The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
00:50:37 · Luke Harries, Harry Stebbings
[00:50:37] S1: We keep the growth teams pretty small. And so, frankly, there's just not enough time to be meddling in the different engineers' work. You're not gonna be doing wireframes or going to be doing, like, large numbers of user interviews because you're too busy running the launches or running the channels for that particular product. So that's one. It's like we keep it pretty busy and lean. And the other one is if you hire great engineers, they just don't need it. And so the best engineers we hire, we put everyone through a product challenge, which is where we give them an imaginary product, and we ask them, step one, work out what features you should build. Like, imagine I'm a customer. Imagine you're doing research. They will actually go through on Google and find competitors and use that to shape the features. So step one is work out what features to build. Step two, they're then gonna turn that into the actual wireframes, and they'll be sketching out in Figma of, like, okay. This is what I think the product should look like. And then step three is you have system architectures. So they're like, okay. This is the different back end, the API design. And if you hire an engineer who, one, you know is really strong at coding because you've screened for that before, and two, you've then tested them through this full product loop. Like, you have confidence that they're able to take those products and those features from idea to launch. [00:51:48] S0: Do you think we will see PMs reduce as a role in future tech companies with AI becoming more and more prominent?
00:51:48 · Harry Stebbings, Luke Harries
[00:51:48] S0: Do you think we will see PMs reduce as a role in future tech companies with AI becoming more and more prominent? [00:51:54] S1: Yeah. I think PMs move to either growth. So you combine the PM role with marketing, and depending on your size and how strong your engineers are, you do more or less at the PM. Or many PMs will actually move towards product engineering, which is they're somewhat technical already. They can actually upscale themselves using tools like Cursor or Lovable and actually go and ship full end to end products. [00:52:17] S0: Can I ask you today, when you think about ElevenLabs code, how much of it do you think is AI generated? [00:52:22] S1: So I still write code probably 20% of the time. The only code I write is through Cursor, where I'm describing in the chat what I want done and then it executes on my behalf. And I'm probably more the extreme because I'm often doing, like, growth features which are adding on top of the core of the product. In terms of core engineering, I would say probably 60, 70% of code now is AI written. So it's still like the majority. Where we don't do any AI written or generated code is our research engineering, where it's just like research engineering and research, where it's a very different style of work. And also these code bases are so sensitive. When you think about the initial ElevenLabs launch and how quickly we got to a billion dollars of revenue, That's pretty much Peter's models being world class, and so that we just don't want to go through any sort of LLM.
00:52:22 · Luke Harries, Harry Stebbings
[00:52:22] S1: So I still write code probably 20% of the time. The only code I write is through Cursor, where I'm describing in the chat what I want done and then it executes on my behalf. And I'm probably more the extreme because I'm often doing, like, growth features which are adding on top of the core of the product. In terms of core engineering, I would say probably 60, 70% of code now is AI written. So it's still like the majority. Where we don't do any AI written or generated code is our research engineering, where it's just like research engineering and research, where it's a very different style of work. And also these code bases are so sensitive. When you think about the initial ElevenLabs launch and how quickly we got to a billion dollars of revenue, That's pretty much Peter's models being world class, and so that we just don't want to go through any sort of LLM. [00:53:11] S0: Was that a million dollars of revenue? [00:53:13] S1: I said billion dollars of valuation. So... And and I think truthfully, at that point, it was, like, probably half the enterprise value of the company was Peter's model.
00:53:13 · Luke Harries, Harry Stebbings
[00:53:13] S1: I said billion dollars of valuation. So... And and I think truthfully, at that point, it was, like, probably half the enterprise value of the company was Peter's model. [00:53:22] S0: Nuts. Yeah. Unbelievable. Can I ask you, we mentioned? I'm really jumping around, but fuck it. I enjoy it so much. We mentioned, like, the enterprise discussions you had earlier moving into enterprise. The big concern with a lot of, essentially, these Gen AI companies today is that they're sugar high revenues. They're unsustainable. We've seen them both. They hit $50,000,000 hour, wow, wow, wow. But it's like, is it real revenue? How do you think about... Do we have a generation of companies where it's pretty synthetic revenue or not? [00:53:51] S1: You know, it probably varies by product to product. [00:53:54] S0: Yeah.
00:53:54 · Harry Stebbings, Luke Harries
[00:53:54] S0: Yeah. [00:53:54] S1: But if they're solving a real problem and the retention is strong, I see no reason why these aren't real revenues. And the thing I've spoken to a few VC friends, I don't understand why they're not upping the rate that they're investing. They're seeing companies which are now... They're like, oh, it's already on $6,000,000 MRR in, like, one year. And I'm like, why aren't you investing on it? And they're like, well, there's this other one which is now on, like, $10,000,000 of MRR. You're like, surely now is the time. Like, the wrong time to be the tiger global play of investing quickly was back in the COVID boom. The exact right time is now when I think you have this unlock of capabilities with the AI models, which has meant new products which just couldn't exist before. [00:54:35] S0: I think the hardest thing is before, there were two to three competitors always. Now there is literally 10 to 15 Yes. On everything. And so discerning winners is much harder when all of them have relatively impressive traction and [00:54:48] S1: revenue. But as long as they're solving real problems and scaling quickly and the customers are retaining, particularly if you're doing b to b or enterprise, you're going to get that cohort of customers for a while. And so if you're already on $10,000,000 of revenue, if we think that it used to be the threshold for IPO was like, oh, let's get to $100,000,000 of ARR, you actually have entire waves of companies which are now reaching those thresholds. What is good retention?
00:54:48 · Luke Harries, Harry Stebbings
[00:54:48] S1: revenue. But as long as they're solving real problems and scaling quickly and the customers are retaining, particularly if you're doing b to b or enterprise, you're going to get that cohort of customers for a while. And so if you're already on $10,000,000 of revenue, if we think that it used to be the threshold for IPO was like, oh, let's get to $100,000,000 of ARR, you actually have entire waves of companies which are now reaching those thresholds. What is good retention? [00:55:16] S0: This is a hard point, which is like in traditional enterprise, it was like, you know, very, very high, very high. And you'd look at like a Lovable with like, I think they're 86%, I think Anton said on the show. In traditional SaaS, that's still shit. We're churning 14% a month.