thinks about the design of things. He thinks about how to make things more compelling, and it was very clear that he had a vision for what this thing would be. A good example is a company called Faire. It's a B2B marketplace. Max Rhodes, the CEO, worked for me at Square.

When he left Square, he actually tried many different ideas, and it turns out none of them were authentic to him. The idea that worked was Faire. Why? Because when he was an undergraduate student, he had an umbrella company that he created, and with this umbrella company he was trying to get distribution for it in local retail. It was extremely hard for a brand. How do you get local retail? There are so many of them. How do you go in and pitch to them?

He realized that this was the problem he wanted to focus on: other manufacturers who wanted to get access to local retail.

Are there any other questions that you love to ask in a first meeting, learning about a company, other than, "Tell me your origin story?"

I think the other one is the idea maze: "Tell me about how you navigated the idea maze." You want to tackle this problem because, again, this is a classic product thing. You start with the problem, but then there are many different solutions, many different ways to solve it. Why did you choose this solution? Why did you choose this way versus the other way?

I will basically try to throw them off course or off-kilter by asking them 5 or 6 other ways to solve the same problem and understand whether they are students of either history or their industry. Why could this problem not be better tackled in another way? I want to understand that they have studied alternate approaches and historical approaches to solve this problem.

I think a good example is the Collisons. I think they bought a book on payments, and they studied exactly why all the payments companies did what they did, how they failed, and how they succeeded. I think the best founders are students of history in that industry, and they understand why all the prior companies took the decisions they did. Ideally, they stand on the shoulders of giants and are able to build this company.

The other thing I always recommend to CEOs is, a board role is like a marriage. Once you get into it, it's very hard to get out of, so never, ever, ever invite anyone to join your board before spending at least a year with them.

Have them join an advisory board. Have them meet with everybody on the management team. Spend time with them. Have them come to a few board meetings. Have them meet with the other board members. Have them come to a board dinner. Have 3 or 4 people on your advisory board, and then make 1 of them a board member if you like them, if you feel they're adding value, if your team feels they're adding value, and so on.

The other thing I've seen with boards over the last 15 years is the management team getting involved. 15 years ago, it would just be the CEO, the co-founder maybe, and the board. We'd meet for 4 or 5 hours, discuss topics, maybe bring in a management team person for 1 slice, the CFO, and then they would leave.

Now, in most companies, they have the management team attend the entire board meeting, except for what is called the executive session. I think that is awesome. Why? Because I think the management team and the board get to meet each other. As part of a board, you want to understand who's on the management team, who could be a successor to the CEO, and what the capabilities of different parts of the management team are.

As the management team, you want to be able to leverage the board for help. I think one of the best practices I've seen—and I've now tried to push other companies to do it—is the notion of a board buddy. Everyone on the board should become a buddy to a management team member, and they would then meet with that management team member multiple times between board meetings, once a month, or even text with them.

They're almost like a sounding board for anything the management team member has. You can see that the different board personas I described map nicely. I generally am the management buddy for the head of product or the head of engineering. Somebody else is a buddy to the CFO. Someone else is the head buddy to the CRO, and so on.

I think the meetings in between the board meetings are actually just as important as the board meetings themselves because a board meeting can have a lot of things going on. Those relationships—that's the other thing I realized—it's not the board meeting that truly matters. It's all the things between the board meetings that are the real, real thing when things get done.

I think the only thing we haven't talked about in this grand art of company building and product creation is the job of acquiring the customer, positioning the product, marketing, and the way it presents itself to the outside world. What's the dispatch from the cutting edge that you're seeing of how people do this? All these things—position, brand, customer acquisition, the ways they do that—what does new excellence look like to you across the many, many companies that you get to see?

18. The Kindest Thing

One of the most interesting things now is that it's different between enterprise-focused and consumer-focused companies. For consumer-focused companies, the big thing is how to scale influencers. I think influencers have become much, much, much more powerful every year in how people, especially younger people, consume products and even choose products.

Somebody said that TikTok is the best local search engine, and I think that's right. My kids have discovered crazy restaurants on TikTok when we go traveling that Google Maps would not really show, or Yelp doesn't show, and so on. How do you reach influencers on TikTok? There's a set of companies that's coming out that's essentially making it easy.

The problem is influencers on TikTok. Obviously, there are head influencers, but there's a long tail that goes viral for different reasons, and you want to capitalize on those viral waves if possible.

So there is a set of companies that is building products to see if they can help brands connect with these influencers in scalable ways. On the enterprise side, I think the most interesting thing I'm seeing is not really an acquisition channel as much as it is an onboarding channel. It is basically presenting an outcome to a customer and saying, “Let's collaborate on outcomes.”

Palantir does that very well. Palantir goes to customers and says, “What's your most important business problem?” “Oh, here it is.” “Okay, great. Give us 6 months to solve it. Engage with us. If we can't solve it, fire us. Don't pay us anything. If we solve it, pay us a lot of money.” So it's truly taking ownership.

And I think this goes to outcome-based pricing: how your product is priced and your confidence in your ability to deliver that outcome, of course. So I think outcome-based selling is one of the most interesting ways of changing how you sell. In fact, one of the top pieces of advice I have for founders reaching out to companies is that you cannot lead with what your product does anymore. You've got to lead with what is the outcome you can deliver, or ideally even have delivered.

I'll never forget this example, and what is crazy is that companies always look to other companies in the vertical. This never will change. For example, if you get J.P. Morgan to use your product, I promise you every single bank will then evaluate your product. But if you get Procter & Gamble, J.P. Morgan doesn't care if Procter & Gamble uses your product.

So even when you go to market, you've got to target. Instead of trying to be too horizontal, unless it's bottoms-up, on the sales side you've got to try to go after 1 or 2 very specific verticals because there is a very clear lighthouse effect. You want to go after the best one and get the best one, and then you basically win all the other ones in that vertical.

I think you might know my traditional closing question that I ask everybody: What is the kindest thing that anyone's ever done for you?

There are so many. I think the best one is a guy called Bob McDonald. I was basically a business school student on the East Coast. I really wanted to get a job. I was on a visa, and I wanted to get a job in Silicon Valley. I was somewhat unqualified. I'd never been a product manager before. I'd been an engineer and never worked in photonics or optical networking before.

Bob basically saw a spark in me and said, “You know what? I'm going to make a bet on you. I'm going to hire you, and I'm going to bring you to Silicon Valley. You're going to be at a Sequoia-funded company, one of the hottest companies in the Valley.” He could have had his pick of anyone, but he bet on me. So I basically have taken this approach that I try to pay it forward, and I have no expectation when I do something for someone.

What created the spark in you? Like, what about your life? Where did the spark come from?

For me, it's all about just knowing how fortunate I am to be healthy, to have a family that loves me, and to know that in almost every run of the simulation, I could be in 1 of a million different worse circumstances than I am today. And so, just gratefulness and gratitude about where I'm sitting.

I mean, we are sitting in literally the top 1% of the 1% of the 1% situations right now and breathing. And so, literally, I think I feel pain when I see somebody suffering. I see, as they say, “There but for the grace of God go I,” in some ways. You basically realize that you're very lucky to be given this 1 life, and you have a responsibility to the world and yourself to be grateful and to lead the best life you can.

Gokul, this was incredibly fun. Thank you so much for your time.

Patrick, thank you. Thank you, my friend.