Everyone Is an Engineer Now | 40 Buyers and Nothing to Sell | What AI Can Never Cheapen

I have a presentation for our team tomorrow, which is about the 20 agents that I have in production and what they actually do, what they've done for me. And in building the presentation, Cloud Code was like, oh, this one's been failing for a week. I was like, what do you mean? Yeah, it just stopped working. It's a problem I have to solve. And I think that's the thing I'm learning is that most of the things you need to know, it's far easier to learn them when you encounter them and bump your toe.

It's the week of June 1st. Welcome to the Learning Corner, a Precursor Ventures podcast. Welcome to the Learning Corner by Precursor. I'm Mia 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 a really well-researched essay by economist Alexi Moss titled What Will Be Scarce? And it makes a really counterintuitive case for what happens to human labor when AI can produce nearly everything a person can. The core argument is that as automation drives down the cost of commodities, real incomes rise and people predictably shift their spending towards things that automation cannot cheapen goods and services where the human element is the value. He calls this the relational sector, which includes nurses, therapists, personal chefs, bespoke craftspeople, hospitality, and care workers. He leans on a behavioral insight from René Girard called mimetic desire, the idea that people want things more when others can't have them. Research he ran found that willingness to pay nearly doubled when people knew others were excluded from the same product. The kicker in this argument is that AI involvement actually undermines that exclusivity premium, making human-made things feel scarcer and more valuable by contrast. His conclusion is that economics does not become irrelevant in an AI world. Scarcity just relocates. He opens with this Starbucks example, which I think is such a good gut check on all the automation assumptions we make, especially with AI. But Starbucks, if anyone remembers, they had technology that essentially removed a large portion of the human element entirely, and then they tried it out and they actually reversed a lot of those changes because They found out the handwritten cups, the interactions with the baristas and ceramic mugs, those things drove more people to sit and stay in their cafes. Charles, do you buy the logic at scale that hospitality or so much of this intangible interaction becomes a genuine competitive moat as opposed to a soft differentiator, which I feel like a lot of people are overlooking today?

I think we already have examples of this in life. Getting a human customer service person for most businesses and services that are scaled is a VIP experience. For everybody else, you get either the automated phone tree or you get the AI chatbot. And I do think because human labor is expensive, people will say, ooh, if that's done by a human, it's bespoke. Remember we talked about, was it your brother who's the glassblower?

Yeah. Mm-hmm.

Like, yeah, your brother's onto something. We are gonna see a resurgence. And I think we already see this. Like, you think about like watches or there's already a lot of handbags. There's a lot of luxury goods already where there's a premium on handmade by humans at low volume. With Craig here. I think that will accelerate. And I do think, I consider myself an economist, as someone who studied econ in school. I do think one of the things I've been trying to figure out, if you think about like a lot of these conversations, particularly around AI businesses and services, most defensibility, it's, I think it's all a question about what's scarce. Because historically in software, computer engineering development cycles were scarce. So if you could accumulate a lot of them and turn their energy into product, it was harder to compete with that. And now computer engineering cycles are less scarce. I wouldn't say cheap. I wouldn't say free. They're less scarce. And I think everyone's trying to figure out, well, what is scarce? I don't think we, we know. So people have gravitated towards taste or energy, these fuzzier things. But I do think we will see a resurgence. It's funny. I went to two coffee shops in Palo Alto. This is obviously a very Bay Area thing. They're both on University Avenue. One coffee shop has one person working behind the bar and a machine makes almost all of the drinks. Very capital efficient. I think I said 2 words to the person in the coffee shop. Coffee shop was pretty full. The people were all working on laptops. They weren't really talking to each other. It felt like a library with a coffee shop in it. The other one I went to is a really old Italian coffee shop on the other end of University Avenue that has no takeaway. They make all of it on like traditional espresso machines. And you go there and everybody's sitting there and talking. And the coffee is made by the people behind the bar. Many of them have been working there for a long time. It's not the fastest coffee shop in the world, but I think it's the most pleasant. And I think the people who go there go there for the experience. I don't know how you invest against that theme, but not everything that happens in society needs to be a moneymaking. So I do think we're going to see what he described. And because human things are expensive and hard to procure, I think people are going to value them even more than they do today.

I think in your example, one thing that's true is both can exist at the same time. For my own coffee shop, it's right across the street. There's only one. They had 5 locations and they sized down during COVID It's very neighborhood feel. There's always a line. It's always crowded. You don't get your coffee immediately, but everyone draws there. And there's a Starbucks a block away and it's way less crowded mostly. And it's just a different group of people who go there. I was talking to the manager and their, their Square tablet had broke. And so they really could only take cash. And there were still lines, like people were still lines waiting. And so that was in their world, a defensible moat that worked for them. People still showed up and a lot of that had to do with the human element of it. The more rational sector piece that Amos talks about of just the extra intangible aspects. That's a pretty spot-on comparison, I feel like, or example that we're making because it's talking directly about hospitality. I guess one thing that got me thinking when I was reading this article is the rational sector, which he kind of argues is going to grow or more people might enter that space. He gives the examples of service jobs. So think in hospitality, nurses, teachers, et cetera. Those people have been historically underpaid for all times. And so I'm wondering, does something like this cause those people to actually be paid correctly, or does it just have more people enter the space because their jobs that they previously worked in are no longer in high demand? We had COVID happen, say that created demand for teachers and learning experiences that we hadn't seen prior, but that didn't really change the salaries that they were paid. So I like the framing of his, his piece because it is optimistic in that things relocate, which is great, but the way he talks about the rational sector, I'm just not so sure if those people will get the premiums that they've always deserved, I guess.

I would love to believe that they will. I'm skeptical that they will, but I do think people who are at the very high end of those services are going to make more money. And I think, you know, you are going to end up with the star economy where I think the people at the very top in some of these other professions are going to find that they can command more. If you're a really great tutor, there might be wealthy families who say, you know, as great as AI tutors are for my kids, I'd I really want a human, but I want like the best human I could find, and my willingness to pay for the best human will be like much higher. In the same way that if you are in hospitality and you work at a really good hotel or a really good restaurant, you can make more money. The standards are higher, the expectations are higher, you have to be better at your job, but you can make more money. But that's more money within relative to that profession. I, I'm very curious to see how this all plays out.

It is interesting because I think the piece does focus also on elitist— is a strong word— but higher income power to drive where the relational sector is going to have the most demand. That feels a little bit ironically dystopian when you think about the article and the takeaway being so optimistic. And then I think, oh, all the power and attention is truly being driven by the people who have the money to kind of position it in that direction. And the question that remains is who will have the money at the end? I think there's this undercurrent conversation that it will be the plumbers, the electricians who've been put behind, and now they will be at the forefront. But I don't, I don't know. I think people who can continue to go to the high luxury resorts will continue to go to those resorts and will pay even more premiums to go to the best ones.

I immediately thought of White Lotus when I read his article.

Yeah, that's what I was thinking of.

I think the other thing that I've been thinking about, which is related to what you said, is if AI is as deflationary as we think it is for digital services, that means those things become cheaper. But if you're already sitting on quite a bit of wealth, you're going to find yourself paying a lot less for stuff that used to be expensive, which means maybe your willingness to do weird things with your money that are esoteric could go way up. You might just be like, well, you know, I don't need to spend as much money on XYZ thing anymore because of AI. What are other passion things I could do with my money? Things I'm interested in. Hey, I'm going to have a Somali come to my house and hang out with me for a week and like teach me stuff. Like new ways to monetize expertise or human experience are going to show up. Do you think they're going to be mostly accessible to people who made money in the pre-IA era who are sitting on it and suddenly are like, AI is making all of the other things in my life cheaper and easier, so I have more money and time for leisure? Like they will be the Keynesian beneficiaries of all this stuff. I do love the way that you worded it.

Earlier, the star economy. This is the time to lean into, if you had a hobby that you're good at, if you wanted to perfect it, it would be the time. Our next piece is by Lisa Kostova titled "40 Buyers and Nothing to Sell," describing an honest take on vibe coding and what happens when you follow the hype into production. Lisa spent 9 months trying to build an AI-powered content system for financial advisors. She showed up to a major industry conference in Huntington Beach with 40 people ready to buy exactly what the title says. And nothing to sell them because the product was structurally broken in ways she could not see or diagnose on her own. The piece is organized around this metaphor of footprints on a glacier in New Zealand. Hikers were essentially following tracks, hopefully looking like they were going somewhere, not realizing that the tracks they were following were actually someone's practice laps on the same path. And the technical failure was really stark in her product she was creating. Her AI coding assistant had quietly created 3 separate services to handle the same API call, never flagged it, and when things did break, it confidently blamed other tools rather than its own output. What actually fixed things was, surprise, surprise, a Bulgarian engineer who looked at her code the way a plumber analyzes a wall leak and declared that it was a total mess, and that if he had received this code from an actual software engineer, he was like, I would kick this back to them and like, fix your crappy code. The real lesson she lands on is that the map everyone is following in vibe coding is upside down. Prompting feels like progress, things we've all talked about before, but engineering an actual system that works in production does require a fundamentally different orientation. I read this piece and I was thinking that Lisa is one of the few people willing to publicly document what actually went wrong when vibe coding meets production in reality, even when there's a lot of excitement and demand coming her way. But for every Lisa, I think there's probably like 100 founders or operators who don't realize that their product is structurally broken until a customer does, which is far more dire. How long, Charles, do you think it will take before the market starts separating the products that were genuinely engineered from the ones that just look like they are on the surface?

My current thinking is hype moves 3 times as fast as reality, which means it's going to take a while because everybody's AI psychosis. AI-pilled CEO believes that these things can be vibe-coded into production, and it's going to take some failure before people say, well, wait a minute, maybe this isn't such a good idea after all. And maybe we overestimated what we could do and we need to go back and take a more fundamental look. And I hope you didn't fire all of your software engineers. And I think some of the soft people will realize, you know what, maybe vibe coding an internal wiki or a software tool that's used by 2 people that we would never pay to have built, maybe that's an okay use. Maybe our core production service that we use to make money that has some complexity shouldn't be vibe coded by someone who's not a developer. Whenever I talk to my friends who are actually trained software developers, the way that they use coding assistants is different. You know, the coding assistants help them write code, but they're still making judgments or they're giving the coding assistant far more instructions on architecture and design. And they're not just letting the coding assistant make a bunch of really important architecture decisions on its own. And I think that's why you get better code versus, I think, using coding assistant when you don't understand the fundamentals of software development or architecture. It will build you something. It will probably work, but that doesn't mean it's maintainable. I was just doing some math. It's about 10 to 1 for me. Support, tinker, upgrade, debugging to build. A lot of my friends are like, for someone with your ability, that's probably right. They're like, if you were better, you'd make better decisions upfront and you'd plan a little better and it wouldn't be 10 to 1, it'd maybe be 3 to 1. But anybody who tells me like, oh, I just like one-shotted this thing and I've never touched it, I'm like, it's probably broken. You don't even know it. So I have a real appreciation for the craft of software development and like what it takes to fix things when they break and how you diagnose the problem and how you figure out what about it you want to fix. And so I think this is the first of many of these stories we'll hear, and they're going to get a lot of attention, just like the, the robots that trashed those people's houses in a desire to get cleaning robot training data. There, there's a deep audience for like AI mishap stories.

Oh, absolutely. I like the metaphor of the hikers and the glacier because it showed that like we're kind of in unexplored territory, not for software engineers. They've been doing this job forever, but for everyone else who's like jumping into this role and upskilling themselves. She had two really good graphics. One was kind of like really invariable looking terrain. Like if you were looking across a valley and in different pockets, she outlined over here, it works well. Like you're up on the top of the hill over here. It doesn't work as well, but I'm not quite sure over here. It's like a deep valley into a pit. It definitely doesn't work, but you're walking the same plane. It could be all one product that you're working in and in different pockets, it's working in different times. It's working previously, at least for non-technical people, it's quite binary. I'm building something and it works or it doesn't. And the customer sees that too. It's like they experience what it is and then it's working or it's not working. I think what's scary is you could be getting an output that just looks like it's correct because it came in a pretty package. It came at the right times, like it could be spat out in 5 minutes and you don't know if it's incorrect. And I think that's where it just takes a little bit more of a technical eye. And my question is, it does seem like people do need to understand the difference between when to use the LLM and when to use more deterministic script. That muscle, I feel like, does derive from having experience of being technical. Do you think that that skill can be taught broadly, or does it just naturally select for people who already have an engineering foundation? And how could a non-technical person upskill themselves to get there? I assume the question is you just continue to have to build and actually take your time and be thoughtful.

So funny. I have a presentation for our team tomorrow, which is about the 20 agents that I have in production and like what they actually do, what they've done for me. And in building the presentation, Cloud Code was like, oh, this one's been failing for a week. I was like, what do you mean? It's like, yeah, it just stopped working. It's not a critical one or I would've noticed it. And so my project this week was I went back and put in much better logging and monitoring into all of my agents and skills so that I get regular updates. Software development, you would put in logging and monitoring. It is really interesting how I feel like every week and I'm like, oh, I have a greater appreciation for this element of software. I understand this concept now in a much more visceral way because it's something that I want. It's a problem I have to solve. And I think that's the thing I'm learning is that most of the things you need to know, it's far easier to learn them when you encounter them and bump your toe.

But how funny is that example that you could, you would share that presentation to us tomorrow and even if things weren't working, we'd be like, oh, how incredible, amazing. Can you send me over your script so that I can copy it and start running it myself for my own workflows?

Like Claudia was like, do you want to include the ones that aren't working in it? I'm like, absolutely. Absolutely. Include the broken ones because I wanna talk about the broken ones and what I'm doing to fix them. Cuz I think that's like part of the lesson too.

Yeah. And I think that's the growth muscle that we're going through collectively is there's this instantaneous satisfaction you get when you build something and it works and you're like, I am amazing. And then when you have to do maintenance on it, it's just so much less fun because the output is the same. It's I already created this 2 weeks ago and now I have to dedicate specific hours to continue to make it work. It's so much of a less fun task, but again, that's where I guess real craft taste judgment is built in those hours, as opposed to just spitting out something that works in the first go. Our last piece is a really good tie to what we just talked about. Now it's an interview with Casey Newton and Boris Cherny, the creator of Cloud Code at Anthropic, and it reflects Boris's take on the software engineering job market from someone who's actively automating his own role. Boris hasn't written a single line of code himself in over 6 months, and he's really clear-eyed that the title of software engineer may start dissolving by the end of this year into something closer to builder, as designers, PMs, and managers all start shipping code of their own. But his overall forecast is quite optimistic. He thinks that there will be 100 times more people writing or directing code in the near future than there are engineers today, just not people with that specific title. He uses the tractor analogy to ground the rate of change for everyone who feels uneasy, pointing out that it took 70 years from the tractor's invention to tractors outnumbering horses on American farms and draws a parallel to how companies that reorganized entirely around computers in the '90s, so completely threw out their files, threw out all their paper and just decided we're gonna focus on this only, saw massive productivity gains while others who were half in, half out saw none or much lower. His closing point that stuck out was about access, the winners of some recent Cloud Code hackathons that they ran weren't senior engineers, but it was like an electrician, it was a doctor, it was a carpenter. That anecdote really landed for me because it does flip the assumption that AI tools are a gift to people who are already only so close to the space. I think we spend a lot of time thinking about Cloud Code is amazing for the best engineer because it makes them 10 times more productive. I would have loved to have seen what the electrician, the doctor, and the carpenter kind of came up with, because I would assume that what they created was actually probably closer search for something that is actually needed because they're kind of building off of their own experience than somebody who's like, amazing, I have these tools, I can be productive. Let me just start creating things for the sake of creating. I was wondering if that was surprising to you, the hackathon anecdote.

I'd read a write-up they did on the hackathon and it was surprising, but I think for me, I thought a lot about Excel and I think it used to be a long time ago that like computation was expensive and computation, we had little people called computers who did computation for And then Excel came out and everyone was like, ooh, can an investment banker use Excel better than the average person? Absolutely they can. Can a consultant use PowerPoint better than the average person? Of course they can. The average white collar worker assumes that they will have to use a PowerPoint or PowerPoint-like product, whether it's Google Slides or PowerPoint or whatever, or, and some kind of app-based computation product like Excel or Numbers or whatever as part of your job. Those are just skills now that you must have in order to be effective as a white collar worker. But it doesn't require that you have the design skills. Of a designer and attention to detail of a consultant or the familiarity with macros and shortcuts of a banker. You can use your mouse. It's okay. You're expected to have some base level familiarity and facility with those tools or you probably won't get hired. And I think the same is going to happen. I think I'm an example of what Boris is describing. The code I've written in my life, I've basically written in the last 9 months. Other than the code I wrote learning the basics of Python, 95% of the code I've written, I've written with the coding It's the only way I know how to really ship code, but I would never call myself a software developer in the same way that business schools started teaching people Python and R for data analysis. I don't think most of those people would say, oh, I'm a data analyst. They're like, oh, I'm an MBA student who's like doing analytical work. And I think that's where we're headed. There will be, I think, people who are professional software developers because of everybody shipping code with variables in the same way that you would never let an analyst ship their final version of a model to a client without people who know what the model's supposed to do. Wouldn't like reviewing it. I think this notion that everybody's going to ship to production with no oversight, it strikes me as like very dangerous. So I do think there will be people who are very skilled at software development, who understand architecture, who understand design, who are going to look at what people ship and say, mm, we're not committing this, or this is pretty close, this needs some massaging and fixing. And it is going to be a little bit more craftsman to make sure that like the body stays coherent. They're going to tinker with these things as they come in. But many more people are gonna be contributing to software products, and it's just gonna become another skill you have to have, just like Excel and PowerPoint.

The one thing that I was noodling on, which was this tractor example, which I think he was using as a way to calm people's fears that things are moving too quickly and, and, and in saying that it took many, many years before tractors were everywhere and people had really started relocating from farming to the back house and in, in the factories themselves. I think that one works for the tractor example back then because it was the high barriers to entry to get a tractor is expensive. It was difficult to get ahold of them. You really couldn't start the tractor unless you knew how to work it. This feels different because there are low to no barriers to entry to start getting smart on code, and you can just go running with no oversight in one direction. So that's where I feel like I think the dissemination and like training of it is, it's gonna happen a lot quicker. Obviously, I think like half a year ago, I don't think most people didn't know what Claude was outside of the tech bubble. Now they do, which shows you that they've done a great job in terms of distribution. I do think that what you said, how it goes through the funnel all the way to production, that's going to be what's more similar to the tractor analogy than just the dissemination of who has access to the tractor or to Claude code.

Also, with 70 years, you can be like, I'm just going to ride this out. That's 3 generations. You could have been a horse farmer. You're like, I see the tractor coming. I don't want change. My child or the person who takes over my farm, I'm going to let them deal with the tractor. I'm going to keep farming the way. And I think this is why people rightly have more anxiety over this. You know, IBM has been talking about the paperless office for 50 years. We're only kind of now just there, but this idea of a paperless office has been around for a long time. And ironically, the introduction of computers is for some people led to more paper because people started printing stuff. So instead of having one copy, now you have like 10 copies or you print the same thing. So I think previous tech waves, you could see it coming and you're like, I might just get out the door before I have to deal with this. This is not that. This is here right now. It's funny, the presentation I made for our team, I basically did a Claude Code session and created a Markdown file with all of the outputs I wanted from my agent development work. I gave that Markdown file to Claude Design, and that was the presentation. I didn't touch PowerPoint. It was like Claude Code to Claude Design, and then some tweaking and polish. Is it a presentation I would ship to our LPs? No. Is it good enough for an internal discussion? Absolutely. I think the reason people are scared is like, if you told them you have 70 years to learn this stuff, you would be like, okay, I have some, I have some anxiety, but I've anxiety for my children and grandchildren. Be fine. I'm gonna like retire without having had to learn this. That's not what's happening here.

That's the scary part. I think it's also the people, I mean, mostly millennials, I think feel like we grew up in a time that was able to grasp the tech wave and figured all of that out. And then right when you're hitting middle management or management your own, you're having to relearn everything and then bestow that knowledge down to a younger generation who doesn't have any of that context. And I think that's where I feel the most anxiety is in within that generation specifically.

For Sure.

Well, we are all out of time. Thank you, Charles Hudson, for stopping in. Thank you to our listeners for jumping in as well. Please like and subscribe, and we will talk with you all soon. Thank you. Bye.