Fast Software Is the New Fast Fashion | VCs Are Hiding Portfolio Value | Kirkland's $500M AI Bet

I think I've gone back to something that I used to spend a lot more time thinking about. There's pain and then there's 10x relief. If you have a deep cut on your forearm, you need a suture, not a Band-Aid. A Band-Aid will lessen the pain a tiny bit. And I feel like lately I've met a lot of companies and like, well, the thing you've built is a Band-Aid. We need a 10x better, undeniably superior product. And if your product is just incrementally better than the status quo, it's not going to be enough to get people to change unless the environment changes.

It's the week of June 8th. Welcome to the learning Corner, a Precursor Ventures podcast. Welcome to the Learning Corner by Precursor. I'm Mia Farnam, and I'm joined by our GP and managing partner, Charles Hudson, where we will be talking about the articles and goodreads that caught our team's attention over the past week. For our first topic this week, Lenny Rachitsky interviewed Tony Fadell on his podcast, Lenny's Podcast. Tony is the person behind the iPod, the iPhone, and the Nest thermostat that Nearly hour and a half long conversation went in so many great directions, but a large center point of it was Tony's argument that AI is making it easier than ever to build things, and that makes taste and craft far more important, not less. All things that we've talked about in weeks prior. He draws this analogy between fast fashion and what he calls fast software, where AI-generated code can work and even look right, but it's brittle, it's hard to maintain, and accumulates technical debt fast. His line was, if you're gonna build a real company, it can't be a throwaway. He also gets into something he calls cognitive surrender. This idea that builders are outsourcing not just the execution, but the judgment itself to AI, which he sees as genuinely dangerous. The piece is anchored in real examples from Apple and Nest and how opinion-based decisions get made when data isn't enough yet, and why sometimes having singular informed points of view still matters even when you can generate 100 options instantly. Going back to that analogy between fast fashion and what he calls fast software, his argument is that AI makes it easy to generate code that works on the surface but is really fragile underneath. He calls it the H&M versus luxury brand problem. You can get something that looks right very quickly, but it falls apart after one wash. And his broader point is that even if you can vibe code a product into existence, You still need all the other functions around it. You need marketing, you need sales, distribution to actually bring it to life and straight to the customer. Where do you feel like founders might be getting this wrong most of the time? Is it on short sight into the long-term technical debt side, or is it this assumption that a great product will sell itself?

I really liked this podcast. It reminded me in so many ways, you know, I, I worked in online games for a long time, and when online games discovered A/B testing, everyone was like, we don't need to think anymore, just A/B test everything and take the outcome of the test. And I'm like, well, that's not judgment. We didn't use the word taste. That's just letting the machines decide. And that did not produce engaging, fun games. It produced lowest common denominator kind of slop. And I think in some ways the just A/B test has turned into just ask Claude or ChatGPT and go with the results. And I think that also leads to like uninspired work. And it's funny, I just wrote a blog post this week about this whole thing about my experience as a software developer and also just learning that you can vibe code just about anything. It doesn't mean it's going to work or be easy to maintain or will be useful. I still don't think there are many great shortcuts to producing amazing quality work. I still think it takes a point of view. And again, I think this notion, I talked to a lot of our portfolio companies. I almost wrote this post called Doing Things Quickly Versus Doing Things Well. I think right now you can do a lot of things quickly. And I meet a lot of companies that are like, this is great. We've never shipped more features and products. I'm like, but are they the right features and products? Are you so quick to respond to customer requests that you're missing the opportunity to go deep and figure out, well, what's behind the ask? What is the thing that they're really searching for? And I worry that like a lot of this instant software stuff is companies are missing out on the chance to learn from what their customers are really trying to tell them. And so I thought the part about the iPod Nano was like, I was like, sometimes like if you have a point of view, the data will not always go your way. Right away. I've lived through this at Precursor. I've had a lot of beliefs about our firm and why we need to exist. And for a long time, people were like, I don't think you're right. I'm like, well, I think I'm right. So we're just going to see what time has to say about that. And we've stuck with it. And over time, people have been like, oh, you're more right about that. Well, there are many reasons to have quit along the way. It's just I had a point of view. I hope that like AI doesn't become a substitute for having a point of view and taking a stand on what good enough looks like or what the customer really needs. And I think it's very easy to just get lazy.

He talks about the example of it's so easy to bring in a group of consultants to run market maps and data insights on thousands and thousands of rows of data, but he's a big supporter of just good marketing and also like getting one singular point of view. And he's like, sometimes it's just really useful for you to purchase the product itself, become the customer, and tell us why you do or do not resonate with it. Point about the, the iPod, I know that they walk through a conversation around when do you know if a market isn't big enough? And Tony walks through the iPod saying that the first and second generations of the iPod were not a big enough market. They really only appealed to the Mac loyalists, the people who stand in line for anything that Apple drops. And a lot of it had to do with where are we meeting the customers that are outside of that circle and are, and how can we make adjustments to it? He also talks about all these different skunkworks projects that Steve Jobs didn't agree with at the time, but he's glad that he kept them kind of brewing in the corner. Things like the Windows connectivity and then the, the stylus itself. Steve was so obsessed with only finger. I want finger only to be the way that we interface with our products. And Tony was really glad that he kept some of those things like the stylus in the background, just working because he did believe in the long end it would be valuable. And turns out with the iPad, that's the preferred use case for the stylus. For, I guess, the founders at the earliest stages, how can you make a more strategic or purposeful attempt at figuring out if the market isn't responding because the idea is wrong or just because the approach is wrong? And all saying that Apple now can kind of iterate on whatever they'd like to in as many runs, but previously they almost went bankrupt. I think people forget about that.

In terms of pushing out the right product to meet people, it's a very narrow gap between persistence and foolishness. And I think oftentimes you don't know. Going back to the stylus, if you think about it, many people tried to bring that— the Trio, the Palm Pilot, the Handspring. And finally, with the iPad, we have pretty good stylus-based devices now. It only took many, many iterations of screen technology, stylus technology, battery, connectivity. You needed like all of these things. So I think you have to ask yourself, have I learned something about this that's directionally causing me to think this is not a good idea? Maybe you were just wrong about the arc. In other cases, you're like, wow, this is just going to take a lot longer. I think about the Vision Pro. Like, I'm not sure if the Vision Pro is early or just it's not something people will ever want. I don't know. I don't own one, so I don't— I'm not qualified. But I will say this. We invested in Fitbit when I was at Uncorked, and I was like, oh, people are going to have things like a Fitbit. Will Fitbit win it all? I don't know. I'm like, people are gonna have these kinds of things. This is just like the V1. I think that's like largely turned out to be true. So I think it takes a lot of conviction. Also, it sure does help to have the financial runway to stick with these things.

Totally.

I think that's like an important, that's an important consideration. Like, is your company set up basically emotionally and financially to persist long enough for your dream to come true? So I don't know. I think about, I think about that a lot, but I'll also ask founders directly, have we learned anything? About the nature of our core hypothesis here that makes this a less good idea than it was when we first started.

Going back to the wearable example, it also made me think about Oura, and I think some things that you're just out of your control. I really do think that they needed COVID to happen in order to really get excited about it. I mean, I think about the NBA partnership that they made with the bubble of having every single person in that bubble wear an Oura ring to track their temperature as the mechanism to see if anyone has contracted COVID. Pre-2020, people knew what Oura was, but I think they thought it was a little weird. And especially if you weren't already wearing jewelry or things on your hands, I don't think it was interesting enough for people to spend the amount of money that they did on the hardware. Post that, I think people saw a use case. And still, some of that is just completely out of your control.

I bought my Whoop because I wanted basically early warning signs for COVID and sickness. Yeah. I'm like, I want respiratory rate. I want like, there's a bunch of leading indicators. This is a pretty cheap and easy way for me to get ahead of anything. And I would, to your point, I wouldn't have Bought it if not for that. That was the trigger. Every good company I work with also usually has some trigger event that's outside of their control that ends up being a tailwind.

And then also knowing when to pivot out of that use case too. I think Aura's done a really great job of capturing women's health. I think now they're like, I don't know if I want all of this market. And I've made a partnership with Natural Cycles in order to do it. Do I need that anymore? So they shed that. They have it built up inside of them. Woo is now trying to kind of build out that audience, but I think Aura's purposely trying to move away from that audience too, because they realize how siloed it can kind of be. Yeah. While also making them successful at the same time. No, there's always just this constant reevaluation of where you are and saying like, is it enough today? And like, when do I have to iterate on it? The last piece I will talk about, Tony's framework for finding what to build is really simple. He really says like, start with pain and then ask what new technology now exists to solve it. Nest is a really good example of it. He was like, there was no way to really measure real life changes to make adjustments to things that could actually really change the financial output of your family where it stands in the, in the health of your home. And AI, while it wasn't spoken about as AI back then, that was a great application of how to do that and expedite the ROI there. In today's build very quickly environment, do you feel like folks are starting with more of like a tech-led instinct on we can build this and then we'll figure it out and then working backwards into a pain point? Because he also does talk about how movies are kind of done backwards where you have this idea and then you build backwards with the actors and the script. So it, it does work sometimes. How do you kind of decide where to actually start?

I think I've gone back to something that I used to spend a lot more time thinking about. There's pain and then there's like 10x relief. If you have a deep cut on your forearm, you need a suture, not a bandaid. A bandaid will like lessen the pain a tiny bit. And is it better than no bandaid? Sure. And I feel like lately I've met a lot of companies and like, well, the thing you've built is a bandaid. What we really need is a suture or a skin graft. Like we need like a 10x better, undeniably superior product. And I've met a lot of people, I'm just like, you have like a 2x better product. It is definitely an improvement, but probably not a big enough improvement. And it's funny, Mia, like a lot of times when I talk to portfolio companies and they're just like, oh, I'm struggling to get people to care. I'm like, it's because the perceived benefit isn't big enough to change. And if your product is just like incrementally better than the status quo, it's not going to be enough to get people to change unless the environment changes. Yeah. And so I think I've met a lot of people lately where like AI allows them to quickly build something that's incrementally better, but because they don't fully understand the nature of the problem, the thing they end up building solves 100% of 2% of the problem, if that makes sense. And I'm kind of like, well, yeah, great. You've like automated or solved this one part that was hard, but like, what about the other 98% of the problem that also sucks that you haven't addressed? And in many cases, the breakthroughs aren't improving the way something works. It's just either rethinking it or removing the process altogether. I told my friend, you know, the best expense report would be one you never have to submit. It'd be one you don't have to submit. So make it, oh, we're running again, easy, better. Actually, the job to be done is to get me reimbursed. And an expense report is one mechanism to do it. If you could just do it without having it. And I think Brex and Ramp to some extent understood that was sort of the job to be done for the person spending money, which is different than the job to be done for the person who's approving those expense reports. But I think the best solutions I see usually like reimagine the nature of the solution and don't just say, oh, we're going to make this friction-filled process suck less. It's not enough.

And this is also not even talking about does your customer even care? It could be something that solves a real pain point. They talk about the smoke alarm through Nest, which they discontinued. There's so many things that we could change in that product and they're in every home and you have a massive market, but do people care enough in order to want to switch that out? And the answer ultimately was no, but like great market.

Yeah. I think about Orbi and Eero and like home mesh networking again becomes way more important when you have tons of devices trying to watch YouTube or you go cordless or you go, you cut the cable again, like certain conditions oftentimes need to be true for problems to surface. And so I don't know, I've become much more. Critical lately on is this thing that this person's building a significant enough improvement relative to the status quo that they're gonna actually get people to move? And in many cases, the answer is no.

Well, our next read comes from our good friend Hunter Walk's blog, and it's a really timely and important topic for anyone operating in or around early-stage venture. The topic's how to communicate portfolio value to LPs when your portfolio is full of SAFEs and you haven't had a priced equity round in a while. The traditional approach is to hold at the last priced round, but Hunter's point is that the market has shifted. Seed is a phase now. It's not just a round, and companies are regularly raising multiple SAFE tranches at different caps before even touching a priced round. His suggestion is a supplemental column he calls SAFE adjusted TVPI, basically using the most recent SAFE cap to recalculate the implied value of your ownership stake while being upfront that this isn't an auditor-approved tactic and shouldn't be replacing standard TVPI in a vacuum. He's careful to note that this isn't the right call for every SAFE. Not ones that feel more strategic than market-driven. And the goal is honest progress reporting, not inflating the picture. Of course, it is a small but meaningful idea for managers trying to tell an accurate story to their LPs without being penalized by accounting conventions that haven't quite kept up with how the market actually works. Obviously, Charles wanted your thoughts here and also reflections on why the market is calling for something like this now versus years ago.

It's so funny. I read it and I almost texted Hunter and said, is this community-adjusted EBITDA? I immediately thought of that. But we have this problem at Precursor, which is Many of our companies raise money on SAFEs, and there have been many cases where honestly it used to not be a problem because you would raise one SAFE, maybe two, and then you would do a Series A round and it would all get converted. So maybe you had a 3 to 4 year window where you were sort of stuck with these valuations and you could explain it, but they resolved themselves. It doesn't work that way anymore. We have companies raising 10, $100+ million rounds on safes. And we have companies going 4 to 4.5 years between price rounds. And most people's valuation policies only allow them to take markups when there's a price round. So how do you like get credit for interim progress? You've gotta have some way to show people. So we, we did a huge exercise this year. It was a lot to do to figure out like, what are these companies actually worth? You know, when is the new safe round valuation indicative of progress? When is it indicative of momentum or heat? What do each of these things mean? And it's, it's not easy, but I don't think we're going back to price rounds anytime. If anything, any of these problems just gonna get worse. I think SAFEs are gonna continue to be a part of life later in companies' financing. And so it's just, it's not gonna get better.

And I think historically SAFEs were really only used for the initial funding rounds and they were all capped at like reasonable values. Now it's not an uncommon thing to see, specifically folks at a YC. They'll raise at $30 and then they'll raise at $100 and they'll raise at $50. All on SAFEs. Those are all on SAFE caps. And I think people would be equally happy to be involved in each one of them. If you got to get in pre-YC and come in at $30, which is high, but a great show if it goes to $150 in the next 3 weeks. Like the question is you should get credit, or not the question, the same as you brought, likely should get credit because it shows access. But how do you communicate that to LPs in a way that they understand? Because that's just a market shift as well. And I think that leads into my last question about this piece, which is beyond doing something like SAFE-adjusted TPPI, are there other ways that you can showcase progress and traction within your portfolio to LPs? Or is that like an exercise that requires a little bit too much handholding than is allowed for a firm in their fundraising cycle?

Yeah, I think one of the hard things is, you know, I talked to some LPs that have teams of people focused on early-stage venture capital. You can have really nuanced conversations with those people because they understand the market, they're in it, and they know more companies than just Databricks, OpenAI, Anthropic, and SpaceX. They have a, a wider knowledge of the ecosystem. Then I meet people where I'm like, they have one person who's responsible for buyout, private credit, early stage, late stage, growth stage venture capital, and real estate. And you could argue that person's portfolio is too big, but that's what they're up against. And I've certainly had situations where I'm like, wow, this person just doesn't have the capacity for nuance. In some cases, they're going to do the, the Claude equivalent of what we talked about, which is they're going to just pull the data from Cambridge, look at the benchmarks, look at your performance and make a decision like, do I want to spend time on this? And in many cases, the answer is going to be, I do not. It doesn't look interesting because the— if just to be clear, the effect of not taking the safe markups is that your performance looks worse relative to other people who have a way to reflect those valuations in evaluation policy., and you can end up in a situation where people just go, this fund doesn't look interesting, not gonna spend any time on it. And you don't even get the chance to make the case. So I think this is important because I hope it allows people to get into the room and make the case.

Well, our last piece is from Reuters, but it's reporting on a Financial Times article that Kirkland Ellis, one of the most powerful law firms in the world with over $10 billion in annual revenue, is committing $500 million over the next 3 to 4 years to build a proprietary AI platform from scratch. They're starting with $100 million this year, and the platform was designed based on input from 250 of their own lawyers alongside 180 internal and external tech professionals. What makes this really interesting beyond the headline number is what it signals about the broader legal industry's relationship with AI. Kirkland isn't just licensing tools or partnering with one of the big AI companies. They're wanting to build something of their own, or at least market it as such. Our other major firms are going a different route. Some have partnered with Anthropic to co-develop AI tools for legal services. The liability risks are also very real and very present. We've talked about U.S. judges sanctioning lawyers in dozens of cases for submitting AI-generated content with fabricated citations. And even Sullivan and Cromwell apologized to a federal judge this last month after AI hallucinations made it into a court filing. $500 million is a massive bet for a law firm, no matter how much they bring in top line. And it's a really different approach than going to Harvey or Lagoora or co-developing with Anthropic like some other firms did. What, Charles, does this tell you about Kirkland and specifically what they want to own as a narrative by building it versus partnering.

I thought it was interesting that, you know, they're hiring real people. $500 million over multiple years is a real financial commitment. And the part of the article that stuck me is they just, they said, you know, we think if the LLMs will raise the floor for law, but we don't get paid for the floor. And I read a couple of other takes in this article, and it seems to me that they're of the mind that like, hey, we have a lot of specialty knowledge, particularly about, they're starting with, you know, fund formation and private equity. We have a lot of like institutional institutional knowledge that we believe gives us an edge, and we would rather capture that edge internally and build our own tools rather than either have that knowledge leaked or distributed to other people through an LLM that we don't control. And I think it's inevitable that there will be industries where people say, we have a lot of proprietary stuff and we want to build the thing that we control that takes our best ideas and our best practices. And I think for a services firm, and a law firm is a services firm, to spend this amount of money on technology is a gigantic bet. Because in a private partnership, that $500 million like could have gone to the partners. Yeah. Could have kept it and shoot its profits. And to me, it, you know, it tells me that in some ways I get it. One way is like it's offense. The other way is it's defense. Hey, we want to have some control over our own destiny. If Anthropic, you know, they're working with Palantir on this, but if Anthropic doesn't solve legal in the way we want, we don't want to be beholden to them. If somebody acquires Harvey and we don't have access to Harvey or LaGora, we want to take our own destiny in our own hands. And I think there are good open source models that they could probably train with their own insights and build something. Something that works great for them. And I think you're gonna see more of this in fields where there's like specialty domain knowledge and where that knowledge is concentrated in relatively small number of firms. Really cool experiment.

And I applaud them for having the courage to do it in a big committed way. I think if they had just said some random number, I think you could kind of like call question, but it's clear, like it's a commitment. There's a whole team around it. I always thought with like the services businesses, especially in the beginning of when we were talking about AI, I always thought, oh, it's a race to the bottom on where your pricing is. And that's ultimately where people are gonna land. From who they want to partner with. But this is a good showcase of how you can kind of become more defensible over time. If you want to continue to charge the premium fees that Kirkland does and deserves them, then you have to kind of create an infrastructure to protect it in this AI narrative era.

And I think they said the reason you hire an expensive law firm is because you're going to get above average advice. If you just need average advice, you don't need to hire a firm like that.

Well, we are all out of time. Thank you, Charles Hudson, for stopping by. Thank you to our listeners for jumping in as well. Please like and subscribe and we'll talk with you all in 2 weeks. We're taking off next week. So we'll see you in a couple weeks. Thank you. Bye.