Gokul Rajaram on the 8 Moats Companies Need & Why Dropouts are 'AI Maxing' the World

tldr

Gokul Rajaram — operator-turned-investor who has worked at Google, Facebook, Square, and DoorDash — builds a new framework for evaluating software durability in the age of AI: the Eight Moats. He argues the public market's "everything software is going to zero" reaction is an overreaction, but the companies that survive will need to score 4+ on his moat scale. He also covers what real AI transformation looks like vs. bolt-on AI theater, the death of Seat Pricing for "Work Products," why Vertical SaaS founders must own the full stack, and why dropout founders are AI Maxing their way to outcomes.

Key Takeaways

The 8 Moats: Data, Workflow, Regulatory, Distribution, Ecosystem, Network, Physical Infrastructure, Scale. Score 4+ and you're defensible. Score 1 or fewer and you're in trouble.
SaaS isn't going to zero — but it's been painted with the same brush. Market overreacted. Atlassian scores ~3, Monday.com scores ~1. Atlassian is oversold; Monday may be fairly priced.
Bolt-on AI only works if you reframe the product, not just add a feature. The companies that win rebuild the entire experience end-to-end. Those that don't are just adding a thin API layer.
There are "Access Products" and "work products." Access products use seat pricing (ChatGPT Enterprise). Work products should use Outcome-Based Pricing (Harvey per contract processed). Seat pricing fails when the user isn't the constraint — the work output is.
Vertical SaaS founders must own the full stack. One function within one vertical won't scale to $10B+. ServiceTitan has 32 products. The ambition has to be: replace all digital labor in this vertical.
Young dropout founders are AI-maxing. They adopt tools faster, live differently, and are producing some of the most impressive early-stage companies Gokul has seen in 15 years.
Durability > growth rate. Triple-triple-double-double is table stakes now. What matters: Gross Retention and Net Revenue Retention. A company growing 10x with bad retention is a disaster in slow motion.
Pure remote kills early-stage companies. Gokul changed his mind on this after watching founder co-founder misalignment destroy companies that had everything else going for them. At least 3 days/week in person.

Detailed Breakdown

[00:49] Career as an Investing Lens

Google: Taught him that remarkable product is the only foundation. "Build it remarkably and they will come." GTM can be figured out. Unremarkable product cannot be rescued. Gmail was 100x vs. Yahoo Mail at launch — that's the bar.

Facebook: Distribution is real. Mark Zuckerberg is the best distribution mind in the world. Multiplayer products create structural switching costs — the best PLG companies are the ones where multiple people using it increases defensibility. Figma was his example of seeing this clearly.

Square: Multi-product portfolio is essential. Square went from 1 product (payments) to 11 products each doing $50M+ revenue. The north star metric became median number of products used per merchant. Key insight: not every product needs to make money. Some products are profit pool products. Some are retention products. Confuse the two and your teams are building for the wrong outcomes.

DoorDash: Operations at its hardest. The COVID decision to waive revenue share from restaurants for a month — painful in the short term, correct long-term. The caliber of people who come out of DoorDash is elite.

[07:58] The SaaS Apocalypse — Overreaction?

Public markets decided that since code is becoming free at the low end, every software company is going to zero. Gokul's read: 100% overreaction. Not all software companies are created equal. The answer is in how many moats they have.

[09:00] The 8 Moats Framework

A riff on Hamilton Helmer's 7 Powers — but updated for the AI era.

#	Moat	Description	Example
1	Data	Proprietary data nobody else has. Gets better over time.	Spotify's Discover Weekly (decade of listening behavior across billions)
2	Workflow	Embedded in operations. Weak alone but powerful if deep.	NetSuite (runs the business) vs. Zendesk (lighter layer)
3	Regulatory	Licenses, compliance, multi-year procurement contracts	Coinbase (MTLs, FCA — basically impossible to switch custodians)
4	Distribution	Proprietary, exclusive channels	Intuit/QuickBooks (trained a network of CPAs who will refuse to use anything else)
5	Ecosystem	Third parties built on your platform	Shopify (can't vibe-code 100K+ apps developers have built on the ecosystem)
6	Network	Marketplace density, liquidity, reputation history	DoorDash (AI can vibe-code the tech but not the restaurant density or delivery reputation)
7	Physical Infrastructure	Atoms create moats software can't replicate	Data centers, logistics, hardware
8	Scale	Costs so low at scale nobody can match	Amazon, TSMC (pure software scale is now commoditized by AI — only hyperscalers qualify here)

Score 4+: pretty damn secure. Score 2–3: real risk. Score 0–1: rebuild now.

Worked example: Atlassian scores ~3 (data, workflow, ecosystem), Monday scores ~1 (workflow only). Atlassian is being massively oversold. Monday may be fairly priced.

[17:42] What Happens to Systems of Record (Salesforce)?

They have a score of 3 — similar to Atlassian. More defensible than most, but they can't sit still. The strategic move: commoditize the complement.

If the profit pool is in workflows → make data storage free, charge for workflow outcomes
If the profit pool is in data → give away agentic/workflow UI and charge for data intelligence

Either way, they need to aggressively build, not wait for others to build on top of them.

[19:42] Bolt-on AI vs. Real AI Transformation

Bolt-on AI has a real ceiling. What separates winners: they reframe what the product does, not just add capability.

Wrong: Add AI search as a feature
Right: Rebuild search as an entirely new experience with new UX primitives

Intercom and Podium both burned the bridges. Podium hit $100M in agent revenue and is on track for $300M. The playbook: don't try to fix the legacy business — build a new business from scratch, then migrate customers ruthlessly, even at a lower price.

The key product discipline: model capabilities improve every 6 months. You cannot have a long roadmap because the next model release will blow it up. Re-evaluate every interaction against new model capabilities constantly.

[31:54] Pricing: The End of Seat Pricing?

Seat pricing doesn't die — it's still the right model for access products where predictability matters to enterprise buyers (ChatGPT Enterprise, Figma tiers).

But seat pricing breaks for work products where the user isn't the constraint — the work output is. If 100 people use your product and process zero contracts, you should charge zero.

The framework:

Access product → seat-based pricing
Work product → outcome/consumption-based pricing (e.g., Harvey charges per contract processed)

[24:16] Vertical SaaS — Can It Still Work?

Yes — but only if the ambition is to own the full stack. One function within one vertical won't get to $10B+. The target: replace all digital labor in the vertical, not just one slice.

ServiceTitan: 32 products, still only ~$10B. Contrast with Robinhood (13 product lines, 100M+ revenue each) or Coinbase (12). Horizontal platforms with broad bases can build product lines much faster.

For vertical SaaS, the TAM calculation has changed. You're not sizing software spend — you're sizing BPO spend + human labor. The AI transition from software budgets to human labor budgets is already happening, in this order:

Cut BPO/outsourcing contracts (easiest, no layoffs)
Don't replace people who leave
Eventually, actual layoffs

[45:51] Non-Consumption Markets — The Biggest Misses

The biggest misses in Gokul's career were non-consumption misses: not recognizing markets that didn't exist before. Shopify (he passed at $1B, didn't realize it was enabling anyone to sell, not just existing merchants). Facebook (couldn't predict $2T from $40B terminal value discussions in 2010). Google post-IPO at $30B — "too expensive."

The lesson: within every category, there are great companies and mediocre companies. Pattern-matching on the category (D2C is bad, food delivery failed before) destroys returns. Quince had 35–40% repeat purchase rates in the deck. He missed it. Never again pattern-match without looking at the data.

[63:24] When to Sell

As an LP-serving fund investor: Don't just optimize for MoC. Optimize for IRR. If the go-forward IRR for an asset you hold is lower than your fund promise to LPs, sell.

Fred Wilson's rule of thumb for fully liquid assets: sell a third, hold a third, trade a third.

Secondary markets now enable this for the best private companies. Use them.

[71:47] Quick Fire

Changed his mind on: Pure remote doesn't work for early-stage companies. Co-founder misalignment killed companies that had everything else. Need at least 3 days/week in person.
Advice to young people: Get 2–3 years of real work experience before starting a company. Life is long. The network and operational experience will compound.
Best fund picks: First Round (seed), Benchmark (Series A), Green Oaks (growth)
Hardest career decision: Leaving Google. "You leave Google only once."
Best CEOs worked with: Larry Page (technical), Mark Zuckerberg (growth/distribution), Jack Dorsey (design), Tony Xu (physical world operations) — next in line: Jeff Bezos
Highest multiple angel investment: Figma. 500–1,000x at IPO time.
Biggest miss: Quince (dismissed as D2C without looking at the 35–40% repeat purchase rate in the deck). Also: couldn't predict Facebook becoming $2T.

Notable Quotes

"If there is not a remarkable product, all the go-to-market and distribution in the world will not save you."

"You cannot be a single product company. If you confuse profit pool products with retention products, your teams are building for the wrong outcomes."

"Bolt-on AI by itself has a real ceiling. The companies where it really works reframe what the product does — not just add the capabilities."

"Seat pricing fails when the product's core value is not about access but about something doing the work on your behalf."

"Young people are AI-maxing more than anyone else. Companies not hiring young people are making a huge mistake."

"Pattern-matching on the category destroys returns. Within every industry, there are great companies and mediocre ones."

One Thing to Act On

Score FeatureOS and SupportWire on the 8 Moats right now. Not as aspiration — as honest assessment. If either scores below 3, the next product decision should be building a moat, not a feature. For SupportWire specifically: can you accumulate a proprietary data asset from support conversations that gets better with every interaction? That's the only moat available to a pre-launch product. Name it. Build toward it deliberately.

Tags

#moats #defensibility #saas #ai #vertical-software #pricing #venture #multi-product #distribution #non-consumption #gokul-rajaram #20vc

Raw Transcript

Auto-captions from YouTube. Folded by default — expand if you need to grep the source or pull an exact quote.

0:00 I call it the eight modes. Data mode, 0:02 workflow mode, regulatory mode, 0:03 distribution mode. 0:04 >> We're on number five. I'm loving this. 0:06 >> Ecosystem mode, network mode, physical 0:08 infrastructure, and the eighth one, I 0:10 would say, scale mode. I think anything 0:11 four or more, you're pretty damn secure. 0:15 >> I'm thrilled to welcome one of the best 0:17 operator turned investors of the last 0:19 two decades, Gawk Rajaram. 0:21 >> You cannot be a single product company. 0:23 Vertical products, you've got to really 0:24 own full stack. It's harder otherwise to 0:26 be a 10 plus billion dollar company. 0:28 >> What's the biggest miss? Is it banter? 0:30 >> Most recent sequins, but to be honest, 0:31 even bigger miss than that in some ways. 0:33 It's not a miss in terms of investing. 0:34 We said 0:36 >> ready to go. 0:49 I've wanted to do this for years and you 0:52 have I mean you've coily played hard to 0:54 get let's put it mildly uh after my 0:57 continuous WhatsApp messages but thank 0:59 you so much for joining me today man 1:01 >> it's my pleasure to be here my friend 1:02 thank you again 1:03 >> now I wanted to start with how some of 1:07 your prior companies that you've worked 1:08 at have shaped your investing mind 1:10 specifically and I wanted to start with 1:12 Google when you reflect on your time 1:14 with Google how did that shape your 1:16 mindset for the types of companies that 1:19 you Like today 1:20 >> I think the best way to think about my 1:22 the Google experience is Google taught 1:25 me that ultimately the best companies 1:27 have a remarkable product at their core. 1:30 Google was a remarkable product. Google 1:31 was definitely a philosophy of build it 1:33 remarkably and they will come. Uh GTM 1:36 was not Google specialty but what Google 1:38 was really good at was building amazing 1:40 product. Sometimes the go to market 1:42 worked sometimes it didn't work. So but 1:44 at the core was remarkable product. So I 1:46 think I ultimately my core investing 1:50 thesis is that if there is not a 1:52 remarkable product all the go to 1:53 marketing distribution in the world will 1:55 not save you. So that's I look for what 1:58 the remarkability is in the core product 2:00 or value proposition of the company. Is 2:02 it 10x 100x better than the alternative? 2:04 I'll I'll tell you a story at Google. 2:06 When I joined in 2003 there was a 2:08 project going on called Caribou 2:10 internally. I was like what the hell is 2:12 this? This was web email which gave 1 2:15 GBTE free storage and back then Yahoo 2:17 mail offered 10 megabytes of storage. So 2:19 it was 100x. I thought it was truly I 2:21 was like there's no way it's possible. 2:23 And turns out it was and it was released 2:24 on if you remember April 1st 2003 and 2:27 that people thought it was an April 2:28 Fool's joke. But that was Google 2:30 literally taking something that was 2:31 unbelievable and making it reality. And 2:34 so that's that's the kind of products I 2:36 like something remarkable something 2:37 unique something powerful. 2:39 >> I like it. It reminds me of actually 2:40 Neil Ma who talks about kind of 2:41 jaw-dropping customer experience as one 2:44 of his kind of core monikers for 2:46 thinking about companies and 2:46 investments. Next we have Facebook. How 2:49 did Facebook impact the types of 2:51 companies that you like? What is 2:53 interesting is that even if you have a 2:54 remarkable product uh you still need 2:56 distribution. Facebook taught me the 2:58 power of distribution. Mark I think is 3:01 Mark Zuckerberg is probably the best um 3:04 I would say uh distribution genius in 3:08 the world. he would look at a product 3:10 and say this is how this product is not 3:12 going to work and it taught me the power 3:14 of multiplayer products in particular I 3:17 think uh most software products are 3:19 single player and as soon as you make 3:21 the multiplayer there is a uniqueness in 3:23 switching distribution etc that comes 3:25 about Facebook by nature you can't use 3:28 it if you only have one person on 3:30 Facebook and so when I saw Figma the 3:34 power of Figma I felt was it was not 3:35 just that a person could use it but it 3:37 was much easier to share with other 3:39 people in your company and I think the 3:40 best PLG software companies are those 3:42 that you can use multiple people can use 3:44 and it increases defensibility. So the 3:47 power of distribution and multiplayer 3:48 products. 3:49 >> What about Square? Square was an amazing 3:51 journey. What did you learn from Square 3:53 that you've taken to your investing? 3:55 >> The power of a multi-product portfolio. 3:58 I think at Square when I joined we were 4:00 a single product company payments and 4:01 payments only. uh when I left we had I 4:04 think 11 products each doing more than 4:06 50 million in revenue and one of the 4:09 interesting metrics was we went our key 4:11 northstar metric went to median number 4:13 of products used by a seller by a 4:15 merchant turns out that the more 4:17 products that a merchant uses the more 4:20 retentive they are the more sticky they 4:22 get so this is the other thesis I have I 4:24 mean this is obviously very clear now 4:26 you have to have you cannot be a single 4:28 product company uh you've got to make 4:31 sure and and most importantly Your 4:33 product number two needs to emanate very 4:35 naturally. It can't be like this 4:37 completely separate product. It has to 4:38 be very adjacent product number one. For 4:41 Square, it was a product called Square 4:42 Capital, which was basically a cash 4:44 advance product that really came from 4:46 the fact that Square controlled the 4:48 payment flows and knew exactly the the 4:51 merchants credit history and could 4:52 underwrite based on basically money 4:54 going in and out. And um it was a 4:56 beautiful product. And the the the 4:59 interesting thing about having a 5:00 multi-product portfolio is that not 5:02 every product needs to generate profit. 5:04 I think there is a uh people always like 5:06 oh it's not making money. Square capital 5:08 didn't make much money but it was very 5:10 good for retention. Some products are 5:13 good for making money. Profit they're 5:15 part of the profit pool and some are 5:16 good for retention. Companies need to be 5:18 very clear which are the profit pool 5:20 products and which are the retentive 5:21 products. If you confuse the two your 5:23 teams don't know and they're built for 5:24 the wrong outcomes. But the power of a 5:26 multi-product portfolio and being able 5:28 to have pro products with different 5:30 goals, retention versus profits. 5:34 >> I I love that in terms of it doesn't 5:35 need to be profitable. I also just see 5:37 so many investors stay being relatively 5:38 inelastic in terms of their mindset on 5:40 margin where it's like oh the margins 5:42 are [ __ ] and it's like I don't you know 5:44 going to our next company negative gross 5:46 margins for the first like first year I 5:49 think. Uh negative gross margins 5:51 >> dude I I don't think Door Dash had great 5:53 gross margins for the first few years 5:54 either. I think Dan Deliveroo did who 5:56 obviously Door Dash acquired and so it's 5:58 just funny that we kind of repeat the 5:59 same mental cycles of oh the margins are 6:01 [ __ ] and it's like yeah so were the best 6:03 companies margins Spotify didn't have 6:05 great margins for a very long time and 6:07 their margin increase has been amazing 6:08 actually dude Door Dash what is the 6:10 lesson from Door Dash it was the most 6:12 operational of all four companies I 6:14 thought I was a good operator but when I 6:16 got to Door Dash I really realized what 6:18 operations means and so a lot of my 6:20 philosophies around how to truly operate 6:23 in a hard hard in hard mode have been 6:25 shaped by it. It really was the 6:27 epithesis or the epitome of I think how 6:29 product and operations can work together 6:31 in the physical world. So how it shaped 6:33 my uh my investing philosophy is the 6:36 kinds of people that that came out of 6:39 Door Dash. I think they are I I just 6:42 think they are excellent and I try to 6:43 get them. It's really around hiring. 6:45 It's around talent. It's around taking 6:46 really hard problems and solving. I'll 6:48 never forget when COVID hit as you as 6:51 you know most restaurants were shut down 6:53 for the first for the first couple of 6:55 weeks and so Door Dash had to make a 6:58 very hard call around how to basic what 7:01 to do how to get these restaurants to 7:02 open and ultimately we decided to not 7:05 take any revenue share from these 7:07 restaurants for a month. Uh even though 7:09 we were a private company we had some 7:11 amount of cash on the balance sheet and 7:13 that really hurt. It was the right thing 7:15 to do in the long term, but it was 7:17 extremely painful in the short term. 7:18 >> You spoke about kind of the skill of 7:21 operators to work both in a physical and 7:24 in a softwarebased environment there 7:25 with Door Dash. You know, with Project 7:27 Europe, which we chat about before, 7:28 we're seeing all of our hardware 7:30 companies be so freaking popular right 7:32 now because everyone's just terrified 7:34 that bluntly Anthropic is going to eat 7:36 their lunch as we keep seeing with 7:38 Anthropic doing security and security 7:40 stocks plunge. I want to talk about the 7:42 SAS apocalypse because my job with this 7:45 show is to learn from people much 7:46 smarter than me and I'm lucky to do that 7:49 here. Is the volatility that we're 7:51 seeing justified or are we in a manic 7:54 hype oversell environment with emotional 7:57 volatility? 7:58 >> Well, as all of our software portfolios 8:00 are deep red, right? All of us have some 8:02 software socks and the reality is the 8:03 public market has decided that since 8:06 code is becoming free uh at the low end 8:10 and becoming much easier to generate and 8:13 create at the high end uh the market 8:15 decided that every software company is 8:16 going to zero. I think this is 100% an 8:19 overreaction because not all software 8:21 companies are created equal and we can 8:22 talk about what the difference are. I 8:24 actually um have spent the last few days 8:26 thinking about the different not even 8:28 last last few months. This is the I mean 8:31 both you and I think about this a lot. 8:32 What are the characteristics 8:34 u of a durable software company? And I 8:37 think there is there's a few that we can 8:39 talk about but yeah I I think it's 8:41 absolutely I think everything has been 8:42 painted with the same brush at this 8:44 point. It is absolutely an overreaction. 8:46 >> You're going to leave me with a 8:47 cliffhanger goal. You're like there's 8:49 some very durable characteristics. We 8:51 can talk about them if we want. I I 8:53 would love it if we could talk about 8:54 them. Can you please help me understand? 8:56 >> It's basically a play on Hamilton 8:57 Helmer's seven parts, but it's slightly 8:59 different. I call it the eight modes. 9:00 The first mode is data mode, which we 9:03 all talk about, but it truly has to be 9:04 proprietary. It has to be data that 9:06 nobody else has access to. I think 9:07 Spotify is a good example. If you look 9:09 at their uh the discover product, it 9:12 uses a decade of listening behavior 9:13 across hundreds of billions of people. 9:15 You can't create that listening that 9:16 that discover product easily. Second is 9:18 the workflow mode, which a lot of people 9:20 argue it's a weak mode. I agree by 9:21 itself. It's a weak mode but the deeper 9:23 you're embedded in the company running 9:25 their operations moving their money the 9:27 the deeper the workflow mode is just by 9:29 itself I don't think it's enough in 9:31 perpetuity but the deeper your embedding 9:33 is for example Netswuite is an ERP that 9:35 runs your business they have a much much 9:37 deeper mode than say Zenesk which is a 9:39 lighter workflow mode so that is a mode 9:41 you can say it's one maybe Zenex is 0.5 9:43 Netswuite is a one third one is 9:45 regulatory mode so licenses uh capital 9:48 require multi-year procurement contracts 9:50 Coinbase when I'm on the board is great 9:52 example. They have MTLS, money 9:53 transmission licenses, state byst state. 9:55 They raise with a phiny CN all of those 9:57 things. It makes it impossible for a 10:00 company to use anybody else and Coinbase 10:02 to custody their crypto because of that 10:03 reason. Fourth mode is a distribution 10:06 mode uh where you have propritor 10:08 exclusive distribution. Intuit is a 10:10 great example. Anybody who wants to 10:12 build an accounting system, think about 10:13 it. Intuit I remember when I started a 10:15 company was after Google. I basically 10:17 tried to use this company called Zero 10:19 XRO and I was like, "Let's use Zero. 10:21 It's like the new it had just started. 10:23 It seems like a cooler interface." My I 10:25 my accountant said, "No, I'm sorry. I 10:27 don't use Zero. You shut it down." I had 10:29 to cancel Zero and go to QuickBooks. 10:31 What a great distribution mode. You've 10:32 trained a network of CPAs to only on 10:35 QuickBooks. I don't know if they have a 10:36 commission or what they get, but that's 10:38 a proprietary distribution channel that 10:40 these guys have. Very hard to displace 10:41 them. Ecosystem mode. If you have a 10:44 platform or ecosystem where many third 10:45 parties have built on and rely on, you 10:48 have a mode. Shopify is a great example. 10:50 You can wipe code an e-commerce hosting 10:52 platform. No problem. But can you wipe 10:54 code the hundred thousands of developers 10:56 and third parties who've built all these 10:58 applications on Shopify? Every Shopify 11:00 merchant I know uses like at least five 11:02 or six other third party apps. That's a 11:04 huge part of the Shopify ecosystem. 11:06 That's a mode. Sixth one is a network 11:08 mode. That's classic Door Dash. I think 11:10 uh do many other modes but AI can vip 11:14 code uh the ability to access 11:15 restaurants but it can't wipe code 11:17 liquidity career density reputation 11:20 history all of those things so 11:21 marketplace density is a network effect 11:24 which is which is structural seventh one 11:26 is the thing you mentioned physical 11:28 infrastructure right atoms wherever you 11:30 have atoms it makes for a mode that's 11:33 hard to displace you can't again I think 11:35 humanoid robots will maybe at some point 11:37 start taking but it's probably a few 11:39 years away and The eighth one I would 11:40 say scale mode. If by virtue of your 11:43 scale your costs are so low that it's 11:46 hard to replicate. I think Amazon is a 11:48 great example. TSMC in semiconductors 11:50 scale modes. So those are the eight 11:52 modes basically uh data um workflow 11:56 regulatory distribution ecosystem 11:58 network physical and and scale. And so 12:01 what you do I think any one of these 12:03 modes is not enough. Uh but what you 12:05 want to do is you want to take a company 12:06 and score it across them. Maybe you 12:08 assign one point to each mode they have. 12:10 And I think anything four or more, 12:12 you're pretty damn secure. But if you 12:14 have a two or three, it's a weak mode. 12:17 And if you have one or less, you 12:19 probably need to really build some more 12:20 modes. So you're not you need to do 12:22 something u to make up for. If you have 12:24 zero, you're you're screwed. Basically, 12:26 I I'm just doing a thought exercise. So 12:27 we have Atlassian and we have Monday. 12:30 They're both down kind of 75%. Um I've 12:33 had both their CEOs on the show. If you 12:35 look at them and you put them across 12:37 this eight kind of rules, 12:40 you would probably say that Atlassian is 12:43 being massively oversold and that 12:45 Monday, as awful as it sounds, is maybe 12:48 being rightly priced in this 12:49 environment. 12:51 >> I agree with that. I think Atllesian has 12:53 you could argue they have proprietary 12:55 data now. They need to use that data to 12:57 build products. They have unique 12:59 proprietary data on all the code out 13:01 there because it's being checked in. We 13:03 there's a lot of stuff they have which 13:05 which they need to use for better. They 13:06 have a workflow mode. Uh they don't have 13:08 a regulatory mode. I don't know about 13:10 distribution mode. Need to think of 13:11 where they have something there. 13:13 Ecosystem mode. I think there's a lot of 13:14 third party things that are built around 13:16 them. So they at least have a score of 13:17 three here. Uh they have a network mode. 13:20 You could do they have a network mode? 13:21 No, I don't think they're not a network 13:23 effects company the way you think about 13:24 it. They have a physical mode and they 13:26 don't have a scale mode. So they have a 13:28 score of three. Monday probably has a 13:30 score of one. I think they have a 13:32 workflow mode. I'm not sure if they have 13:34 the other modes. So, you're right. 13:35 Monday in theory has a much weaker 13:38 um uh score I guess than than Atlacian 13:42 on this. 13:42 >> This is so unfair of me. How would you 13:44 think about Clavio or Clavio in this 13:47 way? Like when you look at bluntly the 13:49 ability for public companies to build 13:51 good agent products, it would seem very 13:52 obvious that Shopify will bluntly build 13:54 Clavio now in the need to reacelerate. 13:58 How would they rate? I don't think 14:00 Shopify will build it. Shopify is an 14:01 investor and Shopify I think has decided 14:04 at least in my opinion that this is 14:06 Shopify has these things called missions 14:08 and I think they decided this is not 14:09 part of their mission to build this 14:10 product. So I don't think the risk is 14:12 Shopify it is that it it has become 14:15 easier to build build Clavio now uh than 14:19 it was you know a year ago. So it's easy 14:22 to build Clavio. They do have I haven't 14:26 talked about brand. I think brand is no 14:27 longer a strong word. I explicitly 14:29 excluded brand. I don't know how strong 14:32 Shopify's promotion of Clavio is. I 14:34 think a lot of it depends on whether the 14:37 proprietary distribution they get from 14:38 Shopify is how strong and tight it is. 14:41 If Shopify is actually going to promote 14:43 them as a you know when you when you 14:45 search for uh messaging or 14:47 communications if they are just like you 14:50 know Google has a morph with Apple when 14:52 you use Apple you basically Apple 14:54 products you get Google search engine. 14:55 If Clavio is a preferred product and 14:57 they have a relationship that makes it 14:58 work, I think it's very hard to displace 15:00 them. It's hard to at least displace 15:02 that part of their business. 15:03 >> Dude, you just throw a grenade in and 15:05 like don't expect me to pick up on the I 15:07 think brand mode is not so relevant 15:09 anymore. I just actually had Elena Verer 15:12 who's the head of growth at lovable on 15:13 our 20 growth show and she said that 15:15 actually brand is the most important 15:17 thing as you commoditize technology and 15:19 it's easier and easier to create. How 15:22 people resonate with a brand is the most 15:24 important. Why do you think brand mode 15:26 is not as important? 15:28 >> Businesses are much more rational uh in 15:30 thinking about it. Less less irrational 15:32 I think uh and the alternatives are 15:34 going to be much stronger. I think on 15:36 the consumer side consumers are much 15:39 more like dollars and there is there is 15:41 dollars and cents but there is a natural 15:43 inclination to just trust brands. I I 15:45 think on the business side it is going 15:47 to get weaker. I think I I actually 15:49 disagree a little bit because switching 15:51 costs are so much lower, right? One of 15:53 Hamilton Helmer's seven powers is 15:55 switching costs. I think switching cost 15:57 is just going to go to essentially zero 16:00 because over the next one or two years 16:02 ability to port data your data as a 16:05 business or consumer from any ecosystem 16:06 to another ecosystem is going to be very 16:08 easy and then people are going to be 16:10 able to replicate almost pixel by pixel 16:13 the experience you have with one product 16:16 in a different product. You'll have 16:17 clones popping up left, right, and 16:19 center? And data portability is going to 16:20 be easy. In that case, what is that 16:23 brand really? It's like in in 16:25 professional sports, you know, you kind 16:26 of cheer for do you cheer for the player 16:28 or the team when they when they switch 16:29 teams. 16:30 >> I need your help because the one that I 16:32 continuously oscillate on is is 16:34 Salesforce. Um, when you say about data 16:36 portability being increasingly easy, we 16:38 had Seb from Cloner on the show. He said 16:39 agents would make data migration from 16:41 systems of record increasingly easy. So 16:43 they wouldn't have the lock in that we 16:45 think or you reduce the switching costs. 16:48 But then I look at your you know eight 16:51 factors and I'm like well they have 16:53 workflow they have distribution they 16:56 have ecosystem they have scale. Well 16:59 they don't have scale they their scale 17:00 means that it is cheaper for them to 17:03 produce. I think they have a score of 17:04 three because that's the thing software 17:06 earlier was a scale game where because 17:08 you had produced a lot of software it 17:09 was cheaper for you to produce a lot of 17:10 software. Guess what? Now everybody can 17:12 produce software as cheaply as anybody 17:15 else. If they had their own data centers 17:18 like the h I think hyperscalers are the 17:20 ones that that basically are able to say 17:23 confidently that they have scale or 17:24 people in the physical world a pure 17:26 software company can't get that scale 17:28 board but yes they are very similar to 17:29 Atlacian where they have a score of 17:31 three I would say. And so do you think 17:32 Salesforce and systems of record like 17:35 Salesforce are inherently attractive or 17:38 less attractive 17:39 given the data portability increasing? 17:42 >> They are more attractive than most 17:44 companies most software companies but if 17:47 they don't build um agentic workflows 17:50 and commoditize a complement by giving 17:52 by figuring out where the profit pool is 17:54 I think they have to figure out is the 17:55 profit pro in the data profit pool in 17:56 the data or the workflows. If it's in 17:59 the workflows, they need to make data 18:00 storage free and basically change 18:02 pricing to an outcome based model based 18:04 on workflows. If they feel the profit 18:06 pool is in the data, then they need to 18:07 give away these workflows for free. And 18:09 so they need to really commoditize all 18:12 the agentic companies at UI and no that 18:14 are trying to build on top of them and 18:16 charge for that. They need to build 18:17 better products using their data and 18:20 make it free. And I think that's the way 18:22 that a Netswuite or a Salesforce system 18:23 record needs to operate. They have to 18:25 commoditize a compliment. they can't 18:27 just wait around for other people to 18:28 build on top of them. 18:30 >> We're seeing buybacks like like never 18:32 before for your work days and your sales 18:34 forces of the world. Is that truly 18:36 indicative do you think of like internal 18:38 company confidence or do you think it's 18:40 a necessity to externally show the world 18:42 that we are confident? 18:44 >> It's both it they are confident 18:46 internally but they need a signal to 18:48 show that they are confident. I think 18:50 the founder buyback is the most the 18:52 strongest signal. It's not just the 18:53 company but also the founder. Did you 18:55 see Service Now CEO's 3 million buyback? 18:58 And then they saw that his garage of, 19:00 you know, classic cars was like three 19:02 times as much. I was just like, "Oh, 19:04 that's a bad comm's move." 19:08 >> Yeah. I think there are there are 19:09 buybacks and there are there are 19:10 buybacks with a capital B. You want the 19:12 capital B one. You want a CEO of a large 19:14 company to do a 20 $50 million buyback 19:17 to show. 19:18 >> I completely agree, dude. We've seen 19:20 we've mentioned Monday. We see companies 19:22 like Notion, like Amplitude. mixing 19:24 publics and privates, but kind of growth 19:27 stage companies and even privates um do 19:30 bolt-on strategies. Hey, our core 19:33 product with a bolt-on of AI. How do we 19:35 determine bolt-on AI strategies that 19:38 work versus bolt-on AI that doesn't? 19:42 >> I think the bolt-on AI, it's an 19:43 interesting thing. The bolt-on AI 19:44 strategy by itself has a real ceiling. 19:47 But I think the companies where the 19:48 bolt-on really works are the ones that 19:51 reframe what the product does, not just 19:53 add the capabilities. So I think you you 19:56 know for example if you just add AI 19:58 search u as there's one thing you just 20:00 add AI search or you build search as an 20:03 experience with new UX primitives I 20:06 think one is just an upgrade the other 20:08 is doing completely something completely 20:10 different. So I feel you got to so 20:13 notion for example I think very highly 20:14 of them. I think notion is adding AI. 20:17 You mentioned they're adding a lot of AI 20:18 agents and I'm hoping that the way 20:21 they've added it, I've actually played 20:22 around with the product. I think it's 20:23 pretty good. But the AI agents need to 20:25 now get better based on how the user 20:27 interact with it and they need to tune 20:29 the model for their customer base. Most 20:31 bolt-on players are not doing it. 20:33 They're simply using a GPT or or or 20:35 anthropic model and they're basically 20:37 just adding a thin layer. You have to 20:40 rebuild the entire experience end to 20:42 end. And I think by you do that by 20:45 identifying something where AI doesn't 20:47 just improve the margin, it changes the 20:50 experience and the economics. Document 20:52 processing was a good example. I think 20:54 Alton recently, you know, you couldn't 20:56 actually extract structured information 20:57 from unstructured documents till about 6 20:59 or 9 months ago. Now you can reliably 21:03 dense legal contracts. So your 21:04 experience around documents needs to be 21:06 fundamentally different. If you're just 21:07 getting someone to upload a document in 21:09 a flow, you need to instantly like give 21:11 them immediate insights from the 21:13 document while they're uploading it 21:15 versus like the same document upload 21:17 thing. That's crazy because now you 21:18 should be able to infer any document 21:20 that enters what the hell is going on 21:22 instantly. So you need to re-evaluate 21:24 every single interaction and say what 21:26 has changed. That's that's the biggest 21:28 difference in product development today. 21:29 you know, model capabilities are 21:31 improving every 6 months because if you 21:32 have too long a product road map, you're 21:34 going around a program and the model 21:35 comes and just blows it out of the 21:37 water. So, you've got to really 21:38 understand what the capabilities are of 21:40 each new generation and make sure you 21:43 can't have too long a road map because 21:44 your road map is going to be blown out 21:45 by the next next model uh iteration. 21:48 >> Speaking of being blown out by the next 21:49 model iteration, how do you as an 21:51 ambassador state educate me? How do you 21:54 ascertain safety from model intrusion 21:58 versus in the way of models and you will 22:01 be eaten with the next update? 22:03 >> If you have some of the other ones which 22:04 are physical and so on, it becomes 22:06 easier. If you're a pure software 22:07 companies, which of those apply to you? 22:09 I think fintech is a good one. I think 22:12 we used to fintech goes through these 22:13 cycles, right? I think fintech, 22:15 especially at Marathon, we invest a lot 22:16 in fintech. We actually think, oh my 22:18 god, fintech is one of the best. If 22:19 you're moving money, you're generally in 22:21 a good place. So anything touches money 22:24 we feel there's a very strong mode there 22:27 much more defensible and so data and 22:29 workflow modes are the two things you're 22:31 really hanging your hat on as a software 22:32 investor because like if you're not 22:34 doing fintech um and then I think it 22:37 early stage it's too hard to know what a 22:39 distribution mode is unless you have 22:40 some hack and these hacks never really 22:43 stand the test of time ecosystem too 22:45 early to say network effects too early 22:46 to say so really physical infrastructure 22:48 they don't have any software company 22:50 scale they don't have any so it's really 22:52 about okay go deep into what is the data 22:54 asset you're creating does it get better 22:56 with time do I believe it get better 22:58 with time are you building your own 23:00 model over time are you fine-tuning a 23:02 model and improving it over time and 23:04 then how deeply are you truly embedded 23:06 in the workflow how deeply are are you 23:08 just a lightweight thing that the 23:10 underlying system of record could create 23:12 or are you building so the so these are 23:14 the two things that you have to hang 23:15 your hat on a data asset that gets 23:17 better with every interaction and then a 23:20 workflow it's hard And I think pure 23:22 software companies are hard. I think 23:24 application software companies are hard. 23:25 You got to really believe that the 23:27 founders can ship with great velocity uh 23:29 to basically build that um and see proof 23:31 of that compounding. 23:33 >> I got in trouble in the partnership the 23:35 other day. I'm quite grumpy generally 23:37 speaking. Um but yeah no yeah know my 23:41 Twitter is getting grumpier and 23:42 grumpier. I know. Uh anyway um but you 23:45 know I've met like you know support 23:47 agents you know voice agents for auto 23:51 manufacturers for dentists for 23:53 chiropractors for the and I'm like guys 23:56 this is like open AI 11 labs for and 24:00 then all of these different verticals 24:02 would you say Harry no no no you're 24:04 wrong they're building verticalized data 24:07 um over time uh they're able to 24:09 fine-tune their own they are deeply 24:11 embedded in workflows 24:13 and they might have distributionary like 24:15 yeah the kind of like dentist cool agent 24:18 is a little bit plaster on top of a 24:21 wound. 24:22 >> I do think these are viable businesses. 24:24 I don't think they're going to be big 24:25 businesses. Um I think as soon as you do 24:28 vertical I don't think you can do a a 24:31 one function within a vertical. I think 24:34 what you want to see there is the 24:35 ambition and ability to truly own the 24:38 full stack, build the whole product for 24:40 the vertical. Service Titan, for 24:42 example, is a canonical example. It went 24:44 public last year, great outcome, is 24:46 still a sub10 billion company or 24:47 something like that. And they own like 24:49 30 they have, if you look at their S1, 24:50 they have 32 different products. And 24:53 even after selling 32 different products 24:54 and really being at least in the US for 24:56 any service pro service like phys field 24:59 services company, they are the canonical 25:00 company, they still are a $10 billion 25:02 company. And I think uh 25:04 >> you know what's you know what's 25:05 astonishing when you compare that to a 25:07 Robin Hood is Robin Hood has 13 product 25:10 lines now doing over 100 million in 25:12 revenue. 25:12 >> 13 Coinbase has 12 doing 100 million 25:15 revenue. Exactly. You're a horizontal 25:16 product. You're serving a broad base. I 25:18 think vertical products you've got to 25:20 really own full stack. I think it's 25:22 harder otherwise to be a 10 plus billion 25:24 dollar company. 25:25 >> I'm going for spice in a world of 2026. 25:29 Can we as venture investors do vertical 25:31 SAS given the fund sizes that we have? 25:35 >> I think you can. Maybe the mega funds 25:38 might say look it might not be $100 25:39 billion outcome. Uh but I think uh if 25:42 you're a if you're a 2003 $300 $400 25:46 million fund, you can absolutely create 25:47 a $10 billion company because remember 25:49 one of the big changes is that vertical 25:52 SAS does take over labor. Um and so 25:56 vertical software, right? It's no longer 25:57 SAS. It's basically idea software as a 25:59 service but it is services. So you're 26:02 going after the services spent. So one 26:04 of the interesting things as you know is 26:05 that verticals mostly especially if 26:07 you're selling to small businesses they 26:09 spend some amount of tooling but they 26:11 spend a tremendous amount on both BO as 26:13 well as on human capital human labor. So 26:16 you need to basically target those two 26:18 spends and I think if you do that and 26:20 you're committed to building the whole 26:21 product you can. Absolutely. 26:23 >> You very kindly said before the show 26:25 that you like the show that we do with 26:26 Jason and Rory. Uh it's very humbling 26:28 when I do the show for 10 years and then 26:29 I find out that it's actually much more 26:30 popular when I actually bring other 26:32 people on to do it instead of me. Uh 26:34 always good for the ego. But Rory always 26:36 says to me with AI, very simple, we need 26:39 to see the transition of spend from 26:40 software budgets to human labor budgets. 26:43 And if we do, the TAM obviously opens up 26:45 immensely. 26:46 >> Do you think we will realistically see 26:48 that 26:49 >> and maybe are seeing it already or do 26:51 you think we will actually remain in 26:53 software budgets as we have been in some 26:55 categories? No, we are seeing that. We 26:57 are seeing that. I think the first one, 26:58 most businesses don't want to lay off 27:00 people. So the way we are seeing it, the 27:02 first thing that's happening is 27:03 businesses are outsourcing to third 27:05 party BPOS. Many of them in India, 27:06 Philippines, etc. That spend is the 27:08 easiest to cut because now you can offer 27:10 the same service higher quality, faster, 27:13 and 20 30% cheaper. The second thing 27:16 they do is when somebody leaves, they 27:18 don't replace that person. 27:20 >> And the third thing they do is layoff. 27:22 So I think layoff is still maybe a 27:24 little bit of while away but you're 27:25 seeing absolutely BO spend. All the call 27:28 center companies that you mentioned all 27:29 the next generation AI um AI customer 27:32 service companies they're going after BO 27:34 budgets. I was shocked when I was doing 27:36 work in this space. How many different 27:38 verticals doctor's office etc use call 27:41 centers outside the US. They already 27:43 have budget clearly allocated and 27:45 there's a better service. So, so I think 27:47 it's BO spend first, don't replace the 27:49 person second, and then potentially 27:51 think about uh laying off. Um, and 27:54 >> do you know Goldman Sachs and Barclays, 27:56 both financial institutions, both have 27:58 over 30,000 people in India? 28:01 >> I didn't know 30,000. I thought there 28:03 was like a few thousand. I didn't know 28:04 30,000. 28:05 >> Isn't that nuts? I was so shocked when I 28:08 heard about that. I I I want to 28:09 understand, but two different types of 28:12 company profiles and what happens to 28:14 them. We've got private companies. Um, 28:16 and I don't want to pick on them, but it 28:18 is helpful to give examples. I'm sorry. 28:20 They're my friends as well, so I can 28:22 kind of do it and they'll love me 28:23 hopefully regardless. Um, but like you 28:26 know your sneaks, amazing security 28:28 business that's got great customers who 28:30 love it, but it was valued at $7 billion 28:34 and it's now 300 millionaire growing 28:36 15%. What happens to that cohort which 28:39 is a great business and serving great 28:42 customers but 15% growth 300 million AR 28:45 and you've got a very high price. What 28:48 happens to that private cohort? 28:50 >> There are two outcomes for these 28:52 companies. All of us have those 28:53 companies in in our portfolio. A bunch 28:55 of them are going to become zombie 28:57 companies and uh they're going to try to 28:59 add AI features as a last resort, not 29:02 succeed and uh and be sold to PE. Even 29:05 the problem is even that might not be a 29:06 good outcome. know because PE itself is 29:08 struggling to digest the companies they 29:10 bought a couple of years ago and the 29:13 prices are resets they do have good 29:15 assets u you know I think they might I 29:18 am seeing in some verticals there are 29:20 companies merging uh with each other um 29:23 I think we'll see if that happens just 29:24 to create more scale but it it'll be 29:27 interesting but the hopefully the better 29:29 outcome that many of them go to is with 29:31 strong leadership you basically can burn 29:34 the bridges and create a completely new 29:37 AI native product. I think you had the 29:39 intercom person uh intercom CEO right 29:41 Finn great example podium another great 29:43 example both of them with their new 29:44 products have grown to 100 plus million 29:46 in a couple of years and basically just 29:48 burn the bridges this is legacy software 29:50 I think the more you fixate on how do we 29:52 fix the business the less you're going 29:54 to focus on how do we create a new 29:55 business so we've got to create a new 29:56 business from scratch you have customers 29:59 you almost got to say I'm going to be 30:01 ruthless about migrating the customers 30:03 from the current business to the new new 30:05 product even if it's even if it it's 30:07 lower price. It's a bright thing to do 30:09 and you got to have a sunk you got to 30:12 abandon sunk cost fallacy. 30:15 How help me on podium that 100 million 30:17 agent revenue. Okay, it triples 300 30:20 million and then it triples again 900 30:23 million. If the price today is five 30:26 billion that I'm paying, I'm I'm paying 30:30 for two years of treble treble ahead of 30:33 time for that asset for that's what it 30:36 would be priced in public markets for 30:39 this business to work. We need the 30:41 multiples in Publix to be way more than 30:44 they are now. Do we not? I think you 30:46 need to assume that they will take over 30:50 uh huge parts of the service budget in 30:52 the businesses and that they will not 30:54 just be a billion dollar company. I 30:56 believe that they'll be a multi-billion 30:57 dollar company because earlier I think 30:59 they were limited to one part of the 31:01 stack and they were f they were on top 31:03 of a bunch of systems. Now they're 31:05 taking over the entire software stack 31:06 and that's the thing I like about you 31:08 want founders who are ambitious enough 31:10 to go after the entire stack not just 31:12 the earlier piece of the stack they were 31:14 in. and you want to be the only product 31:17 uh that the company uses and you want to 31:19 replace as much of the digital labor as 31:22 you can possible. That's the ambition. 31:24 So what you want to say is what's your 31:26 market size here? How much in all your 31:28 customers? How many people that do they 31:31 have that are doing digital work and do 31:33 you have the ability to replace all of 31:35 that payroll over time and all of the 31:37 other things they're doing and take a 31:39 part of the payments transaction 31:40 revenue. And if you think that's a big 31:42 enough opportunity, that's when you 31:43 invest. 31:44 >> Do we see the total death of seat 31:46 pricing, my friend? I hear you 31:47 completely in terms of that movement 31:48 into services. Does seat pricing die and 31:51 we actually have consumption based 31:52 pricing as the primary pricing 31:54 mechanism? 31:54 >> Seat pricing doesn't die. You know why? 31:56 If you look at Chad GPT enterprise, Chad 31:59 GP enterprise is priced based on uh 32:01 based based on seats because seats 32:02 provide predictability for enterprise 32:04 buyers. U but they don't drive expansion 32:08 revenue by themselves. So you basically 32:10 have to bundle a lot more into each 32:13 seat. So Chat, Jupyter, OpenAI sells 32:16 seats based on different tiers where 32:18 they have different functionality. I 32:19 think Figma sells three different types 32:21 of seats. So you're going to see 32:23 different kinds of seats. Now the big 32:25 challenge is seatbased pricing which you 32:26 alluded to is it breaks when the 32:29 product's core value is not about access 32:32 uh but it's about uh something doing the 32:35 work on your behalf. So at that point 32:37 charging per user doesn't make sense 32:39 because user isn't the constraint 32:40 anymore. It's a work output. So at that 32:43 point you got to go to outcome based 32:44 pricing. So for example if I'm something 32:47 like Harvey I don't know how Harvey 32:48 prices I bet that they price based 32:50 purely on how many contracts uh they 32:52 process versus how many people are using 32:54 it. For example even if 100 people are 32:56 using it and they process zero contracts 32:58 in theory they should get zero. So I 33:01 think you have two kinds of products. 33:03 You have access products and you have 33:06 work products. Access products um is 33:09 basically seat based uh like I think 33:11 Chad GB enterprise is a good example and 33:13 then work products like Harvey are 33:15 probably more outcome based and not seat 33:17 based. 33:17 >> You mentioned Harvey there we are seeing 33:20 increasing competition within certain 33:22 categories. If I think about you know 33:25 law it's Harvey and Lagora. If I think 33:27 about customer support, Sierra and 33:29 Dakagon and there are dominant funded 33:32 players. How do you think about the 33:35 ability for firms to kingmake? Is 33:37 kingmaking complete [ __ ] Is it not? 33:39 I'm just intrigued to get your thoughts 33:41 on that. 33:42 >> Kingmaking is a thing. I think it is a 33:43 thing. You see uh earlier than earlier 33:46 companies are getting these rounds that 33:49 are valuing them, you know, at at 33:51 valuations which really are eye opening. 33:54 That said, I think it won't work unless 33:57 the the company executes on the promise. 34:01 I think uh other firms can take it as a 34:03 signal and decide to pile on or not and 34:06 you know but but ultimately the the 34:08 company has to execute on the vision. If 34:10 it doesn't then it's just a bad bet. So 34:12 yes, it is there but it by itself just 34:15 because you're kingm doesn't mean they 34:16 are the king. They still have to execute 34:18 and justify it. If not, then I think the 34:20 reality is everyone's playing different 34:22 games, right? You and I know being 34:23 venture capitalists now that, you know, 34:26 somebody with a $10 billion fund is 34:28 playing a fundamentally different game 34:29 than somebody with a $400 million fund. 34:31 Um, and if you try to play the same game 34:34 they are, you're going to lose. You got 34:35 to play the game that you're best 34:36 equipped to play. Benchmark plays a 34:38 different game than than say Andre 34:39 Hollowitz, but both of them play 34:41 different games and both of them do well 34:42 at their game. 34:43 >> We mentioned podium earlier and going to 34:45 100 million with their agent first 34:46 product. The growth is incredible and 34:48 the growth across this cohort of 34:50 companies is dude we were doing this 34:52 eight years ago when you know you went 34:54 from 1 to 10 at Slack and it was like 34:56 holy [ __ ] that's amazing. Now it's like 34:59 1 to 10 is is still great but there's 35:01 quite a few who've done 1 to 10. How 35:04 does your mindset change around growth 35:06 expectations for the companies that you 35:08 invest in? Is a world of triple triple 35:10 double double dead. It's not dead, but 35:13 it you it's no longer elicits the 35:16 jaw-dropping uh awe that it used to a 35:19 few year few years ago. As you said, I 35:20 mean, you know, the 1 to 10 is becoming 35:23 more and more common. Uh and and you 35:26 know, those numbers will basically get 35:28 you I mean, lovable could probably go 35:29 public with that kind of trajectory. 35:31 Now, the bigger question for me is uh 35:34 durability. Um, and it's not even 35:36 quality, it's durability. Because like 35:38 you and I discussed, margins can improve 35:40 over time and will improve over time. So 35:42 it's not about margins, it's about is 35:44 this revenue durable? And so retention 35:47 is is basically very important for me to 35:49 understand are people using it as I 35:51 think we saw in the first wave of AI we 35:53 saw many uh chat GPD like there was a 35:56 company called Jasper not to pick on 35:57 them but they went from 1 to 40 and then 35:59 they came back from 40 to 10 or 36:01 something like maybe 1 to 100 and 100 to 36:02 40 within very quick time frames. So 36:04 there's a lot of tire kickers out there 36:06 especially in proumer products who test 36:08 the product and then move on to 36:10 something else. So what you want to look 36:12 under the hood beyond behind all these 36:13 numbers is two things which I think are 36:16 the fundamental indicators of business 36:18 quality. Customer retention or gross 36:20 retention and then net revenue 36:22 retention. Um basically if those two I 36:24 think are the biggest indicators of 36:26 quality. I would always take a company 36:28 that's just just I should say just as 36:30 crazy doing a triple triple double 36:32 double with excellent gross and 36:34 excellent net revenue retention than a 36:36 company that's growing 10x in a year 36:38 with really bad customer retention and 36:41 less than 100% or less than 90% net 36:43 revenue retention. 36:44 >> How do we think about ceilings on those 36:46 markets? And I'm specifically thinking 36:48 about one that haunts me which is 36:50 granola. I was one of the first 36:51 ambassadors to meet Chris and clearly 36:53 granola has crushed it and is an amazing 36:56 product but customer retention skyhigh, 36:59 revenue retention skyhigh. 37:02 Honestly, if anthropic or open AAI did 37:05 an enterprise product that's not taking 37:07 and it's connected to all of the 37:08 different suite of products that they 37:09 have, I think that heavily threatens the 37:11 market size that Granola is able to 37:14 expand into into large enterprise. How 37:16 do you think about the worthiness of 37:18 those retention numbers if there are 37:21 alternative factors like that that could 37:22 impact it? 37:23 >> Yeah, I think you want to keep you want 37:25 to basically weigh the retention in the 37:27 light of what they have encountered. 37:29 You're absolutely right. I think you 37:30 want to see just like you want to weigh 37:32 the growth of a company uh in in light 37:35 of have you gone through any seismic 37:36 events. If they've not gone through any 37:38 seismic events, you've got to then take 37:39 it with a grain of salt. What 37:41 comparative threats have you faced? Has 37:42 a single competitor come out? Have you 37:44 been able to ward off that? has your 37:45 retention stay strong in light of that? 37:48 I do think I think some of these some of 37:50 these products like granola etc have uh 37:53 we'll see if they are the case but they 37:55 are these unique products that really 37:58 open up non-conumption markets which 38:00 means that I would never have actually 38:02 bought a noteaker before uh a separate 38:05 noteaker outside of zoom or something 38:07 cuz zoom comes with its own noteaker gym 38:09 it comes with note taker but gorilla is 38:10 so powerful that it basically got me to 38:13 consume a separate note takingaking 38:15 product and it's probably true for many 38:16 of us and and I I think that just 38:18 changed the market opportunity for them. 38:21 I think Uber and so on are great 38:23 examples where they just saw 38:24 non-conumption markets 38:26 >> also do gamma you've got Google slides. 38:29 >> Exactly. Great. The very good parallel 38:32 non-conumption market you would never 38:33 assume. Why would you ever buy a 38:35 PowerPoint thing or a presentation thing 38:37 separately? It's a non-consumption 38:39 market. It's a zero market. But the 38:40 product is so good, so remarkable that 38:43 it gets people to to buy it separately 38:45 as a separate SKU. Power to them, more 38:47 power. I think we need more of these 38:49 standalone remarkable products. Intuit 38:50 is a great example. As you know, 38:52 Microsoft tried to crush into it again 38:53 and again and again back in the 80s and 38:55 '90s with bundling everything into 38:57 office. But in Turboax survived and 38:59 thrived. 39:00 >> Okay. But does that go against what you 39:02 said earlier about the need to be 39:03 multi-product? You know what you've done 39:05 with Gamma is you've taken slides out of 39:07 Google's G Suite and made it on steroids 39:10 amazing very deep featurerich in a way 39:13 >> multi product they can't just be a 39:15 single product and go they will need to 39:16 have a second product I'm sure of that 39:19 they will need to have a second product 39:21 in it has multiple products 39:22 >> okay but what's what's grod and gamma's 39:25 multi-roduct 39:26 >> I don't know why gamma would not create 39:27 a s just like they've they have taken 39:30 their riff on powerpoint why can't they 39:32 have their own take on documents or 39:35 slides the same way. I have to assume 39:37 it's basically the different kinds of 39:39 content that people create or even 39:40 websites. 39:42 >> It's hard to see. Um 39:44 >> we'll see. 39:45 >> Both both of them are in a wave of 39:48 incredibly hot attractive companies 39:50 which have lower margins than we are 39:52 used to in traditional SAS mines. How 39:55 has your mindset changed or stayed the 39:57 same around margins? How should I think 40:00 about margin assessment when looking at 40:02 companies today? Yeah, I think uh 40:04 inference costs are dropping. So you 40:06 automatically uh assume that margins in 40:09 theory should go up. Uh but I think it's 40:12 it's not about margins in year 1 or two. 40:15 The more defensibility or leverage you 40:18 have u in some ways over your customers 40:20 and what choices they have. Uh the more 40:22 pricing leverage you have. So I would 40:24 rather see margins go up with price 40:27 increases than cost decreases. A good 40:30 example is PayPal rolled off both of us 40:33 on our board at Square and he told us 40:34 that PayPal back in the day raised 40:37 prices five times in 3 years because 40:40 there's such stickiness. They knew their 40:41 customers really couldn't do anything. 40:43 And you see I mean those are I mean Uber 40:46 I have to say I don't know how if they 40:47 have raised price or not but I know that 40:49 they've basically changed the economics 40:50 of how much they pay drivers over time 40:52 so that their their margins have just 40:54 expanded continuously and they've also 40:56 raised prices in different ways. So I 40:57 think you two ways of increasing 40:59 margins. So first of all you and I both 41:02 we don't look at margins in year one and 41:03 two. I mean it doesn't make sense even 41:05 years four and five. But you want to 41:07 have on one side the ability to increase 41:09 prices. On the second side you want 41:11 ability to cost to get lower. I think 41:13 that second thing is happening by by 41:15 nature. What you want to see is in 41:17 addition the ability to have such a good 41:20 product and ideally multi-product a 41:22 story which makes switching really hard 41:24 and you can raise prices. Can I be 41:26 blunt, dude? In the environment that 41:27 we're in today, the two things that you 41:29 said there, ability to increase prices 41:31 and margins in years 3, four, and five. 41:34 Dude, the world is changing so much. I 41:36 have no idea about their margin 41:38 structure in years 3, four, and five. 41:41 It'd be poetry. 41:42 >> You don't you don't look at margins. You 41:44 look you look to see whether or not they 41:45 have they have a product that is 41:47 compelling enough. For example, I think 41:49 obviously Disney Plus is a Disney Plus 41:51 for every year. I think they increase 41:52 prices on me. like they've gone from $20 41:56 to 30 to 40 or something like that. 41:58 Amazon Prime is another one. But this 42:00 takes many years. But you want the 42:01 potential. You want to evaluate the 42:03 potential. Do they have the potential? 42:05 Do they have the ability to increase 42:06 prices in the future? 42:07 >> My dear friend, if you have 42:08 >> margin is much less, it's not something 42:10 I worry about. I think durability and 42:12 defensibility is much more of a worry. I 42:14 think good companies if they're 42:15 defensible, they will have the ability 42:17 to increase margin. That's what I was 42:18 saying actually. 42:19 >> Have you ever shopped in Chanel Gawkle? 42:22 >> Uh 42:23 I I have gone into I think they do have 42:26 stores or mini stores. Yes. I've never 42:28 shoed myself for life as Yes. 42:30 >> I buy my mother every year Chanel 42:32 handbag for Christmas and birthday. Do 42:34 you know what they do every 6 months? 42:36 Prices up 10%. Every 6 months. 42:40 >> 10%. 42:41 >> Dude, 42:42 >> 10% for the same product. 42:44 >> I I used to buy a handbag and it was 42:45 £500. Now it's £10,000. 42:48 >> Holy cow. 42:49 >> My question is how we should invest in 42:51 LVMH. Clearly um 42:54 >> durable, defensible 100 plus years, 42:56 right? 42:57 >> 100%. And dude, in a world of increasing 43:00 wealth inequality, actually 43:03 awful statement. Can you have good 43:05 businesses selling to 43:09 non-wealthy people? You've worked in 43:11 fintech before. You 43:12 >> I think Robin is a good example. I think 43:14 you've got to have massive scale. I do 43:16 think a u I'm an investor called Atlas, 43:19 which sells to billionaires, is the 43:21 opposite. It's a great business. It just 43:23 you you need massive scaling. You need a 43:24 veg product that is almost cheap or 43:27 free. I mean, Robin Hood, you have to 43:28 have free canonically in your thing. 43:30 Robin Hood obviously offered free stock 43:32 stock trading for a long time and that 43:34 was their core pitch and that allowed 43:36 them to basically get a lot of people. 43:38 You need to have something free or some 43:41 hook that is really low cost that allows 43:42 you to expand. But it is a harder 43:44 business because you can't make the 43:46 arpoo or the average revenue per user is 43:48 low enough that you need millions if not 43:50 tens of millions or hundreds of millions 43:52 of people. When you sell to very rich 43:54 people or wealthy people or large 43:55 enterprises look at Palanteer which is 43:57 the business equivalent of selling to 43:58 wealthy people they have I think what 44:00 less than a thousand customers maybe 44:02 even less and each customer pays them 20 44:04 million or 30 million or 100 million or 44:06 billion. It's an easier business to 44:09 obviously it's a very hard business but 44:11 you can see I like those businesses. I 44:13 mean it's uh look at Viva they sell they 44:15 went public with four customers four 44:17 customers. Do you bother to do market 44:19 sizing today given the transients of 44:21 markets? As you said there, some of the 44:23 best companies make you pay for things 44:25 you never thought you'd pay for. Do you 44:27 bother to do market sizing? 44:29 >> I think non-conumption is the biggest 44:30 challenge. But yes, you can't not do 44:32 market sizing. I do bottoms up with a 44:35 specific segment and I always know I 44:37 think any customer base that has more 44:40 than 10,000 customers or a few thousand 44:42 customers, you got to segment them. 44:43 There'll be a few different segments. So 44:45 you want to understand within each 44:46 segment what the bottoms up propensity 44:49 to pay is. What's the problem you're 44:50 trying to solve with them? And then 44:52 you've got to talk to them to understand 44:53 what the budget is. You got to do the 44:54 work. 44:55 >> What's your biggest misread on market 44:57 size and how 44:58 >> Shopify? I remember seeing Shopify at a 45:00 billion and I was like how many how many 45:03 e-commerce merchants are there really? 45:05 Um and that or maybe even before a 45:07 billion one of the early rounds. So Tam 45:08 felt really concerned. I think what I 45:10 missed was that Shopify was not just 45:13 selling e-commerce. It was basically 45:15 allowing anybody to sell. So it 45:17 basically changed any entrepreneur on 45:19 the planet, anybody who wants to sell 45:20 something went. So it it wasn't just 45:22 existing e-commerce merchants. And 45:24 that's what you want platforms to do. 45:26 They literally make it possible for 45:28 every person every person to think of 45:30 the possibility of selling or renting 45:32 their home out or taking a ride which 45:34 they never would have thought before or 45:35 or installing a buying a new 45:37 presentation app or a note-taking app. 45:40 They are the biggest hits. They're also 45:41 the biggest misses. If the bet doesn't 45:42 play out, they're screwed. The bet plays 45:44 out, they could be bigger than anything 45:45 else. Google non-conump. Many of them 45:47 are non-conumption market. They're new 45:48 behaviors that didn't exist before. 45:51 That's in some ways what venture is all 45:52 about. It's not about existing. It's 45:53 about new behaviors and betting on them. 45:56 Facebook non-conumption market. I mean, 45:58 right? I mean, think of all of these 46:00 iconic companies. 46:02 >> The thing that's amazing with Facebook 46:03 is the ease for you to dismiss it for 46:05 being the 52nd social network. I mean we 46:09 forget now that Franster and MySpace and 46:11 everything before it had been bluntly 46:14 there had been so many and 46:16 >> the biggest difference was identity and 46:18 friends and MySpace you didn't know who 46:20 the people were they didn't have the 46:21 real photo they could come up with their 46:23 thing I remember when Facebook had to go 46:25 into Japan Japanese cultural norms were 46:27 that all the Japanese social networks 46:28 back then were incognito you couldn't 46:30 for some reason maybe saving face or 46:32 something you could not share um your 46:34 your real name or your photo so everyone 46:37 said Facebook, you've got to adhere to 46:38 Japanese cultural norms. You've got to 46:40 change Facebook and make it similar. 46:42 Mark said absolutely not. Even if it 46:44 takes us longer and they they succeeded. 46:47 They succeeded. 46:48 >> How do you prevent prior wins or losses 46:52 impacting future decision making? My 46:54 biggest mistake is I lose or make money 46:56 in a market and it inherently makes me 46:58 attracted or not attracted to it in a 47:01 way that could subvert decision- making. 47:03 How do you avoid? 47:04 >> Yeah, this is it's very hard. I think 47:05 it's a mental thing where you've got to 47:06 take every every opportunity at first 47:09 principles. We all struggle with it. I 47:10 think the I think the best venture 47:13 capital I someone asked me what's the 47:14 best venture capital bets. I talk about 47:16 a paradoxical one. I think it's Mike 47:17 Murit's betting on Instacart. Why? 47:20 Because he lost 370 million on Webban 47:23 less than a decade ago. He burnt it 47:25 through. Same space Apura comes to him. 47:28 He bets on it. He bets on it after 47:31 losing hundreds of millions of dollars. 47:33 It is not all sequent 47:36 to the ground. And think about the first 47:38 place he was thinking needed and the 47:40 courage needed to make that bet. I think 47:43 it's brilliant. 47:43 >> You got you've got to laugh at being a 47:44 Sequoia partner. You're going, "Dude, 47:46 not this [ __ ] again. Come on, Mike." 47:48 Like 47:50 really. I'm so curious to see how he 47:52 like just incredible. 47:54 >> You're like, "We know you did Google and 47:56 like we love you, but come on. Not food 47:59 delivery again. 48:03 online grocery shopping. Exactly. 48:05 >> Dude, market is one way we trip 48:06 ourselves up. Oh, market's too small. 48:08 Market's too small. The other one that I 48:09 always make mistakes on is price. Uh, 48:12 how do you think about when you reflect 48:14 on you've done so many good deals? Are 48:17 the best deals the most expensive in 48:19 your experience? 48:20 >> I think there are two ways I've now 48:22 realized after many years of doing this 48:25 at CDNA, price almost doesn't matter if 48:27 you're right about the company. So, I 48:29 think you just invest at whatever the 48:30 price is. I for example I invested in 48:32 the seed round of fair at back in the 48:34 this is about 8 nine years ago 20 48:36 million which is very expensive for a 48:38 seed round at that time it was the 48:39 highest price YC deal at that point I 48:41 think it's been a 100 or 200x for me so 48:44 I think you you invest in a company 48:46 which is great you have conviction you 48:47 invest now I think the B I think B+ 48:50 that's when price starts destroying 48:52 returns I think by then you got real 48:55 revenue real traction you got to you can 48:57 pick a generally good company and still 48:59 get crushed 49:00 For example, one of my friends invested 49:02 in this security company at basically at 49:05 uh when they had 100 million in revenue. 49:08 He invested in them at 4 billion. I 49:11 think they've gotten to 500 million in 49:13 revenue, but guess what? They're still 49:15 at 4 billion. Uh and so basically, they 49:18 will not make 1x the capital they 49:20 invested. And so that's a challenge I 49:22 think. But guess what? Even in VW, 49:24 Benchmark made money. Benchmark made 49:27 money at Weiwork because they invested 49:29 early enough. So I think at subund 49:31 million and maybe that's an arbitrary 49:33 number you can if the company is good 49:35 you'll make money regardless. 49:36 >> You mentioned we work there. We'll get 49:38 to selling because I used it in an as an 49:40 example in a show we did with Miles 49:41 Clements from Excel. Um but I just want 49:44 to touch on like the A market there and 49:46 you saying about pricing kind of where 49:47 it matters where it doesn't. I'm with 49:49 you 100%. But we're seeing 100x ars for 49:52 3 million revenue companies and they're 49:54 being priced at three $400 million in 49:57 this new environment. How do you advise 49:59 me as a series A lead investor to 50:02 operate in a market where A's are not 10 50:06 to 20 now on 100 to 150. They're 50:09 actually 300 to 500 and 30 to 50 million 50:13 rounds. 50:14 >> I don't think an a a investor can do. 50:17 There are two kinds of deals that a 50:18 investors have to do. One is I think 50:20 where there is less legibility on the 50:23 company where you're betting on there is 50:25 some early product market fit. There's 50:27 not 3 million revenue. There's half a 50:29 million revenue and that's like you said 50:31 when you get it basically for 50 or 50:34 subund million a million or so. But then 50:37 as soon as it gets to three or four, it 50:38 gets you maybe you can do a couple of 50:40 deals like that. But I don't think you 50:42 can build a series A fund doing deals at 50:45 three or 400 million. Uh because it's 50:47 not going to be enough ownership. These 50:48 some of these are going to fail, etc. 50:50 But most deals I think you you've got to 50:52 you got to have double digit ownership. 50:53 You got to invest slightly earlier. It's 50:55 a tough it's a tough game, but got to be 50:57 patient. I think the good news is like I 51:00 said I think uh it's concentration is 51:02 your friend in some ways but it can be 51:04 your friend because then you don't feel 51:06 that you got to do 10 deals a year you 51:08 can do four deals a year and do 15 51:10 companies in the portfolio 51:11 >> dude you've got 250 million bucks in the 51:13 fund you've got 200 million when you 51:15 actually look at investable cash if you 51:17 don't have any reserves you've got say 51:19 10 20 million series A checks if you're 51:22 wanting to get ownership double digits 51:24 that we all say that we 51:27 Is that enough? I'm not being cynical. 51:29 I'm asking for my own advice. Is that 51:31 enough? 51:31 >> You have to have reserves. We have 35% 51:33 reserves. So, you have to have a mix. 51:35 No, I've got Come on. We need a bigger 51:38 fund. This doesn't work. 51:40 >> You've got to have a mix of You got to 51:41 have a mix of seed and incubation bets 51:44 and a mix of series A bets. So, the 51:46 incubation bets are bets you take on 51:48 founders who are basically the best in 51:50 the world at what they've done. A good 51:52 example, I think, was um invested in a 51:54 company. This was before marathon but 51:56 with my marathon partners who were part 51:58 of the marathon they led the round with 52:00 venode venode coastline and Mickey 52:02 Malcad ribbit and a company called lead 52:04 bank which was my colleague Jackie 52:06 Reese's 52:06 >> yeah I I interviewed her she's amazing 52:09 so that was inception round at like some 52:11 crazy valuation I mean very strong 52:13 valuation not a crazy valuation but a 52:15 strong valuation because Jackie was 52:16 Jackie and she had built the bank at 52:18 square she built this bank again so you 52:20 want somebody in a industry where they 52:22 know the inner workings of better than 52:23 anybody else in the world and you say 52:24 you back those people and that's a much 52:27 much better riskreward there. So you 52:29 want to do a few of those in addition to 52:31 you want to do a few incubations and 52:32 seed in addition to u in addition to 52:34 series I think you're doing that right 52:36 in some ways I think you got to mix it 52:37 up as an early stage fund I think every 52:40 firm I I think every early stage firm I 52:43 think you got to have an access to 52:44 founders and they you're their first 52:46 call when they go to start a company and 52:48 then you meet founders you don't know 52:49 them before you're kind of betting on 52:51 traction and so on you got to have a mix 52:53 of both kinds 52:54 >> do you buy the proprietary found again 52:57 this is where I get grumpy as [ __ ] but 52:58 I've done 3,000 shows, dude. You know, 53:00 at some point you have to get cranky. 53:02 Like every venture investor sells the 53:04 >> There's no proprietary founder access. 53:05 What is proprietary is your ability to 53:07 add value. And I think founders, you 53:09 basically have to I think build if 53:12 you're just capital and assuming 53:14 founders will come to you, you're not 53:16 going to win. But what you have to offer 53:17 them is something. What you offer them 53:18 is something very different and unique. 53:20 What I offer them is something very 53:22 different and unique. We all have I 53:23 think you got to hone as investors. What 53:26 is it that we're offering? Is it 53:28 council? Yes, that's free. Is it a Is it 53:31 distribution? You offer incredible 53:32 distribution. Is it a network of 53:34 customers that you can get them access 53:36 to? Is it like talent hiring? What is it 53:39 they can do? 53:40 >> Do the best founders need you? Keith Ro 53:42 always says the best founders do not 53:44 need a venture investor's help. You've 53:46 worked with the best. How do you feel? 53:49 >> They may not need it. I I don't know. I 53:51 think uh in I generally agree with Keith 53:54 that on the margins investors don't add 53:57 value and the value they add gets less 53:59 and less as a company grows. But I do 54:01 think there are a few points where a few 54:05 things you can do on the margin. For 54:07 example, helping them choose between 54:09 this candidate or that when they're 54:10 hiring, helping them think about go to 54:13 market that could make a difference 54:15 between the company being a mediocre 54:17 exit or an outcome or being a generation 54:19 company. You don't need to do everything 54:21 for them, but just those one or two 54:23 things that you can help on in on the 54:24 margins can hopefully be the difference. 54:28 >> What would be your advice to LPs when 54:31 they are consistently sold by GPS like 54:33 me and you proprietary founder access? 54:36 Oh, I have the best network. I'm a super 54:40 smart fintech expert, so I know it 54:43 better than anyone. What would you 54:44 advise them on manager selection when 54:47 everyone says proprietary founder 54:48 access? 54:49 >> Very simple. Go and talk to the 54:50 founders. Go and talk to the founders 54:53 and see why they chose especially the 54:55 earliest stage founders. Uh go to the 54:57 seed founders that they invested in and 54:59 the inception stage founders and ask 55:01 them what was different why did you 55:02 choose them? What are the options you 55:04 have? And I think you've got to use the 55:05 data to basically you can't just rely on 55:08 everybody. You're right. Most VC pitches 55:11 look the same. What you want to do is 55:13 dig one level deeper and talk to the 55:15 founders themselves and understand for 55:16 each of the last five companies that 55:19 they invested in, why did this founder 55:21 pick uh this firm? 55:24 >> Can I push back on the model that you 55:26 have and just pretend that we're a 55:28 hypothetical partner? Okay. You have 35% 55:31 reserves. Why is that optimal over just 55:34 having more lines in the portfolio? When 55:37 you hear about the 100x 200x multiple on 55:39 fair, I'm like focus on ownership, have 55:43 more increase diversification and take 55:46 away the reserves. Why do you 55:48 >> I think there are two ways of operating. 55:49 I'll give you an example. Uh 55:52 trade desk where I'm on the board had 55:55 two seed investors. Uh I ventures Roger 55:58 Arinberg who's absolutely a goat and 56:00 then uh founder collective again goat 56:03 firm. both of them. So they have two 56:06 completely different philosophies. U 56:08 founder collective only does first 56:10 checks they never do any praa afterwards 56:12 period. Roger on the other hand doubles 56:15 down again and again and again. So trade 56:18 desk raised I think two or three rounds 56:20 of financing. That's it and went public 56:21 very early. It was very hard for them to 56:22 raise financing. So the multiple that uh 56:27 founder collective generated was 56:28 incredible because they only invested at 56:29 the seed round and they got it at 5 56:31 billion or something like that or even 56:32 more. or I think they held for longer 56:34 while Roger generated a huge dollar 56:37 return even though his multiple was 56:39 different. So there are two 56:40 philosophies. I do think if you look at 56:42 uh my philosophy is more if you look at 56:45 founders fund which I think is one of 56:46 the best performing funds a huge part of 56:48 their success is basically doubling down 56:50 on the companies that matter and I think 56:52 uh the unsung hero of founder fund is a 56:54 guy called Napoleon Ta who leads the 56:56 growth practice and Napoleon basically 56:58 is the one who decides which of the 57:00 companies should we double down on and I 57:02 think if you double down properly it 57:04 changes the complexion of the fund 57:06 because you have much more insight I 57:08 would argue that if you if you work 57:09 closely with these founders 57:11 you have much more insight into these 57:13 companies and how they're going to do 57:15 and even how they think about the future 57:17 opportunity because you've thought about 57:18 it with them than a random company you 57:21 meet. Now there is a balance there. You 57:24 don't want to be overconentrated but I 57:26 would argue that if you have x number of 57:28 bets shoot I mean and you work with the 57:30 founder and you think highly of them. 57:32 That's why each founder fund founders 57:34 fund fund is named after the company 57:35 that makes it like in in colloquial 57:37 terms. There's the Andural fund, there's 57:39 a SpaceX fund, etc. Why? Because that 57:42 one company is the one that makes it. 57:44 And that's the reality, Harry. I mean, 57:46 literally, um, you have one company most 57:48 likely, one or two companies that will 57:50 drive most of the returns of any any 57:52 given fund. The question is, do you just 57:54 want to have the initial stake? Do you 57:55 want to increase your probability of 57:56 finding the initial company or do you 57:59 feel like you you there is a explore 58:01 exploit thing, right? Where do you stop 58:03 exploring and when do you stop 58:04 exploiting? So, different people have 58:05 different points of view there. My fund 58:07 one could have been, which you're an 58:09 LPN, and I'm very grateful to you for 58:10 supporting me when I was 18, 19. Um, but 58:14 it could have been at one point the 58:15 Hopin fund. It could have been the 58:16 clubhouse fund. Um, and it turns out 58:19 that it will most likely be the linear 58:22 fund. I think I think Linear is a great 58:25 business and we were very early there. 58:27 But my point being with the transitions 58:29 in name, it wasn't obvious. And so my 58:32 question to you is with preemptive 58:34 rounds coming so fast, how accurate do 58:38 you think you can be in predicting the 58:40 winners? Because it definitely wasn't 58:42 obvious to me. You got to be thesis 58:45 driven first and foremost. I think what 58:46 we are is we think about what is the 58:49 thesis. In other words, you've got to 58:51 have a good sense of who the other 58:53 companies are and players of in the 58:54 space and you've got to understand why 58:56 this company is better than every other 58:59 company. What on what dimensions is it 59:01 better? Is that the dimension? Is that 59:03 durable enough over a venture time frame 59:05 which is 7 to 10 years or even you know 59:07 maybe 10 to 12 years now. So you've got 59:09 to do work. You've got to be thoughtful 59:11 and patient. Remember what being 59:13 concentrated does. It gives you more 59:15 time. It gives you more time to meet 59:17 companies. It gives more time to think. 59:19 It gives you more time to be helpful to 59:21 companies. But you don't feel the 59:23 pressure to deploy on a monthly basis. 59:26 So if you look at a 30 portfolio company 59:28 portfolio fund over 3 years which is the 59:30 initial deployment period you're 59:31 basically almost investing one company a 59:33 month and so that's incredible like I 59:36 almost feel there's pressure on the 59:38 folks who do 30 to 40 to do basically 59:40 one company as a partnership per month 59:42 if you're doing green oaks I think is 59:44 you know if you would ask me who's one 59:46 of my favorite Neil you mentioned Neil 59:47 was on the show six there seven funds 59:50 have basically what 65 companies overall 59:53 our six funds are 65 companies so 11 59:55 companies per fund He's an absolute 59:57 beast. You He's got 10 companies that 59:59 have returned over $2 billion. 60:02 I The [ __ ] thing about my life, Gogle, 60:04 is I hang out with these people and I 60:06 just leave feeling like a total loser. 60:10 >> You're just like, you leave Mickey Maler 60:12 and you're just like, "Yep, no, I didn't 60:14 do the Robin Hood." 60:16 >> Yeah. I mean, just 60:17 >> I think you've gotten better with every 60:18 one of those interviews. I've seen just 60:20 your style and just your your investing. 60:23 Who do you learn from? 60:26 >> Entrepreneurs most of people in the 60:28 arena. I think uh for me since I'm so u 60:32 and our firm is so trend driven and we 60:34 uh we really care about what is the 60:36 thesis, what's the market, the best way 60:38 is I think you can talk to investors but 60:40 they're always one click away. You've 60:42 got to talk to people who are in the 60:44 trenches building products. You've got 60:46 to understand what's changing in their 60:48 lives, how they're thinking about the 60:49 customer. Um and so I think you got to 60:52 today's world you've got to stay close 60:53 to the model companies for example. So I 60:56 have people that I meet with at each of 60:58 the model companies to understand what's 60:59 coming down the pike, how they thinking 61:00 over the world etc. Because like you 61:02 said you got to also understand I we 61:04 always use a joke um uh this joke like 61:08 this was 10 years ago. If you're a 61:09 startup which is directly in Google's 61:11 road map like directly you should not be 61:14 building it because Google is very good 61:15 when something is directly in road map. 61:17 like a tank. They would just roll over 61:18 you even slowly slowly but doesn't 61:20 matter. They were implacable. They would 61:22 roll. But if you're even 10° to the 61:25 side, it's very hard for them to like 61:28 move like that. So, you're generally 61:29 safe because it's just not they're just 61:31 rolling in one direction. I think you 61:32 just want to know what direction the the 61:34 turrets or the guns are pointed at uh 61:36 for each of these models. I mean, 61:38 there's one part of the is what the labs 61:40 are going to do. The second part is what 61:41 a customer is going to do and what 61:42 customer behaviors there are. And so you 61:44 want to talk to I think you do want 61:46 learn from entrepreneurs a little bit to 61:48 better understand who are the companies 61:49 but then you want to go and talk to the 61:51 companies themselves and you want to 61:52 understand you talk to three or four 61:54 companies in the same space you very 61:57 quickly I think I'll never forget I 61:58 think people who met Tony at Door Dash 62:01 and many of the other folks who were 62:03 fundraising at the same time they always 62:04 felt Tony had a deeper understanding of 62:07 the same market than others globally and 62:11 so you basically want to meet that's why 62:13 I My biggest regret is one of my biggest 62:15 regrets is actually passing on Vanta uh 62:18 because I met Christina but I had 62:20 already committed I was like this person 62:22 is going to win the market but I had 62:23 already committed to another company in 62:25 the space and uh sadly enough I I 62:27 believe in like once you invest in a 62:28 company as an angel I didn't have the 62:30 time to scan the landscape and meet with 62:32 all the companies in the space and 62:33 that's what I do now. I have time but 62:36 you know Vanta is um will be definitely 62:38 part of my anti portfolio. I think it's 62:40 part of yours too if I remember 62:41 correctly. Oh, D. Elad sent me and he 62:44 was like, "Dude, this is amazing. This 62:46 is amazing." Every time Elad sent me 62:48 something and said, "It's amazing. Just 62:49 [ __ ] do it. Do not think you're 62:51 smarter." Is my takeaway there. Um, 62:54 there are other people when they do it 62:55 just like, "Do think you're smarter." 62:58 Like, I will leave them out of it, but 63:00 there's some friends where I'm like, 63:02 "You've consistently sent me this. It is 63:03 shit." Um, anyway, we mentioned we work 63:06 earlier. I use Wei work as an example 63:09 with Miles Clemens about selling. I'd 63:12 love to hear your thoughts on how do you 63:14 think about when to sell? Obviously, we 63:15 have investors, you have LP stake, they 63:17 care about DPI today more than ever. How 63:20 do you think about liquidity and when's 63:21 the right time to take chips off the 63:23 table? 63:24 >> As an angel, I used to hold till IPO. 63:26 So, Figma had many liquidity 63:28 opportunities during the years, but I 63:29 kept kept holding it for 13 years till 63:32 it went public. And uh I think at the 63:35 IPO, it was actually priced very nicely. 63:37 Unfortunately, after the IPO, it has 63:39 been uh it has been more challenging 63:41 price-wise. Uh but I think as a as a 63:44 fund investor, it becomes interesting. I 63:46 think there are two situations. One of 63:48 the things I think most early stage 63:49 firms get wrong is they just focus on 63:51 mo. They don't focus on irr and mo is 63:54 multiple invested capital. I think IRRa 63:56 matters a lot as you know. And so you 63:59 can have an LP told us about a firm that 64:01 gave them a 7x mo over 20 years and that 64:05 was a teens IRRa and that is like okay 64:08 you know there's something crazy here I 64:09 mean that is an a venture venture 64:11 venture firm and so you've got to look 64:13 at go forward IRRa and your projection 64:15 if your go forward IR at every liquidity 64:18 opportunity is lower than what you are 64:21 basically promising your LPS or what you 64:23 think your fund should have I think you 64:25 should sell I think you have an 64:26 obligation to LPS be to at least sell. I 64:28 like Fred Wilson's uh strategy around 64:30 selling which is sell a third, hold a 64:34 third and trade a third. This is when 64:35 the asset is completely liquid. Uh but 64:38 in this case, I think you want to sell 64:39 at least part of it. Especially if the 64:42 asset is a uh is a is a is an is a 64:45 company that'll return a chunk of your 64:47 fund. So if you're going to return 20 30 64:49 40% of your fund, I think you you you 64:51 owe it to your LPS uh to sell a piece of 64:54 it, especially if the go forward IRRa is 64:56 not um is not compelling. 64:59 >> On top of that, you then have the hold 65:01 period after the IPO where you have, you 65:03 know, obviously you're 65:04 >> Exactly. And you don't have Exactly. And 65:06 that's another uncertainty. So I do 65:08 think the secondary markets have been 65:10 one of the best or most interesting 65:11 developments over the last few years. 65:13 And so now all these great companies 65:15 have pretty liquid secondary markets 65:17 where you can sell sell shares. 65:18 Obviously there's Roofer and so on the 65:20 company has I do think uh there are 65:22 these hyperlquid periods in the market 65:24 and now is one of them. So I do think 65:27 one should be very careful and 65:29 thoughtful about what the go forward 65:30 IRRa is uh for any asset one owns and 65:34 really think carefully about whether one 65:36 should sell or not. 65:38 Can I ask you when you look back at the 65:39 angel portfolio, what's the biggest 65:42 regret? 65:42 >> Pattern matching too much. I think a 65:44 good example is Quinc recently raised at 65:47 10 billion. I saw Quins 4 years ago when 65:50 it was at 100 million valuation and I 65:52 was like a D2C company D2C companies are 65:55 kind of on the downswing. How how good 65:57 can this company be? What do you you 65:59 know? So I I literally just dismissed 66:01 it. I didn't even look deeper into the 66:03 company. And you know I think um yeah I 66:07 think I 66:08 >> what's the what's the takeaway from that 66:10 then? 66:11 >> The takeaway is that you can't just take 66:14 an industry and say it's good or bad 66:16 within every industry within every 66:18 category. There are great companies and 66:20 there are mediocre companies and you've 66:22 got to understand each company's 66:24 remarkability and differentiation. 66:26 Quinc, for example, had an incredible 35 66:28 to 40% repeat purchase rate, which was 66:31 like higher retention than most consumer 66:33 apps. And so I should have paid more 66:35 attention to that versus dismissing it. 66:37 It was literally in the blurb. And I was 66:39 like, okay, you know, so what how big 66:41 can this get? They were. And so that's 66:43 non-conumption. What I didn't like and 66:45 respectfully when you said about kind of 66:46 your trend style of investing is 66:49 actually some of the biggest misses I 66:51 also have is when there's an amazing 66:53 founder that's clearly amazing but 66:55 operating in a bad space and they pivot 66:58 3 months later into a good space but I 67:00 turn them down when they're in a bad 67:01 space and I'm like I I don't care what 67:04 they're building. I don't care what 67:05 trend it is. Goal, you're amazing. If 67:08 you're selling pillows, I'm in. If 67:11 you're a seed investor, if you're a pure 67:12 seed investor, I think you have to do 67:14 that. I think that's what VC does, 67:16 right? I think that's actually the right 67:18 way to do pure seed investing. First 67:20 round capital, I think, is a one of the 67:21 best seed firms. Guess how many 67:22 companies they have in their each fund? 67:24 80 companies. Why? Because you got to 67:26 take 80 bets, which means that some of 67:29 them and YC of course best seed investor 67:31 of all time or preede. I mean, you've 67:32 got to take hundreds of bets because 67:35 most companies will pivot. But I think 67:37 as a concentrated portfolio, you've got 67:39 to basically understand the business and 67:43 you've got to bet on both the business 67:44 and the founder. Unless again the 67:47 founder is an N of one founder in a 67:49 space. I I I think there are two 67:52 categories of actually there's three 67:54 categories of founders in some way. 67:55 There are repeat founders who know a 67:57 space exceptionally well who've done it 67:59 before. You're betting on them again and 68:01 again to do it. The security is a great 68:02 example. Security is full of repeat 68:04 founders. They know the market. They 68:06 know the they know the history. Uh they 68:08 know the customers all of that. Uh but 68:11 then there are consumer companies. 68:12 Consumer internet is full of first-time 68:14 extraordinary founders. Mark Zuckerberg, 68:15 Larry Page, all of these are just 68:17 first-time founders. So those are two 68:19 archetypes. The third archetype that has 68:20 come up is AI labs researchers. uh uh 68:24 and so I think you you know that that 68:26 one is a more interesting bag where you 68:28 have of of course Daario and and and and 68:30 the open AI folks but then you have a 68:32 bunch of other labs that came up and 68:34 ultimately didn't turn out to be 68:35 anything. Um so you want to for these 68:38 kinds of folks you as a as a seed 68:41 investor you just want to blindly write 68:42 a check into them a brilliant young 68:45 person uh who's done something 68:47 extraordinary um a repeat founder or 68:50 maybe an AAB researcher 68:52 >> will will you do a frontier model uh a 68:55 neolab uh periodic labs uh uh ineffable 69:01 where they're clearly [ __ ] amazing 69:03 people pedigreed to the hills but the 69:05 price is in the billions. 69:07 >> Not possible. I don't think with our 69:09 fund it's possible. I think the 69:10 ownership is just literally the first 69:12 round for these companies is like you 69:14 said a billion dollars. So I think it's 69:16 just too high a the riskreward is just 69:19 not worth it. 69:20 >> Are you seeing the mega funds 69:21 cannibalize the business model of our 69:23 series A? 69:24 >> Uh they're playing a different game. I 69:26 think uh smart founders are look I think 69:30 I I think very highly of the mega funds. 69:32 Yeah, 69:32 >> they've basically gone and they deploy 69:36 $15 million checks almost as an option 69:39 and a lead lead generation for the for 69:42 the next round. And their strategy is 69:43 obviously to have an index at the A of 69:47 every single good A company um and then 69:49 double down on the ones that truly 69:52 matter and triple down and quadruple 69:53 down and do SPVS in them and and do 69:55 specialized funds in them and so on. And 69:57 I think it works it works for LPS in 70:00 some way. It's a different asset class 70:01 than the funds that the early stage 70:03 funds but it's different. I think smart 70:06 founders, good founders have started to 70:08 look beyond just the fact that they can 70:10 get 10 million very quickly from a mega 70:12 fund and say what am I getting here? We 70:15 see many examples actually in mega funds 70:17 of um partners leaving the fund um and 70:21 the companies orphaned within the mega 70:23 fund because their partner has left and 70:25 now they don't have a single person to 70:27 advocate for them in any way, shape or 70:28 form and they're a drift and anybody who 70:31 so repeat founders actually what is 70:32 interesting is repeat founders are most 70:35 likely to uh essentially know uh and see 70:39 beyond just the glitz of a mega fund in 70:41 some ways because many of them have gone 70:43 through especially the ones that have 70:44 started a 70:45 in the in the last five or six years uh 70:47 we have many stories where the the 70:49 mid-level partner at the mega fund has 70:51 left who was their partner and then 70:53 they're like okay [ __ ] you know I 70:55 basically don't have an advocate in the 70:57 fund and I now am left with a person who 71:00 I don't know um and they don't know me 71:02 and uh they're joining their board my 71:04 board and that's a tough one you've seen 71:06 Nikico Bonatos you've seen Max Gazour 71:08 you've seen Arif Yan Muhammad you've 71:10 seen or are we just seeing the start of 71:12 this continuing wave 71:15 and there will be a huge amount more 71:16 spin outs or do you think we're going to 71:18 see that cel? 71:19 >> I think we're going to see a more spin 71:21 outs. I do I do think there is a limit 71:23 but I do think you're going to see the 71:25 you're going to see these mega funds 71:27 train again more waves of investors and 71:30 these investors are going to realize 71:31 that being a mid-level partner at a mega 71:33 fund is not what it's cracked out to be. 71:36 So they're going to spin out and go back 71:38 to, you know, the way of doing things 71:40 that venture used to be 20, 30 years 71:41 ago, which is a small group of partners, 71:44 uh, building a deep relationship with 71:46 entrepreneurs. 71:47 >> Uh, listen, dude, I want to do a quick 71:49 fire. So what have you changed your mind 71:51 on in the last 12 months? 71:53 >> I used to think pure remote would u 71:55 would scale for early stage companies if 71:57 you had the right culture, but I don't 71:58 think that anymore. I think you've got 71:59 to be in person at least a few days a 72:01 week. 72:01 >> What what what made you change your mind 72:03 there? I think just seeing a few 72:05 companies where literally the companies 72:07 died because the founders were unable to 72:09 agree. They had everything going but the 72:11 founders were just not in the same place 72:13 and they were just not able to move fast 72:15 enough and agree and align on the 72:17 strategy and change things. So iteration 72:19 speed just suffers massively if you're 72:21 pure remote. It doesn't need to be 5 72:23 days a week but at least 3 days a week. 72:24 >> Biggest advice to a young person leaving 72:28 university today. 72:29 >> I know it feels exciting to start a 72:31 company. Everyone's doing a startup. AI 72:33 is a new way but my strong advice is to 72:35 first get two to three years of work 72:36 experience at a company at a good 72:38 company you won't regret it you'll learn 72:40 a lot both the experience and the 72:42 network of people will be invaluable for 72:44 you so just two or three years don't be 72:46 impatient life is long work get some 72:48 work experience before starting the 72:50 company 72:50 >> you've got to answer an unfair one 72:52 you've got to invest in three different 72:54 types of funds a seed fund a series A 72:57 fund and a growth fund which three are 73:00 you choosing and they can't be mine 73:02 capital 73:03 >> first round capital and benchmark where 73:05 I win I win an LP in both and then green 73:07 oaks first round would be the seed bench 73:09 with the sea and green oaks would be the 73:11 would be the growth 73:13 >> wow that wasn't a hard one was it you 73:15 thought about that mo most people are 73:16 like oh I can't do 73:18 >> asked us this question uh they basically 73:20 said I don't know whether they were 73:21 asking us for they literally asked us 73:23 okay what's so I I've actually answered 73:25 this question LP before 73:27 >> dude that's fantastic what has been the 73:29 hardest decision that you've made in 73:31 your career you've left amazing 73:32 companies you what's been the hardest 73:35 decision 73:36 >> leaving Google leaving Google I think 73:37 there was a saying you leave Google only 73:39 once so leaving Google to start a 73:40 company I was on a pretty incredible 73:42 trajectory there I was learning a lot 73:44 really enjoying it but it was really 73:46 tough to leave Google I don't regret it 73:47 but it was very very hard to leave 73:49 Google 73:49 >> who's the best CEO you've ever worked 73:51 with 73:52 >> all four of them Larry Mark and Tony and 73:56 now Brian Armstrong um Bill and Ben at 73:59 Pinterest it's a hard one I think all 74:01 four of them have different superpowers. 74:02 I would say the best technical CEO Larry 74:05 Page, the best growth centric CEO, Mark 74:07 Zuckerberg, the best designcentric CEO, 74:09 Jack Dorsey, and the best uh physical 74:12 world operational CEO, the most likely 74:14 be Jeff Bezos next Tony Shu. 74:16 >> I'll take that. That's a good one. 74:18 What's the biggest miss? We've said 74:19 Vanter. Is it Vanta? 74:20 >> Quinc. Quinc, man. Uh most recently 74:23 Quins, but to be honest, even bigger 74:24 miss than that in some ways. is not a 74:26 miss in terms of investing. It's that I 74:29 couldn't predict that Facebook could be 74:30 a $2 trillion company. When our company 74:32 was going to be acquired by Facebook, I 74:34 was arguing with the corp dev team at 74:36 Facebook and we were arguing over the 74:37 terminal value of Facebook and we had to 74:39 put China into the mix saying entering 74:41 China will get us to 40 billion in 74:43 market cap and we were arguing whether 74:44 it was 20 billion or 40 billion in 74:46 several years from then. This is in like 74:47 2010 and turns out in less than 10 years 74:50 11 years it was a trillion dollar 74:52 company. So when these things work, they 74:54 work in a in a scale that is 74:55 unimaginable. And even at Google, I 74:57 remember very well after the IPO. I was 74:59 there during the IPO and we were sitting 75:01 around with a bunch of PMs and we're 75:02 saying, man, the the company is valued 75:04 at 30 billion. It's too expensive, too 75:06 expensive. And so you just these things 75:08 compound and uh it's just incredible to 75:11 see these things become trillion dollar 75:12 companies. So Facebook and Google in 75:14 some ways the biggest misses in terms of 75:15 not being able to predict that they were 75:17 going to be multi-trillion dollar 75:18 companies. Which angel investment is the 75:22 highest multiple? 75:24 >> Figma. 75:26 >> What was the multiple? 75:27 >> Uh five of between 500 and 1,000x at the 75:31 time of IPO, but it has sadly gone down 75:34 since then. 75:36 Oh, 500 to a,000x. Jesus Christ. Um, 75:40 tell me final one. What most excites you 75:43 about the next 10 years? I like to be 75:45 optimistic. I think we have too much 75:46 pessimism. What do you like? I'm really 75:48 freaking pumped about this. 75:50 >> The most ambitious entrepreneurs are 75:52 finally tackling the hardest problems. I 75:54 feel uh there is the ambition with AI 75:58 especially as unlocked is just 76:00 incredible. So uh the ambition of 76:02 entrepreneurs tackling the hardest 76:04 problem facing humanity and society is 76:06 just absolutely incredible. How can you 76:08 not be optimistic when you have you know 76:10 Elon going? I mean I think we have now 76:12 these entrepreneurs who are role models 76:14 who are not just you know building these 76:16 you know small companies but they're 76:18 truly taking on humanity problem. So the 76:20 answer to Peter Thiel's question of we 76:22 were looking for flying cars and we got 76:25 140 line you know what 140 word uh uh 76:28 140 character apps I think has been is 76:30 finally I think we're coming it's coming 76:32 into focus. I've got to ask one more, 76:34 but you said about Peter Teal there. 76:35 Obviously, he has the Teal Fellowship 76:36 and a preference for young ambitious 76:38 founders. We're seeing this massive 76:40 movement towards very very young 76:41 founders. We mentioned Mccor earlier who 76:44 are brilliant. Are you in line with the 76:46 shift to the earliest youngest founders 76:50 and how do you feel about that shift to 76:52 super young founders? 76:53 >> I think u I actually am a huge fan of 76:55 it. I I feel even at companies I feel 76:58 some the companies that are not hiring 77:00 young people they're making a huge 77:01 mistake because young people are more 77:03 AIed as you could call like looks maxing 77:06 AI maxing than uh than anybody else. Uh 77:09 in fact I think the younger people are 77:11 adopting tools better and they just live 77:13 and breathe differently than than 77:15 others. Um so I I'm a huge fan. I've 77:17 actually invested in more dropouts as an 77:19 angel now over the last uh few months 77:21 than I have invested in um in in in my 77:24 rest of the last 15 years I've been 77:26 investing. So I don't think it's a right 77:28 thing to be honest for many of them to 77:30 be dropping out and starting. I do think 77:32 they could benefit socially, 77:34 emotionally, etc. But some of them are 77:37 just exceptional. I don't think all of 77:39 them are, but I do think this crop is 77:41 going to produce some incredible 77:43 founders. 77:44 >> Gole dude, I so appreciate you. I've got 77:46 so many notes I had to go on different 77:48 sides. Um, this has been fantastic. So, 77:51 thank you so much for being so amazing, 77:52 dude. 77:53 >> My pleasure, my friend. Look forward to 77:54 doing stuff together.