AI Fatigue Is Real and Nobody Talks About It | Patient Capital Will Eat VC | How GPs Close LPs

And one LP told me a funny thing. He goes, you know, 100% NPS as a GP is a bad thing. I was like, what do you mean? He goes, it means you probably have never had a hard conversation with anybody or had to make a hard decision, and you've prioritized keeping people happy over doing the hard parts of your job.

It's the week of February 9th. 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 through the articles and Goodreads that caught our team's attention over the past week. Our first piece is from Siddharthkar titled AI Fatigue is Real and Nobody Talks about it, exploring the paradox that AI is making engineers more productive on paper, but more exhausted in practice. Siddharth shipped more code last quarter than any quarter in his career, yet he's never felt more drained. His core argument is that AI reduces the cost of production but increases the cost of coordination, review and decision making. And those costs fall entirely on the human. He's gone from being a creator, creator and maker to a full time reviewer. And reviewing AI generated code is actually harder than reviewing human code because every line is suspect, because it all looks plausible. And beyond review fatigue, he talks about the FOMO treadmill spending weekends evaluating every new tool and update across cloud code, Google, Amazon, Q and others, only to see maybe a 5% improvement before the tools update again and invalidate everything he learned. But what scares him most is what he calls thinking atrophy. He describes being asked to work through a problem on a whiteboard with no AI, AI, no laptops, and realizing he'd outsourced so much of his initial thinking and reasoning from scratch felt genuinely difficult. Ultimately, Carr argues that the real skill of the AI era isn't prompt engineering or knowing which model to use, it's actually just knowing when to stop. So we talked about brain atrophy and AI fatigue in a previous article where it was coined brain obesity, but that was written from the standpoint of more non technical individuals. This being written by someone whose entire career is technical, did it make you think about AI fatigue in a different way?

You know, yesterday I talked to one of our portfolio companies that's in the DevOps space. I don't think I can say their name yet, but they were doing something in the DevOps space around continuous integration and code quality. And I was talking to one of the co founders and he goes yeah, we Sometimes will see lines of code when we're trying to do continuing integration. And we'll go, no human would ever construct a line of code that way. That must be AI generated. And he's like, it's part of what's made this whole thing hard, which is the cost of writing code and generating code has come down tremendously. So, as is often the case when the cost comes down, you just get more of it. But more doesn't always equal better. And I had this experience, you know, I'm like this bad junior engineer, intern level engineer now. And I was talking to a friend of mine who's been a software developer for his whole life, who gives me a lot of advice. And I said, you know, every time I finish something like 30 days into using it, I know exactly how I should have architected it from the beginning. And he's like, welcome to software development, my friend. And what I've realized is that these tools make it so easy to get started. You can skip all of the work you would have done around planning and thinking. And there's the temptation of like, oh, I can just sit down at the terminal and bang away and I'm going to get usable code. It used to drive me nuts, Mia, when I. I ran a software company and the engineers at my company, I'd see them sitting there staring at the wall or like throwing a tennis ball in the air. I'm like, get to work. Write code. And they'd be like, hey, we're like thinking through a bunch of architecture decisions. I'm like, it looks like you're sitting in your chair. And now I fully understand. I mean, I always believe them. I gave them a hard time, but I didn't understand the importance of planning and thinking through, like, if this is the end state of the system, there are decisions you can make up front that will lock you into a path that will be hard to get out of or can give you the room to grow. And that's the kind of stuff that like, whiteboard thinking and planning really helps with. And I do think we've gotten to it. We're doing, I think, the natural knee jerk thing, which is computers can write code. Computers can write lots of code. Let's just have computers write and ship lots of code. I'm like, well, someone still has review and clean it all up. The same CEO I talked to yesterday told me something else interesting. He goes, this AI generated code works the best for people who already have mature code bases where the code can learn style, convention and understand how it works. I worry about these people who are starting from scratch who have predominantly AI generated code for that very reason. The other thing that I've come to realize is when I finish something with Claude, it's ready to go the instant. It doesn't need to take a coffee break, it doesn't need to go home. And it can feel oppressive because these tools are able to work 247 with no complaint around the clock and humans can't do that. So the question is like, do you want to try to match the output and style of the tool or do you want to use the tool to match your work style? And I think a lot of people are like, well, I'm just going to work all the time. Because my agents can work all the time. I'm like, but they're machines and you're a human. Yeah, I told you guys, like when all the multiple stuff happened, I was like, this is the first thing. I'm going to let other people be the guinea pig and early adopters and figure it out. I'm just going to watch it for two weeks because I think the cost of becoming an expert in this is high relative to like my perception of the benefit of doing so.

Yeah. My Instagram has been flooded by everyday people running out to buy Mac minis to get their cloud bots started. So I'm happy to stay out of that for a little bit.

I'm good.

It's fascinating. Sitan also talks about there's a point of diminishing marginal returns with just working with AI generated code too. And specifically from my perspective, which is one of a non technical background, I totally resonated with that. I can get like to 80% creating something and then to get my output exactly where I want, every time I re prompt and recontextualize, it gets me like half of a percentage point closer. Half a percentage point closer. And then he talks about how you spend way more time trying to write the perfect prompt versus creating the best output. And that's where my technical expertise just shows because I don't have that engineering baseline. And I love the way that you said even though you think that you're like a junior AI intern, he does say that the best engineers that he works with kind of think about their AI coding assistants as truly like an intern, that they have to do the first draft and really take a close look, but they can push out things pretty quickly. I asked, actually, one of my friends is in engineering. I'm like, do you feel like it's like managing a person Then, and that's that part of the issues, like you really didn't have to manage as many people in your role and now you have to. And he's like kind of worse, I would say, because I can't hop on the phone and have like a 30 minute conversation with somebody about, okay, so where did you learn how to do this? And like, why do you structure code like this? Let's walk it through why this doesn't work here. So to your point about if there's like a deeper database of historical code that's so much richer, but if you're building something from scratch, you don't have that, so then you don't even know where things are shifting and changing from your perspective. And we tinker around with AI tooling all the time, I guess. How do we strike the balance between knowing when to stop and not giving ourselves the best shot at upskilling in a consistently evolving space? I mean, this is likely a question for both technical and non technical folks. So understanding. Okay, let's take a break. We don't actually have to spend every weekend reviewing every single update and refresh of a new tool that came out. I find myself spiraling as well, just wanting to be up to date on what's the latest. But you know, at what point are you sort of like, what additional benefits are you gaining from spending five to six more hours on reviewing a new tool that do effectively the same things that you've been doing? Something else that works half as good?

I found that I've gotten more productivity and joy from getting better at a smaller number of tools. And instead of trying to be proficient in everything, I'm like, how do I get the maximum amount out of cloud code? How do I get the maximum amount out of make? How do I get the maximum amount out of like the things that I use as opposed to being like, oh, I can use like a hundred tools. I just have found it's more satisfying to feel like I think I've gotten much better at Claude code and I've gotten much better at using APIs than I was before, which was like not capable at all. There are lots of other people like, what about this thing? I'm like, that sounds wonderful. I am not going to learn how to use that. I'm going to stick. Because right now I feel like for the things I want to build, I have sufficient tooling to build the things. It's actually a nice constraint. And if I ever find there's something, oh, I can't build it with API access to notion, airtable make and cloud code. Then I started to ask myself like, is this really something I should undertake?

Yeah.

Is this like an educational project? Is there a simpler way to build this? Is there somebody who sells me a software product that would meet this need? And maybe I think sometime next year there's going to be this builder backlash which people are going to have a bunch of things that they vibe coded or spun up but nobody wants to support anymore. And it's just going to everyone's going to be like, oh, we built this thing but nobody wants to support it. And now like a bunch of the underlying APIs have changed. It turns out maintaining software is not trivial. And I think sometime in the next year people are going to realize even vibe coded tools require maintenance.

Well, our next piece is an article written on X by Will Manidas titled Patient Capital Will Eat the World and I genuinely enjoyed this one. The article argues that traditional venture capital is fundamentally broken and must transform into something resembling patient permanent capital structures or die. The structural trap is that VCs are stuck in a doom loop. LPs demand capital deployment on schedule regardless of market conditions. Fund structures prevent obvious smart moves like concentrating capital, sitting on cash, buying secondaries at discounts, or returning capital. In general, VCs must stay busy, they must deploy, must have opinions on every deal, even when the smart move is maybe likely to do nothing. Will claims that the winners will be firms that build massive patient product agnostic pools of unconflicted capital structures similar to Berkshire Hathaway sovereign wealth funds or generational family offices. No deployment clocks, no mandates, no LP politics. They can hold forever or sell tomorrow. They can concentrate, they can diversify, and most importantly, they can wait. Ultimately, traditional 10 year venture funds will survive at the margins, but the bulk of capital and highest returns will move to permanent capital structures. The funds that will survive are already making structural changes to their firms, he said, to prepare for this change. And the funds that won't will continue raising and deploying and marking up positions until it all fades away. I mean, this was largely an existential piece on venture as an asset class. If you read it, we've talked a lot about how much venture has changed since it was a mere cottage industry. How much of Will's perspectives and predictions do you personally agree with Char?

I think a lot of the things he mentioned are already happening. They're just only happening if you're really, really inside the tent. If you think about, you know, there's a bunch of like snapshots in that post if you look at the kinds of deals that General Catalyst is doing, the kinds of deals that Andreessen's doing, the structural changes at Sequoia that become a very different structure, these are all structures that are being done in recognition that like the ten year fund life cycle, I guess I would say there's a difference between the 10 year fund cycle being dead and the notion of a closed end venture fund with a defined start and end point being done. I read this and I'm like, who, what, what fund manager wouldn't love what he articulated the question, do LPs love it as much? And I think we're going to try everything else before we get to the state of the world that Will describes. I think we're trying right now a couple of things. We're trying extending Fun Life and just saying, let's just all agree that 10's not real. It's probably more like 15. You know, the nice thing about a 10 year closed end fund is if in theory, if you're an LP and it's the end of 10 years, you're like, I actually don't want to give you the ability to roll my money. I actually just want it back. This is the end of the road. I'm not going to give you permission to keep managing it. Your job is to now liquidate these holdings and give me my money back. It's kind of nice to know that your money is locked up for 10 years. And it's unclear to me that LPs will have. You know, as someone told me, there's a big difference between a 3x fund in 10 years and a 3x fund in 15 years. If you look at it from a purely IRR basis, then the person's not wrong. So at some point, I think what he describes works great if it's your money, right? But with someone else's money, the ability to hold it for a really long period of time is hard. And so I think we need to solve two different liquidity problems in venture. One is liquidity at the single position level, which I think he touched on, which is there's just not that many companies that the public markets seem enthusiastic to take. So we have to find a way to get liquidity for private companies. And you're seeing these big tenders and you're seeing, you're seeing these things happen. And we have to figure out how to get liquidity at the LP holdings level such that if an LP has been in a fund for 12 years, but these problems both have the same Root, which is, if you're in a fund that you've been in for 10 years that doesn't have a bunch of interesting companies in it, who's going to buy it from you? The manager's not going to buy it from you. It's not like a mutual. There's no redemption like who's going to buy it from you. And I think with this, all the vintage folks call it concentration squared, and I like that term. We live in a world where there's a really small number of assets that people seem very excited about and maybe they're irrationally enthusiastic about them, and they have a bunch of other companies that are good and nobody cares and nobody wants to buy them. And the ones that nobody wants to buy, it's hard to get them public. It's also hard to find a buyer for your shares. So you end up with a bunch of stuff where you're like, you have 100, a company that's doing 150 million in ARR in your portfolio, but if there's no bids for it, what's it worth? I could give you a pretty wide spectrum of ways to think of what that company's worth, but if no one will pay you for it, it's probably not worth very much, even though it's a good company.

And some of these new strategies that are emerging, like the continuation vehicles or opting into secondaries, it's interesting that we're going to see more funds tried out because those aren't options available to every type of fund as well. There'll be newer things that come up that work for one type and then do not work at all for a different archetype. So I left reading the article wondering if you weren't a mega fund and you look a lot different, how imminent is the death of your type of venture world in the way that he describes?

I mean, when I talk to LPs, I tell them it's like the classic being in the middle is really hard. If you're really small, like us, you can sell into secondaries. You have all of these, like, liquidity options that you can use because you're not on the board and because you're not a big shareholder and nobody, basically nobody cares. And if you're really big, you can create vehicles and you can invest in infrastructure that will allow you to hold on to things, and your LPs will will allow you to do that. And also oftentimes you're such a big shareholder that you can't sell in the middle. It's tough because you're big enough that if you sold, people would maybe notice and care, but you're not big enough to have all the infrastructure of a fully built out RIA and a bunch of other products. So the problem is, like all asset management businesses, the goal is to grow assets under management without regard to the quality of opportunity set out there. I talked to someone over the break and I was like, well, what are you doing? Because I'm just not investing right now. I'm like, what do you mean? He goes, I don't see anything I like, so I'm just not investing. You can do that as a GP if your LPs believe that you're not deploying capital because you don't see opportunities versus like you're just too busy skiing or hanging out or you just, you feel lazy. And there I've had friends who at the end of, you know, in 2021, did the hard thing and just said like, this is too insane. We're just like not doing it. This is like too insane.

One last question on this one. From your perspective, what might be harder being an emerging manager, building a fund with no playbook to work off of right now since the industry's evolving, Evolving just as you're getting kind of like your training wheels in motion, or being an established firm and needing to reimagine what's worked in your fund in the past to survive? Looking into the future, I would definitely.

Rather be established right now because I think the other signals that LPs have been able to use in the past are usually like markups by big funds. And I think the big funds in a lot of ways are setting the agenda for venture because they have all the money and they're like, these are the things we care about. If you want to be a part of our club, you should invest in the things we care about. If you don't want to be a part of our club, you. I have to. You don't have to.

Well, our last piece is from Julia Maltby, titled 5 Tips for GPS Building a reference list that actually helps you close LPs. Exploring how reference checks are one of the most important parts of LP diligence on fund managers. So much of the job is what she calls professional gossiping, texting someone who knows someone to get the kind of ephemeral information you can't extract from a memo or a deck. Julia walks through how to build a list that actually moves the needle. Pick references that showcase both investment skill and strong firm operations. Because being great at sourcing deals is very different from being a dependable, communicative fund manager. Include co investors who reinforce your market reputation and back up the edge you're communicating in your materials. Make sure everything checks out. If you told a specific sourcing story, the reference better tell the same story. Choose LP references with nuanced context on why they invested, not just name brand firms that could appear on anyone's list. Include founder references where things didn't go as planned because the most powerful thing a founder can say is they choose you again even after a difficult outcome. And lastly, prep your references with specific stories so they can speak to you in the ways that will be most impactful for yourself. Having the experience of an LP and gp. Are there any specific tips referenced here that sound intuitive in nature but actually much harder to put into practice in real life?

There's this tension between the people who know you the best and the people who, like, look most impressive on paper and, you know, through the stuff I do at Screen Door and my small personal LP portfolio, sometimes people will say they'll put a reference and I'm like, wow, that's like a big name person. And I'll email the person, yeah, we did a deal together. He's great. I'm like, what else can you tell me? No, he's great. He asked me to be a reference and I said yes. And I'm like, okay, all right, thank you very much. Super helpful. That didn't tell me anything at all about why I or any other LP should like, give this person money. That. That feels like a miss.

When they list that. Does that speak volumes to you more of like that decision to list them?

It means one of a few things. And I think Julia's point is funny. What she described is now the way I did not do all of the things that she mentioned in here in our early funds, partially because I didn't know better and partially because it would have been hard to do. I do all of these things and more now, and I think it's really important. And the best GPS I know will send me an email. Hey, I'm fundraising. This is the story I'm telling. I would like you to be a reference. Or even if you're not on my list, I think you might get called. And here's the story we're telling. Here's the companies we've worked on together and I've thought really hard, how do LPs reference GPS and what can I do to make sure the people who want us to be successful are armed with what they need to be successful. And I've also met people who've given me references and I'm like, well, this person's like, kind of obscure. And you get on the phone and the person knows so much about that individual. And I think you want to strike a balance. For me, the most important thing on the reference calls is that the person can actually speak to your work. I think it's also always cool when you get LP perspectives from people who were on the GP side before or on the investing side before. I'm enjoying Julius writing because I think it comes from a different place than a lot of other LP content.

We also offer references for founders who are considering an investment from Precursor. So how deep is that overlap between the pool of people that you would recommend, as in references for founders, and then references for LPs? Like, what are the biggest distinguishing factors that you feel like would make you want to put one on one list and not on the other?

It's interesting. I think with LPs, there are some LPs I talk to who only want to talk, talk to really successful founders. I talked to him. It was very eye opening. And the person was just like, look, if you have a bunch of founders who think you're the greatest thing in the world and you're super helpful and all those companies fail, I don't know what to do with that message. It means you're like, very relevant to companies that are very unsuccessful. And I thought about, I was like, I know what you mean by that. And that's like a fair criticism. And I know some people who have one or two silver bullet references with really famous founders who will basically just say, this person's great, you should give them money. And that's, that's, that's good enough. What I've learned is everybody's looking for something different. I know some LPs who are like, I want to know who the most successful founders think their most useful investors were. Because if that person can make Dario or Sam Altman really successful and they say this person is really helpful, that means something. And I'm sympathetic to the idea that hearing that somebody is an amazingly available, helpful, useful vc and none of those companies have been successful. You're getting like two messages at once. So I know, I also, just as a general rule me, I assume that I just did a call with an LP this morning and at the last five minutes they asked me about two fund managers. I know I'm not on those people's list. They're just Like, I know you're connected to them. What do you think of them? And I was like, oh, I didn't come into this prepared. I was like, what do you want to know? So I just assume that like most LPs who really want to know who you are and what you're about are going to ask many people who are not on your list what those people think of you. And that's just, I think, a good incentive to just try to be a good person every day and try to be helpful because you never know. One LP told me a funny thing. He goes, you know, 100% NPS as a GP is a bad thing. I was like, what do you mean? He goes, it means you probably have never had a hard conversation with anybody or had to make a hard decision and you've like prioritized keeping people happy over doing the hard parts of your job. And he's like, a 20 is bad too. I did. It did make me think you're right. If I'm doing my job correctly, at any point in time there will be founders that are upset with me because I'm not doing our pro rata and their follow on round or I'm not supportive of some decision that they want to make. And I get it.

I mean this is really relevant to her tip, which was including a founder reference where the story isn't a successful exit does feel like a more vulnerable move from what you had said earlier. Any guidance on actually what LPs would actually find remarkable on those calls? I'm trying to just differentiate between qualities that actually separate one of one fund managers from what's actually just like a really low bar for VCs when an investment doesn't go as planned, which is just like being a good person. So I find with the folks where it's not a successful exit, the things that I would want to hear if I was an LP would be like, lead with grace really gave me the time of day. Is there any, anything else that should surface as more of like, this is truly exceptional?

I think the best LPs I work with generally are like, tell me what it is that you do and why you think you win. And do the people who I talk to tell me the same story. And I asked one of our LPs like, well what do you talk to the founders who it didn't work about? A lot of it is just a, what role did you have in helping them figure out that it wasn't going to work out? How long did you sort of stick around during that process and more importantly, how does that founder feel about. And one person was like, yeah, when the company wasn't working out, he ghosted me. And that LP was like, I'm actually okay with that. If it wasn't working out, while it would have been good from a relationship standpoint for that person to remain involved, I'm actually okay. That person just said, this is not worth my time. And while the founder didn't love that, the LP was like, this is the GP who's doing with their time what I'd like them to do. I could give you other LPs to be like, oh, that's a terrible long term relationship management decision to make to cut bait. And part of it ends up being every LP has a set of beliefs about what makes GP successful and what are the most important things to do and get right. And to some extent, like, they're looking for people who embody their values. And if their value is that, hey, we stick with people till the end, the person who cuts bait isn't going to rate as high if their value is, hey, we just want you to spend time on the best companies that are winning and the ones that don't win, if they hate you, who cares? And I've met LPs with a wide variety of views on what a good healthy founder VC relationship looks like. And they tend to pick people whose values are aligned with theirs.

Well, we are all out of time today. Thank you Charles for jumping in. Thank you to our listeners and subscribers for joining as well. Please like and subscribe and we will talk with you all next week. Thank you, Sam.