Founders Are Pulling All-Nighters to Keep Up With Their AI Agents

Educating the market is kind of a loser's bet if you're an early-stage company. How do you attach the thing you're doing with the thing that people already care about so that they pay attention to the thing that you're doing? And I've told a lot of founders in our portfolio, there's nothing wrong with your company except that nobody cares. It's just not that important.

Yeah.

And you just gotta understand that like there just might not be a venture audience because most venture firms just need outrageously large exits for the math to work. And if you're not one of those companies, you're literally not worth their time.

It's the week of August 24th. Welcome to The Learning Corner, a Precursor Ventures podcast. Welcome to The Learning Corner by Precursor. I'm Mia Farnham, 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. We're back in AngelList studio for a second week in a row, so we love being part of this. Thank you so much for everyone for making this happen. Our first piece this week comes from The Wall Street Journal, written by Catherine Binley, titled Startup founders are working harder than ever to keep up with their AI agents. The piece profiles a handful of founders who are discovering something paradoxical about the AI era. The more their agents can do, the more the founders themselves are working. From pulling all-nighters to monitoring agent requests from an Apple Watch mid-run, these founders describe managing AI agents in almost romantic terms. It's seductive, it's intoxicating and all-consuming. The piece captures this growing tension in the startup world where the speed unlocked by agents is real, but so is the human cost of keeping pace with them. One founder summed it up really well. Quote, every minute that I'm not working, I'm missing out on not doing a week's worth of work. We've talked a lot about the importance of judgment and taste a bunch of weeks in a row. This piece really isn't. necessarily saying that these founders aren't leveraging those things, rather that they're feeling like the clear bottleneck on progress since their quote unquote workers never take breaks and they do. They literally reference 8 hours of sleep as being a bottleneck. Is this just becoming the norm for what burnout looks like in this new era for founders?

Yeah, I just think when we talked about 996 a while ago as this sort of anxiety-related cope, Mm-hmm. I think some of that, people know that these tools are always available. So it feels like if I'm not working all the time and I'm not giving my agents work to do, I'm missing out and I'm falling behind. Sometimes it's true. I think the other thing, and I've had this experience, sometimes you wake up in the middle of the night or you wake up early and you have an idea and you're like, oh, I should just give it to an agent and have it start working on this thing. Or I find myself sometimes in line for coffee. I'm like, ooh, I have a thought. Let me ship this. off to Hermes and have it work on this. It's not always healthy, but I understand if you believe that everybody you're competing with, either directly as like a market competitor or more broadly in that marketplace for attention and money, is working all the time and has agents doing things all the time. It's natural to also feel like I should be doing the same. Whether that work all needs to be done, is useful, is well thought out, is a whole nother question.

And I think in the piece, What's left out is remembering that you yourself are not an agent.

Yes.

Because your brain can atrophy over hours of time of overworking. And so I guess, how do you figure out, Charles, or determine when's the best time to have an idea and just wait a little bit versus I need to run back right now and push this?

It's really funny. One of our portfolio company founders and I were hanging out and he asked me, what do you use for ticketing? I was like, I don't have a ticketing system. I build all this stuff by myself. He's like, you need a ticketing system. I was like, why would I need a ticketing? Who's gonna fix the tickets? He's like, well, agent's gonna fix the tickets. And he's like, more importantly, you need a place to write down all of these ideas you have so that you don't have to immediately start working on them. And he was right. Before, I just wrote them down in Obsidian as, as scraps. But now with tickets, it'll say, oh, you have these 3 tickets that are all kind of talking about the same thing. There's a theme here.

Mm-hmm.

And I just found the ability to write these things down and put them into a backlog or put them in a place where I could review them. Has saved me from sometimes going down a rabbit hole on something that actually in the end, given an hour to step back, actually wasn't that important.

Do you find that this is mostly relevant for you because you're both the customer and the founder of your own tools? Do you feel like if you have like one human in the loop besides yourself, it would alleviate some of that?

Uh, I think it'd either make it worse or better. I was talking to somebody who had the misfortune of building a CRM for their company. They got rid of HubSpot. And he felt really great about it. And I was talking to him this week and I said, what have you learned? He goes, I don't have time to be a part-time product manager. I'm like, what do you mean? He's like, all these people on my team who are using the CRM, they have all this stuff that they want me to build.

Yeah.

I was like, oh, you mean like new features? And he's like, yeah. I'm like, well, you're not a product manager. That's what you were paying HubSpot for. So that kicked off this huge conversation around the future of headless CRM and things like that. But I do think no one ever tells you no. The agents won't tell you, No, they might say, hey, this isn't that important, or maybe you should take a break, or that's quite a bit of work, but they'll almost never tell you, no, I won't do this for you unless you've hit your usage limit.

The article does reference unhealthy amounts of time put into just relieving the bottleneck, which can amount to like skipping meals or just not sleeping altogether. It made me think of the part of this of just pushing things to production is largely related to how long you can just Be at your desk and just hit yes. It reminds me of, I don't know, like investment banking of like, it's not necessarily the work itself when you're an analyst is all that difficult, but it is who can stay here the latest to push things out so that you're never the one that folks are waiting on for things to move through. Again, we balance so much conversation around judgment and taste, but this article specifically just focused on brute force of sitting in your chair and continuously hitting things so that things don't get— locked up by you. One of the founders in the piece described working with AI agents as a drug, like using one. And another said he can't be present with people outside of work because the agents demand his attention. Where do you think the floor is here, or is it just the new baseline for early-stage founders where we are today? Sometimes I think, and we think of this for 996 as well, a lot of it is chatter and just posturing a little bit. What do you think?

I think being a founder puts most people in an emotionally vulnerable position where you're constantly anxious about whether this thing is going to work out. Am I good enough? Am I working hard enough? Am I doing the right things? Am I taking advantage of all of the tools? And in some ways, like AI agents are the perfect drug for an emotionally vulnerable person because they can work endlessly. They mostly are sycophantic and want to please you. They can do just about anything you imagine or can articulate. And you know, everybody else that you know claims to be using them, whether they're actually using them or not, it's another issue. And so I think it, in some ways it's like the perfect storm for emotionally vulnerable founder who's trying to get a business off the ground and trying to find product market fit. It's, oof, it's kind of scary.

Also different from actual human employees. You as a founder can run around being like, I was so productive. You should have seen all the stuff I did. Your agents aren't gonna be like, we didn't really do that much. Just like ran in loops for hours. Our next piece is from Gorghe Oros from his Pragmatic Engineer newsletter titled Headed for the Exit: The Great Engineering Leader Career Break. The piece reports on a real wave of CTOs, VPs of engineering, and heads of engineering walking away from their roles, often with nothing lined up next, and digs into 10 common reasons why. The list runs the gamut from founder mode making the CTO role feel low ROI to equity grants that preference stacks have rendered nearly worthless to a striking new pull factor. AI labs like Anthropic are now paying individual contributors a median of $1.25 million a year in total comp, which is more than most executives earn at mid-sized startups. The piece also names a growing burnout pattern that threads directly with our first article, where the intensity of AI-assisted work is producing real crashes, with one engineer describing months of late nights and nonstop prompting followed by a hard collapse. There is a hopeful counterpoint in the story of Matt Boyle, who joined Gitpod as VP of Engineering, helped it pivot and rebrand Ona, and watched it get acquired by OpenAI. It's a reminder that the right company with the right mission can still make the role feel worth it. The piece makes a pretty pointed argument that founder mode has essentially made the CTO and VP of Engineering roles feel low ROI at most startups. It's funny because I look at The number of talent asks we get from our founders, and they're all engineering related, but they're mostly for entry-level front-end and back-end combined support, less on the leadership end because they're really just trying to ship things quite quickly. Is this a feature or a bug of how startups are being built right now across the market? Oof.

Can I say both?

Yeah.

It's funny, related to your other question, like, would it be better or worse if there was someone else using my infrastructure? I remember when I saw Gary Tan's G-Brain and all of the reading these things about people shipping 60, 80, 100,000 lines of code into production databases. And, you know, I would describe myself as a very, very, very junior software developer at this point. And I know the cost of mistakes and that's mistakes without a power dynamic.

Mm-hmm.

And I read that one part where there's one engineer talks about like, what do you do when your CEO just rage pushes 60,000 lines of code into the production database, feels great about it. And then you're kind of left to clean up the mess and to deal with all the tech debt, all the things that could be imposed. And if anything, learning how to code with AI has made me so much more respectful of like the craft of software development. It's speed, quality, or cost.

Mm-hmm.

And I think right now we're running this experiment, which is like, what if we just went all in on speed? What if we decided like speed is the most important thing? Quality. Yeah, sure. It matters. But speed is the thing. And that felt like the through line in all of those interviews is everyone's like, well, we've just gotten rid of all the things that made me valuable as a VP of engineering, which was the ability to say no to things.

Yeah.

Or the ability to prioritize the roadmap or the ability to like enforce some quality standards. Those things are not as important at the moment as sheer speed of shipping. And I think we will come back to a more normal state of affairs where there's a better balance between quality and speed, but that is not where we are at the moment.

And the prioritization on speed right now, I think, feels so experimental because we've never had a world like this before. It's usually come at such a premium of you need to hire somebody, you have to interview them, you have to see if they're technically advanced enough to join your team. Now you can just start running. So we've never actually faced this before. And I think for folks who've grown up and become leaders, Across tech, um, and specifically within engineering, it's difficult to like learn this game and then advise people on what the strategy is as well. Um, but across startups, especially on the early stage front, very little roles are being hired for people with that caliber of experience because they want people pushing things to production.

Yeah.

As opposed to providing counsel on soundboard, mapping everything out, and then making sure everything is correct, which as much as I'm an investor and I want people to be thoughtful about this and intentional. It's really hard to sit across the seat from a founder who knows how many competitors are building in the space that they are and tell them to wait 4 or 5 months to really be specific about what they're trying to build and hire somebody of this caliber to join the journey. I don't know if you've had to balance those conversations.

We have, I think, in the pre-AI world, Planning was at a premium because engineering cycles were expensive because humans could only write so much code. It took a long time. And if you built the wrong thing and you built it poorly, the costs were expensive. And so planning had this like huge, huge benefit, which I think favored experienced engineers who understood process, who understood architecture. And now we've gone to a world where I wouldn't say planning has zero perceived value, but planning's value is probably—

collapsed 100x. Yeah.

So people are like, well, what's the point of planning when I can just rip code? If I don't like it, I'll back it out. I'll redo it. I'll fix it. And I think what's lost is some of the like elegance of architecture and some of the thoughtfulness of like anticipating problems so that instead of having to fix things, you build them right the first time. But in a world where like, well, look, well, if engineering cycles can run 24/7 and I can spin up more agents, what if I get it? The cost of fixing a mistake has gone down. And the question's like, well, now what's the role of planning? When building is cheap and fixing is cheap, I don't think the answer is zero, but it feels like the market is essentially saying for most cases, the value of planning is far less than it used to be.

And this still hasn't even touched the feature of when customers are paying for it too. Like, this is all just like running things through a simulation, but the moment you start introducing people into paid versions of what you're offering, you running experiments day to day to compromise on quality and not doing the planning like that. Really can collapse your value. One piece that stood out to me was the fact that these leaders are feeling that their original skillset isn't as valued in the setting as founders require everyone to be completely AI-pilled across the organization, or that they're sitting in organizations that have not evolved with the market.

Mm-hmm.

And so they feel like their skillset will become atrophied if they sit there any longer. So it's almost like 2 sides of the range.

Mm-hmm.

Is the answer always to leave the current role though, if you're in either bucket? And I guess you could expand that to not just leadership, but if you're entry level as well.

Most people can't do AI transformation. Either they themselves don't know how to do it, or the organizational antibodies will get in the way. And I do think there is something to be said for starting from scratch, but again, I don't think starting from scratch means you throw away every best practice that worked before. I do think it is easier to build AI-native products with people who have fully embraced AI.

Yeah.

And if you go to an AI-native startup, you're by definition surrounding yourself with people who've said like, we're going to build this way. And I think what's hard for some companies is like, you know, you've got 10 people in the org, 3 of them are AI-pilled, 3 of them are AI-curious, and 4 of them are like, this feels like a total joke and a waste of time. And depending on the configuration of any given team, you might get maximum AI throughput or you might get maximum AI resistance. This is something that's come up in our portfolio companies. Some of the more mature ones were like, hey, if we didn't go hardcore in AI, we knew we had a lot of engineers who were thinking about their future. And if we didn't lean into this and give them the chance to build and develop these skills here, we were going to lose them. And I've been involved in a handful of our portfolio companies where I told them, you might need to think about reorganizing the leadership of your engineering team.

Hmm.

To put people who are more front-foot AI in charge because the leader's gonna set the tone and gonna dictate the pace at which this stuff gets implemented, adopted, used, and just a general policy and tone on it. And we've had a couple companies where it took switching a leader to get them to really embrace the technology.

And then on the other side of this, if you are like a seasoned engineering leader at like a large scale, call it like a Fortune 500 company, and you know things are moving Slowly. Do you leave and abandon all that and go run back into startup world? And I'm talking about people who have like 30-plus years of experience. Like, is that what you do? Or I don't know, you just call it what it is and then decide what to do. Like, I think it's really interesting reading the article, walking through different people's perspectives on what's important to them now and like, what do they want their career to look like in the future and, and what will actually matter from a skill set and trying to hedge what's the best I think big companies are trying to do tech transformation too.

That's why Anthropic and OpenAI and all these big labs have these forward-deployed engineer consulting teams and they partner with the Deloites. So maybe the real answer is if you stay at a big company, the opportunity is to learn how do you do AI transformation in a scaled organization?

Yeah.

And that's just probably a different answer than how do you build from scratch an AI-native startup. And you might conclude, hey, I've been in big companies for 30 years. The idea of going to a 2-person startup isn't something I want to do. Lean into the chance to learn, well, how do you actually get a 10,000, 30,000, 50,000-person company to adopt AI in a meaningful enough scale that it actually impacts the business?

Yeah. And that's a skillset that doesn't collapse just because brand new tech has been released into the ecosystem.

Yeah.

Well, our last read this week comes from Yoni Ovitz-Slow on his Substack 99D titled What capital wants. The piece is a take on what the current fundraising environment actually demands from founders right now, and the answer is both simple and a little bit daunting. Growth investors have essentially converged on one bar: show a credible path to a $20 billion company with you as the emerging winner. The piece also makes the point that category-based investing has quietly died, where fintech investors no longer just want to be fintech investors, healthcare investors no longer wanna just do healthcare, and everyone is chasing whatever the current wave of conviction happens to be. He closes with 3 paths for founders navigating this moment: find true believers, educate the market, or find a way to attach yourself to the things capital already knows it needs to do. He mentioned the internet hype chart that we talked about a couple weeks ago of category winners not being funded in the period where there was the most heat around them.

Yeah.

Ultimately, this turns into the phenomenon that he supports, which is that seemingly category investing is over and everyone just wants to be in the current thing. I've never had so many conversations with my friends who work at a generalist fund, just like call it what it is, but don't wanna say that that's the thesis.

Yeah.

They've created like every orientation of what they could say without saying the word generalist. How can you turn this into advice for founders who are building out their lists of possible investors to go pitch in front of? And be realistic that there's hidden meanings and things that investors are looking for.

I think investors change strategies faster than founders can keep track. I, we announced our second nuclear investment this week and someone called me and says, I want to see more of the energy deals. This is someone who is at a specialist venture capital firm that couldn't be further from energy based on everything they've done for the last 15 years. And I was like, wow. And I think if you think about everything we've talked about with concentration, this is the natural Endpoint, which is, okay, if there's only a very small number of companies that are gonna matter from a return standpoint, if your strategy keeps you out of those companies, it's a bad strategy. And so I've never met more people who are like, I'm like, wait a minute. I thought like, nope, we went from SaaS to crypto to AI to robotics. We are on that pipeline train. And as much as I think it's cynical, I also think it's realistic. It's, it's, I think venture is realizing that, hey, being in a lane that doesn't produce breakthrough huge companies is not a way to remain relevant in venture today. And it puts your firm's strategy at risk. So to me, the only 2 answers are everyone is an AI firm because AI becomes a meaningless term because AI is everywhere and it encompasses everything. Or that this is a temporary blip where people just say, God, I have to now rebrand my health tech firm as a space and defense firm.

Yeah.

And yeah, I've met a lot of people who pivoted and they're like, well, you know, we've always been interested in this category. It's just, we're reframing what we do. And I'm like, it's also just okay to change.

Totally.

It's okay to just change. And I find it hilarious that when I started Precursor 12 years ago, the 2 things people told me is you have to focus on ownership and you have to have a sector focus, otherwise you'll never make it. And I'm like, well, it appears right now like neither of those things is true. But one thing I liked about Yoni's article is like, it presents 3 choices, but I think for most founders, you only have one. Educating the market is kind of a loser's bet if you're an early-stage company. Like, you, you won't. Finding the true believers is great, but not reliable. The third thing is the only thing that's available to most people, which is how do you attach the thing you're doing with the thing that people already care about so that they pay attention to the thing that you're doing? And I've told a lot of founders in our portfolio, 2 things. One, there's nothing wrong with your company except that nobody cares. And they're like, what do you mean? I go, nobody cares about your category. Nobody cares about this market. It's just not that important.

Yeah.

And you just gotta understand that like there just might not be a venture audience. And second, I think founders have a hard time grasping this. $5 billion company is the minimum. That's like the minimum interesting scale for a venture-backed company. And really, What Yoni said is true. People are underwriting $20 billion plus outcomes, and in many cases you are far better off being an infinitesimally small probability $20 billion outcome than you are a high probability $5 billion outcome. Mm-hmm. Because most venture firms just need outrageously large exits for the math to work. And if you're not one of those companies, you're literally not worth their time.

I just have a lot of empathy for founders too who are building at the earliest stages, 'cause I was even thinking about Handful of the exits or acquisitions that have happened over the last handful of, of weeks, or the Hugging Face potential deal that's at like, what, $13 billion?

Mm-hmm.

I don't think when anyone started that they were like, this market is worth $10+ billion, but it evolves over time. And I think that's never been more important to founders in the first meeting with investors pre them doing anything and setting out a roadmap of like, I need to showcase how this is $5 billion plus. Market, that's hard. There used to be a time with an early-stage venture where you could say, this is where it starts, this is what I think it could grow to, but you could be a bit more measured on what you're thinking. I just hopped off a conversation with a founder who's building within forestry and agriculture, and technically the value of that market is uncapped, but where we were focusing was. And so that's the conversation I had to have was, if we focus here, I think you have to be realistic on what that market size is and communicate that and be ready for some investors to just say no.

Yeah.

You could also just venture into the land of building out this more elaborate story, which speaks to the third one of like where this could go. Now, is that what your intention is and what your dreams and hopes are for the company? I don't know. I'm not you, but that's a new dynamic that I think every founder is existing in. And it's even for the folks who sit outside of consumer and CPG, which I know are the ones who feel the most out of favor here. It touches all of them.

I also just feel like if you started a company 2 years ago, Precursor, pre-OpenRouter, like the world has like literally changed while you were building.

Yes.

And you're like, oh, well, when I started this company, I knew I needed a multi-billion dollar outcome. And now the world is telling me I need a multi-tens of billions of dollars outcome. No fair. I'm like, well, yes, it isn't fair. The goalposts shifted dramatically. And I think what we're going to see is just a continued churn at the bottom end of the market of companies that get started. That make real good progress, but they don't look legible to Series A and Series B and future investors. And they're just gonna say, this is a great $500 million to a billion dollar company. It's just too small. And those founders are gonna scratch their heads and say, do you know how life-changing that outcome would be for me and everybody at my company? And the VC's gonna say, that is both true and I don't care.

Yeah. For founders who are trying to convince investors about the market, And get them on board, what would be your advice?

One is you tell people a story that, hey, the thing I'm working on is going to be the next thing. The open weight, open source stuff became really important when people started to care about the cost of AI. It was, I think, predictable that at some point people would care about cost. So you had to be around and be in business to be OpenRouter or to be OpenCoder, to be whoever you might be to take advantage of that trend. So one thing is I think you have to tell people we are the next hill. Like, hey, we're doing security and compliance for AI agents right now. Nobody cares as much as they should because there's so much desire to push this stuff out. But any day now, CFO is going to get involved. The chief investment security officer is going to get involved. We are the next event horizon for this category. Or you just have to find the small set of VCs who are just not totally at that hill. I would say there's maybe, 20% of our industry who's like, I'm open to things other than enterprise AI. I wanna see things that don't fit that mold either because I know how to get those businesses financed or I have some other reason to believe that I can help those companies succeed.

Or I'm just patient enough to see how that works.

Yeah.

My last question is, when you're talking to other folks who are shifting around their strategy on the investor side and talking about being more generalist and not being very specific on like, What within generalist land they're focusing on. Do you think more of that play is for founders or for LPs of just, I find a lot of people are like, we're generalists, but like right now we're really focusing on defense, deep tech, et cetera. And I'm sort of like, I'd wish you'd, like they change their theses on their website every single day, it feels like. And I'm sort of like, is it signaling for founders or is it signaling for the fundraising cycle?

I assume most of this is done in the name of both.

Yeah.

But I think if you go into an LP meeting, let's say you're a health, uh, let's say you're an education fund and everyone goes like, well, where are the big exits in education? All of the action seems to be in enterprise AI. You have to say, don't worry, the exits in education are coming. Or you have to say that, hey, AI is going to transform education. It's going to create a bunch of companies that will matter. Or you have to say, you know, everybody's learning about AI right now. So therefore AI adoption is an educational theme. And so this is an extension of our strategy. to invest in AI harness companies and like, you know, and token counters or whatever it might be. And I think at the end of the day, most LPs want to make money and they're trying to compose. This is the other irony. LPs would always tell me before, well, I want specialist managers so that I can create my own portfolio. I want a health fund, a fintech fund. I'm like, well, everyone's doing the same thing now. So no matter what strategy you thought you were backing, you're probably backing AI.

Yeah.

I assume More of this energy is going into marketing these firms to LPs, 'cause I think you can change LPs' minds faster than you can change founders' minds.

Mm-hmm. Well, we are all out of time today. Thank you, Charles. Thank you, AngelList, for letting us hop in again. Thank you to our subscribers for jumping in as well. Please like and subscribe and share with friends, and we'll talk with you all next week. Thank you.

Bye.

Bye, everybody.