The Exchanges

Q So speaking of test driven development, a lot of the questions were about the various engineering methodologies. Agile development, lean startup, test driven development, extreme programming. How do you guys think about those? Do you, do you adhere to any particular methodology in your company?

A Uh, I would describe us as agile-ish. Um, so we do not fully adopt any of those methodologies. Um, basically we find what works well for our team and aren't super dogmatic about, you know, is this agile or is this not? You know, we run sprints, we have daily stand-up, so we have some elements, um, incorporated in our development. Um, but I think kind of with anything, you know, A lot of these problems, you're going to be solving a lot of problems for your organization, and you have to kind of find what works best for you. Um, and, uh, so I think it makes sense to take bits and pieces that work for you, um, and adapt it to your organization. Um, and then I guess it's just the one other thing I'd mention on this is, um, as you grow, uh, a big, a big challenge is just constantly Evolving how you work, right? So it's very different to work on a small team of four, um, where everyone kind of knows everything, um, and everyone, everything is in everyone's heads versus a team of 30 or 50 or for some of us, you know, hundreds, um, that is very different. So, uh, you kind of constantly need to be assessing is this style of working or this methodology, um, is this right for this stage of the company?

AI assessment note: “I would describe us as agile-ish. Um, so we do not fully adopt any”

Q And yeah, what's the experience going to be like for an attendee? What can they expect?

A So I think the most, the most compelling reason to go as an attendee is if you think you might be interested in working at a startup, but you're not really sure which one or how to think about it. If you come to the work at a startup event, you'll get in a short period of time, like a really compressed view of what's out there on the market, what your options are as an engineer right now in, in, in, in, in, and you'll get to talk to the founders because after After the talks, we're going to have just like a mingling session where you can just walk up to the founders of any company and head to know them.

AI assessment note: “after the talks, we're going to have just like a mingling session”

Q I was not convinced that this is, that this is going to work given the current strategy. Um, you're not, you shouldn't say, I think, You're a bad founder. Would not say that. Um, but I think there's ways that you can convey that you just were not convinced in an honest way. Because that's basic, if the team cannot convince you, I mean, how do you say this, Jeff?

A Yeah, I think this is, that's actually a great, and one of the hardest questions. We actually struggle this with YC all the time, because sometimes we interview people, and we think it's a great idea, but we just don't believe in the founders. And it's the last thing you want to tell someone. One, because you might be wrong, and you know, you shouldn't stab someone in the heart. Two, because it's not a good interaction, right? And let's just keep in mind though, there are always a million reasons not to invest in a company. Most companies will fail, and it's a lack of your own imagination if you can't come up with a reason why you're not Believing in their version of the future. And that's what you should, that's what you should communicate to them. I, I think the, the bottom line is, every interaction with the founder should be as helpful to the founder as it possibly can be. And if you can't be helpful, then just don't say it. But if you can, if you can find some way to help them, even if the real reason is not that, so much that, but it's still valid, I would use that. There's always something.

AI assessment note: “find some way to help them, even if the real reason is not that”

Q Historically, it wasn't this the government that funded all this stuff?

A That's right. So, so one of the, and this is, you know, part of why I'm building protocol labs, which is that, um, there's this massive gap between Um, there's this huge open area where, like, stuff is not getting funded around, yeah, building large-scale infrastructure. Like, you couldn't, my, my claim is you couldn't build something as free and open, and that works as well as the internet today, um, because no group would fund it. Um, and, and what you would end up with is a, a massively stunted version of something that is highly centralized and controlled by a couple, by a couple of groups, um, and that wouldn't have the, the amazing generality of something like TCP IP. Like, Part, part of what's beautiful about TCP, IP, DNS, like that whole era of protocols was that people worked super hard for months and years at a time to think about the interfaces and refine it so that you could end up with something sufficiently abstract to, to support a ton of use cases and sufficiently concrete to actually work today. And that kind of development is not super fast and takes a lot of work and takes a lot of money. And that's not something that, you know, VC funds. VC funds, clear application.

AI assessment note: “That's right. So, so one of the, and this is”

Q now. Um, what's going to happen to all the programming jobs? Like, it used to be the case that if you were a CS major, there is a very clear path to, like, a very stable, like, upper middle class background where you get, like, a good stable job as a, as a programmer. Um, but, like, Are those jobs still going to be here in 10 years? Like, yeah.

A Yeah, like my, my parents were really proud when I, uh, you know, graduating, I, you know, got my degree, and then I got my job at Microsoft, and I was a level 59 PM, uh, you know, lowest of the low, but I had health insurance, and my parents were really, really proud of me. And, you know, one of the fears, frankly, like, that we're hearing, uh, and it's sort of, you know, Coming out in the numbers is that will there actually be jobs? I think it's a tricky thing right now with the advent of intelligence, you know, some of the simplest things that people rely on entry level people right out of college for, uh, they're not hiring as many of them anymore. And, you know, the craziest stat, I think this came out of, uh, uh, the New York fed in February of this year, um, computer science majors, uh, You know, obviously this is not the people in this room. This is just, like, out of, like, you know, a normal distribution of all computer science majors. 6.1% in unemployment in February of this year. Art history, in contrast, was only three point oh percent.

AI assessment note: “6.1% in unemployment in February of this year. Art history, in contrast, was only three point oh percent.”

Q So fast forward a little bit. I mean, at the start of COVID delivery demand cratered and then skyrocketed and you cut commissions in half. You ran a TV campaign that advertised your competitors. What gave you Conviction about that. What, you know, what was that period like?

A COVID was a bit of a blur. COVID, 2020 was a bit of a blur. Um, we were actually in 2019 preparing, you know, to go public. Obviously, COVID, um, shelved those plans. But, but, you know, COVID was probably next to 20 13. So the, the, the COVID year, 2020, that is. Next to 20 13 was probably the year where it felt most like DoorDash NYC. You know, it was seven days a week, 10 a.m. to two a.m., all hands on deck, you know, multiple, you know, all company, you know, meetings per day. And sometimes in crises, I actually find that it's a lot easier operating a company because it's very clear what to do. Number one, job number one, keep everyone safe, right? Get tens of millions of units of PPE. Make sure that we can ship no contact delivery or contactless delivery. We ship that product in four or five days. Number two, Got to make sure that everybody gets liquid. Why? Because the average merchant has 17 days of cash on hand. So every hour of cash is very, very important. Same thing for dashers, um, who, who, um, a lot of them are furloughed or laid off, um, because of COVID, and, and so getting them instant liquidity. The third thing was making sure that we could take care of the community, actually, and so we partnered with Dozens of the largest hospital networks, um, from UCSF or Stanford here in, in California to Mount Sinai on the east coast where we wanted to make sure that all…

AI assessment note: “in crises, I actually find that it's a lot easier operating a company”

Q we can't make it to the 50% of the folks who are at HBS, um, know how to code, can we create a class where we grab people who do know how to code from elsewhere in Harvard? And that's like a required, um, ratio. Cause I, I do think that, man, it's really hard to meet people outside of class where you live and like the normal school activities.

A I, I think it's a great suggestion and there's definitely, A lot of that percolating at Harvard and at MIT, by the way. MIT does, I think, a better job of integrating technical people into the MBA program. What's happening at Harvard is the School of Engineering and Applied Sciences is physically moving across the river, across the Charles River to be co-located with the business school, and that's a multi-year project that's Been funded by folks like Steve Ballmer and John Paulson and others. A lot more social engagement as that opens up in 2019 and 20 20. In fact, I'm already seeing classes that hit exactly being considered by the business school that hit exactly what you're saying. I think He's taking classes at the MIT Media Lab. I love it because it shows that they're mingling with the engineers and with the visionaries and with the futurists, and I think we all know great startups and great ideas come from cross-pollination, and that's why the silo thing is so heartbreaking to me because we have so many Narrow lanes, and I don't know what you see geographically with the YC group, but I think if you're not in cities where you're seeing a lot of cross-pollination, It's a real, uh, disadvantage for those entrepreneurs.

AI assessment note: “I'm already seeing classes that hit exactly being considered by the business school”

Q Um, so you mentioned that you should, uh, try to charge your early user. But if you, if even your final product, you're planning to make it a free app and monetize with some premium content, uh, what should you do?

A So the question is, put it more generally, um, should you be going free if your final idea for your product is to be free? What I would say is this. If your users are users who you never plan to charge, then it's totally fine for you to be free, but if you do plan to charge them in some way, it's really helpful to charge them as soon as possible, because you want to know whether or not they're willing to pay, and certainly if their business depends on it, It's especially helpful to charge them. So that's the measure that I would use, and there are all kinds of little tweaks and, and so on and so forth, but at a high level, do you ever plan to charge them? I charge them. If you never plan to charge them, you plan to monetize based on ads, which is really usually the way that you never plan to charge them, is you can monetize with ads. If you're not going to monetize with ads, you probably should start charging them. Alright, next question.

AI assessment note: “if you do plan to charge them in some way, it's really helpful”

Q Could we maybe talk a little bit about, um, like the humans behind the robots here? Like, um, how did the company get started? Like who are the, who are your co-founders? How do you all get together and what skills do you each bring to such a complex problem?

A Sometimes the joke I make here is that the human behind the robots are also robots. Not really. Um, yeah, so Pi is a very, I would say, entraditional company. We have a, like, larger than average founding teams, and some of us worked really closely together when we were at the robotic team at Google, and the robotics team at Google was, I think, a really, really great environment for seeing the sign of life and creating the relationships and the community. That allow the robot community and like these advances to flourish. There is Locky, uh, which we met when we, uh, were thinking about starting the company and has just been really instrumental in making sure that we're a good business. And there is Adnan, our hardware lead, um, that came over from Android. And Adnan has a really difficult job because if you want to work on cross embodiment, you know, remember my, uh, joke about how if you want to add two years to your grad school, Bring on one more robots. The, the hardware problem and the operational problem for us is how do we build, improve, and scale a fleet of Heather Joe genius robot. You know, it's just not one robot platform. And because we built the organization from scratch in the beginning to, to, to support that, like, I think we're able to do it, but it's just a really hard, uh, problem. Um, because there's just like No two different robots in the fleet. How do y…

AI assessment note: “There is Locky... And there is Adnan, our hardware lead”

Q Oh, you could. You totally could. What if it's OpenClaw and, um, Obsidian and Markdown files and like, you know, a brain.md with like ontology that's custom to your use case, and what if it's a hundred OpenClaws in the background that you orchestrate?

A I think there's two sides to this. The first is that we already see a little bit of a side of life, where for simple failure modes, um, during evaluation, if you can describe the way that the robot failed in text very precisely and very clearly, Then, you know, you can ask a language model to make very reasonable recommendation about what the next step is. Um, but the, the, the flip side is that this only works for simple cases today, and the reason why that's the case is because I think it's pretty, um, fundamental limitation of the model that we have today, which is that they are not at the core model that take action in the world and see the consequences of its own action, especially action that changes the physical world. Um, and, and so I, I think this kind of very fundamental understanding about how the physical world works is missing from the really large foundation model. Um, and, and I think that that's one of the ingredients that's missing to, to be able to build this automated robot research scientist.

AI assessment note: “this only works for simple cases today, and the reason why that's the case”

Q out. I guess it becomes more nebulous when you go a couple degrees off where there are fields that are not naturally formally verified and you need to come up with a, again, with some sort of a function. To come up with that reward that makes it verifiable with very fuzzy things like, let's say English language and composing the perfect essay. How do you make that formally verifiable?

A Yeah, yeah, absolutely. I mean, writing SS is, you know, the typical example of a domain that's not verifiable. And so what you're going to see is that progress of reasoning models and base LLMs on this type of, of, of domain is, is, you know, it's going to be very slow because the stack we're using, like the LLM stack is very, very reliant on its trained data. It's basically just operationalizing the trained data. And for writing SS, the trained data is coming from Uh, human experts, like annotating, uh, answers, and that's costly. So you're going to see this very, very slow progress. Maybe, maybe it's even going to stall. But for any, any very favorable domain, like take code for instance, which was the big unlock is, uh, when, uh, when people started creating this code-based like training environment, uh, for, for post-training. Uh, where the, the, the reward signal, the verification signal is provided by things like, uh, unit tests and so on. And so that means that, uh, the model was not just working from human provider annotations. It was actually trying some things, uh, verifying the answer and, uh, and generating a lot, lot more string data in the process, a much denser coverage of the problem space. And not just coverage in terms of like, is, is the answer right or wrong? But also starting to build models of the execution traces, right, so that the models could start in…

AI assessment note: “writing SS is, you know, the typical example of a domain that's not verifiable.”

Q Yeah, what do you do to be so productive with it?

A One is if you're able to use, uh, just generally far less code and plumbing, Um, so a lot of what I do is like deploy stacks on like Vercel or Next.js or like Cloudflare workers, where there's kind of like already a bunch of boilerplate, like taking care of for you. And then you don't really have to think that much about like, Hey, I need to stand up like all these different services and deal with like service discovery and like registering on like some sort of central endpoint or like all these databases. It's like, Oh, like everything is pretty roughly defined in this, like one or 200 lines of code. I, Tend to operate more towards microservices for that as well, or like individual packages that are fairly well structured. I think it's also worth knowing like what the LLM superpowers are. Like in general, coding agents are, I think Andrej Kharapathy just tweeted about this. They're like super persistent, so they will keep going no matter what. They end up, uh, typically just making more of whatever's there. So if you're trying to direct them to do something, it's worth like I mean, I can pick on OpenAI slightly in this example. OpenAI has, like, a giant monorepo. It's been there for a few years now and has, like, I don't know, thousands of engineers who are committing. Some of those engineers are like super senior meta folks who came in and are like, know exactly how to write …

AI assessment note: “One is if you're able to use, uh, just generally far less code and plumbing”

Q Do you have thoughts, just because you know the, the agent's so wild, like, what, what types of engineers are going to benefit more than others, um, from these tools becoming popular?

A In general, I think that kind of the more senior, senior you are, the more you benefit, um, because the agents are so good at taking some sort of idea and then putting it into action. If you're able to prompt that in a few words, it's kind of like, oh, now suddenly I had this like idea. I find this so often open AI, like strolling through the code base. It's like, oh, like, here's the thing that I wish were different. Here's the thing that I wish were different. Here's the thing that I wish were different. Like just being able to kick those off and then have them come back, I think is Super empowering and multiplies your impact. I think also being able to detect like which sorts of changes are good or bad architecturally is very important or like have a sense for where you might want to flag something to an agent. I think engineers who are more organized, like manager ish, uh, and there's probably just a missing product to be built here. Uh, maybe something like conductor, uh, where it's like spread across all of your sessions and kind of reminding you like, Hey, you were working on this thing. It's done. It needs your input here. Oh, you should switch your attention over to this other thing. I think that is.

AI assessment note: “the more senior, senior you are, the more you benefit”

Q So if you were going back to your, like, college days and studying CS again fresh and you, like, were picking your own, like, syllabus or curriculum, like, what would you, what would you study?

A Personally, I think still understanding systems, uh, is very important, um, and just having some conception of, like, how, like, Git works, you know, or, like, HTTP, or databases, like, queues, like, all of these different systems. I think that those fundamentals are still quite important. The other thing that I'd probably do is just have a semester where like each week you're just building something and you really try and push the models as far as they can go. There's a sense that you have whenever you're doing something that you could always just like go up the layer and ask the model to do it and like go up a layer and ask the model to do it. You know, it's like, oh, I have like a implement command where it like implements the next phase of the plan, but then I could have like an implement all command and it like goes stage by stage and creates a new sub agent. And then I could have like a check your work kind of thing and like. And I think knowing where the models can and can't accomplish that is such a moving target that it's worthwhile just to, like, tinker a lot.

AI assessment note: “Personally, I think still understanding systems, uh, is very important, um, and just having”

Q So the, the common knock that you hear on this from people who don't really know what they're talking about is like, oh, it's just consulting dressed up with fancy marketing speak. Why is that wrong?

A I think before I say, I don't want to tell you glibly why that's wrong, because I think there's actually a real risk that it's right, right? And I think, you know, if you, if you go back to 20 15 and you talk to people about Palantir, maybe you would hear two things. One, that Palantir is evil. Um, but the second thing you hear is that it's a consulting business that is never going to scale, you know, that it's actually like a bad business. It's not a software business. And we spent a lot of time trying to understand whether that was a correct Characterization or not. From a business model perspective, one of the key things that you will see, that you should see, is that it may be the case that you're, when you go into, you do a new deployment at a customer, that you're actually losing money early on. As the longer you're at the customer, first thing is your product, because of the product discovery, gets better suited to what the customer does. And so you no longer need a large team of people at the customer site figuring out what the customer is doing, you know, paving, you know, writing that. Code. The second thing is that you should be earning the right, as Sean would put it, to have access to more important problems at the customer site. And so you should see basically that your cost per value of the outcome you're delivering is going down. And so your profit margins start…

AI assessment note: “your product, because of the product discovery, gets better suited”

Q Are there specific areas that you think a lot more builders could go into and build with these new models? I mean, there's a lot that has been done, let's say, for coding tasks, but what are some tasks that have a lot more greenfield that are just getting unlocked right now with the current models?

A I come from a research background rather than, uh, rather than business, so I don't, I don't know that I have anything very, uh, very deep to say, but I think that, like, in general, any place where, um, it requires a lot of skill, um, and it's a task that mostly involves sort of sitting in front of a computer, interacting with data, I think finance, uh, people who use Excel spreadsheets a lot, um, I think I, I expect law, although maybe, maybe, maybe law, ah, is, is, is more regulated, requires more, ah, more, more expertise, um, as a stamp of approval, but I think all of these areas are probably green field. I think another that, that I sort of mentioned is, how do we integrate AI into existing businesses? I think that, like, when electricity came along, there was some long adoption cycle, and, The very first, simplest ways of, say, using electricity weren't necessarily the best. You wanted to not just replace a steam engine with an electric motor, you wanted to sort of remake the way that factories work. And I think that probably leveraging AI to integrate AI into parts of the economy, um, as quickly as possible, I expect there's just a lot of, a lot of leverage there.

AI assessment note: “I think finance, uh, people who use Excel spreadsheets a lot, um, I think”

Q Now, one aspect about, uh, scaling laws, they've held for over five orders of magnitude, which is wild. This is a bit of a contrarian question, but what empirical sign would convince you that the curve are changing? That maybe we're getting off the curve.

A I think it's a really, I think it's a really hard question, right? Because I mostly use scaling laws to diagnose whether AI training is broken or not. So I think that, uh, Once you see something and you find it very, it's a very compelling trend, it becomes very, very interesting to examine where it's failing. But I think that my first inclination is to think if scaling laws are failing, it's because we've screwed up AI training in some way. Maybe we got, ah, we got the architecture of the neural network wrong, or there's some bottleneck in training that we don't see, or there's some problem with Precision and the algorithms that we're using. So I think it would take a lot to convince me at least that scaling was really no longer working at the level of the sort of these empirical laws because so many times in my experience of the last five years when it seemed like scaling was broken, it was because we were doing it wrong.

AI assessment note: “it would take a lot to convince me at least that scaling was really no longer working”

Q You were at the retreat we had, um, where the founder mode, it was not called that at the time, it was just Brian Chesky talking, PG turned, uh, he coined the phrase founder mode, but you were there, and again, you worked at Airbnb in the years where they were growing so fast. Do you have any thoughts on the whole thing?

A It absolutely resonated with me, uh, and I also, at the same time, I love Airbnb. Like, I could not be more happy, uh, that, uh, you know, that, like, there's so much change, uh, that was executed and delivered within, within the, you know, within the company, uh, but certainly I can totally see how, um, You know, there was, there was like, there was a need for change, right? Uh, and like a need for like creating conflict within the company to actually ask the hard questions and say, Hey, do we need these folks or are, or, you know, are we, are we actually doing this project because somebody actually needs to execute this project so that they can get promoted versus like the comp, this is the right thing for the company.

AI assessment note: “It absolutely resonated with me”

Q one is like how you kind of dig yourself out of The whole, let's say you've launched your app that you've been out there for a little bit of time and you still just don't have users or growth. Like, how do you know when it's time to keep tweaking and how do you know when it's time to just like call it a day and pivot into something else?

A I think that's a good question. It's one of these things that I think is more art than science. The way that I would prompt a founder that was in that situation is something isn't working, right? You're out there, you've been trying, you're pulling every lever you can think of to pull. And just nobody's biting, nobody's using your product. Which one of your assumptions was wrong, right? And that like, it doesn't mean that this is a bad idea. It doesn't mean that you need to pivot to something else necessarily, but it does mean that you've made a bad assumption and you need to figure out what that is and develop a new hypothesis about the world.

AI assessment note: “It doesn't mean that you need to pivot to something else necessarily, but”

Q So you've launched, your product is out there, and nobody uses it. What do you do?

A Um, crying. Um, again, like the best founders just view everything, like we were talking about earlier, like they're learning their sponges, and so I think they just treat this as something, like a problem they need to solve and diagnose. You want to look at, like, where is, like, why are you not getting any users or customers? Where's the drop-off? I think you treat it like an analytical problem, like, is it because people aren't returning my, like, replying my emails, like, asking to give them a demo? Okay, like, Can I tweak the messaging? Is that what's going on? Am I targeting the wrong people? And then you just like week by week, you're like, okay, like you tweak one variable. Like, okay, I'm going to try a different messaging, copy my emails, see if I get more people booking demos. Um, and if that works, great. And if it doesn't, you're like, okay, like maybe I'm just targeting the wrong people. Maybe I'm trying to sell into like thousand person enterprise companies and I should be targeting smaller companies. Like I always find like, it's like tweaking those variables one by one is like how you kind of dig yourself out of The whole, let's say you've launched your app that you've been out there for a little bit of time and you still just don't have users or growth. Like, how do you know when it's time to keep tweaking and how do you know when it's time to just like call i…

AI assessment note: “treat this as something, like a problem they need to solve and diagnose.”

Q platform to help people learn software development. Okay. We've already launched our MVP. It's basically a publication text plus YouTube. We get about a thousand users on our channel monthly. Shall we focus on building new features and innovations, or should we go after currently Implementable revenue sources like subscription. Oh, okay. So should we start charging for things that we're already offering or should we build new free features?

A This is a tricky one because it really goes to like what your current situation is, AKA do you need money, right? What, how big of a startup are you looking to build? If you're looking to build a lifestyle business, then like I would definitely charge right now. Um, and kind of what your vision of what the product should be. Um, and so, like, this one's hard to answer just straight away. If you're thinking, oh, I don't need this to be a venture-scale business and I want it to be something I can do with all my time, so it needs to be able to make money so I can quit my job, then charge. Great. If you're thinking, like, oh, I'm looking to build a free product for teaching people to learn software development because I want to monetize it in this other way, Then I'm not so sure. Um, and then in terms of building new features, like, that's a question for your users. Like, what, what do your users want? What are they, how are they frustrated with your product today?

AI assessment note: “If you're looking to build a lifestyle business, then like I would definitely charge”

Q In consideration of getting an MVP up as soon as possible, what are the pros and cons of using white label services?

A So this is like a, this really brings up kind of a classic thing. So I would say, I would say a couple things to this point. One, if you're just trying to test any demand at all, and a white label service can, like, let you do that, great. But I'd argue that, like, if you want to test any demand at all for this type of product, probably, like, a Google Sheets spreadsheet could do it too. And so if you're just looking to test is there any demand, Like, go as simple as possible. If you have seen demand, and you're thinking to yourself, okay, I want to make this a company, if you want to make it a software company that can scale, that can raise money, it's a lot easier if you're building your own software. What's next?

AI assessment note: “if you want to make it a software company that can scale... building your own software”

Q deliver, then, like, we'll give you X or whatever, but it seems like a lot of companies would want to, like, plan specifically around, like, okay, if I need these organisms or whatever, like, do I have a plan that's definitely going to get me the organisms, or am I, like, Maybe if I haven't, that's great. That seems like a weird situation. So how do you get around that?

A Yeah, so I'm, I'll try to paraphrase the question. Um, it was like, it seems like getting an LOI would be hard, because the company doesn't know if you're actually going to be able to deliver it, and they have to, like, plan around that. Um, and the answer is yes. It is hard to get LOIs. Like, even though LOIs are not binding, they're actually pretty hard to get. Um, and, The weird thing is that the very fact that they're hard is it makes them valuable. If they were easy, they wouldn't be worth anything. So the, the reason that they're kind of valuable is that it is hard to get a company to do that. They'll typically only do it if you're solving a really critical pain point for them. If it's just like a nice to have, it's gonna be hard to get an LOI, um, which is actually really good signal for you to know that you're working on something that's a really big problem for them.

AI assessment note: “They'll typically only do it if you're solving a really critical pain point”

Q And so obviously like Niantic, you're working on games. Um, at what point do you think AR becomes this kind of like first class citizen in media, you know, alongside video, photo, I guess audio?

A Yeah, I think it will, um, it will take time, because we're at the early stages. I think the progress of, um, technology has, like, different cycles. I mean, if you read some of the work by, um, like, Carlota Perez, who's an economist, talking about innovation cycles, there's, like, big two stages. Stage number one is when technology gets installed. Quote, unquote. Uh, so installation types of technologies are when the applications are not ready to be built because you really need the tooling to be able to express those applications, and examples of types of installation types of technologies are just like network infrastructure, operating systems, uh, programming languages, all these levels of abstractions that are needed because you're not going to, let's say an example, today it would be crazy to program a website based on assembly, right?

AI assessment note: “it will take time, because we're at the early stages.”

Q change topics to the right way to work with non-technical co-founders. Um, and I think the topic that came up most commonly in the Startup School Forum was the one that Ralph alluded to, which is how to deal with deadlines and timeframes, particularly with non-technical co-founders. Um, and I, I think particularly in the, in the early days. So does anyone have thoughts on that? Any, any orders? Good.

A I'll volunteer for this since my, my technical, non-technical co-founder is staring at me over there. Um, a few, a few ideas here. If you've got someone that's really non-technical, and I'm not saying my co-founder is really non-technical. I'm sure she can use a computer. Um, but you know, if it's really non-technical, this is an amazing testing opportunity for you. I mean, I remember I would write this software and I'd be so happy about it and I'd hand it to her. And as soon as she touched it, it would break. I mean, I don't know what she did. She shook the iPad, she rotated it three times, but she had a way to break the stuff I wrote, and, um, that was amazing for testing in the beginning, so that's one way to engage your non-technical co-founder. Um, you eventually have to learn how to start pouting your deadlines. That's really hard to do. I never got very good at this. I'm always optimistic. Even to this day, I'm like, oh yeah, it's gonna take me a week. So I never really got good at this. I'm just aware that I'm Optimistic about it, and then I'm always kind of off by a week, and she's aware of that as well. Um, so eventually I've talked to other engineering leaders that keep two books. They keep, like, a separate set of books they talk to their technical co-founders, non-technical co-founders about. Say, okay, here's the deadlines here, and then you have another set of bo…

AI assessment note: “You eventually have to learn how to start pouting your deadlines.”

Q Let's go back. You, you, you were working on distributed systems. This was interesting to you. How did this turn into the company? Like, what was the thing you applied to YC with, right? What was the timeline there?

A I applied to YC with, with the plan of doing, um, you know, this, of building both IPFS and Falcoin and a company called Protocol Apps. I mean, it was right away from the beginning. Um, it was like this large scale plan of going to do, build a whole bunch of different things. Um, all around, um, distributed peer-to-peer systems, all about decentralization, and with a business model of taking a portion of currency. Um, and this was in December, when this was a very new thing. People weren't doing this. Um, there was basically Ethereum and a couple other groups that had also gotten to the same conclusion. And I mean, it, aside from a few side projects that we've started and so on, and like basically like delaying our timelines in terms of like software taking a lot longer to build than, than expected, we've pretty much followed the plan, um, in that it, you know, from the beginning we had both IPFS and Filecoin, um, and the, the, you know, I guess connecting to, to what I was saying earlier, so I had this problem around data sets and, and versioning and so on, and that led down the rabbit hole of like really thinking through, um, how information moves Uh, in, in the network, how information moves on the internet in the first place. Um, how does addressing, um, how does it, how, how do we do addressing in general? Um, it turns out like with HTTP and so on, we don't, we do all this…

AI assessment note: “I applied to YC with, with the plan of doing... building both IPFS and Falcoin”

Q But like, that came out of the internet, right? The Linux kernel, That exists because the internet.

A Yes. The Linux Journal is an awesome example. I think like, um, you had the ability to, to undergo, undertake these major, major infrastructure projects, um, and like things that take a long time to, to create and mature, um, on the internet. And a whole other interesting avenue here is how do you fund these things? Like how do you, how can you fund these Long-term endeavors that are much more open-ended and on the internet and so on. And that's what Bitcoin and Ethereum, um, proposed one example of how you fund that. And this is, this goes back to what you were starting to bring up earlier, which is the idea of, um, you have a protocol and you have, uh, you take that protocol and you say, hey, it's gonna create a whole bunch of value. And it also has this This native token that's going to address a whole bunch of that value. Not all of it, but some subset. And that native token is going to be of limited supply. So because we were creating this token, we can take some of that token and give it to the people building the protocol, which then helps, you know, they can sell it for dollars or whatever to then feed themselves. And then that way they can like actually fund the development of the project. And this is effectively what, what Ethereum did, right? That kind of funding model, um, allows, um, people to remain very close to the, the actual protocol layer and to think deeply …

AI assessment note: “Yes. The Linux Journal is an awesome example.”

Q Wait, so to touch that though, so what, you're looking for something that doesn't use human intervention whatsoever? It's a purely algorithm of the cancer?

A I think it's fine to feed in human intervention along the way. You know, there's interesting research It's done on large companies and governments where you have all these, you know, peer reviews, uh, you know, kind of like, and manager reviews and, and, uh, all this kind of, you know, three, six, a review kind of perspective. And out of that, you can get good signal, right? Like, otherwise, if we didn't, weren't getting good signal, then there's no hope for, for any kind of company that's large, right? And so surely something's working. Um, and, and there's good research that shows, like, you can definitely get interesting human feedback on the, in the loop, and you can take that as a signal. But the hard thing is, I claim that what we need to do is, is allow the collection of that feedback to have humans in the loop, but do so in such a way that it is extremely difficult to game, because, you know, again, that's, if you, if you give people, people will quickly learn that they can just, like, give each other really high ratings, and that will translate into really big boosts and promotions and so on, or, like, you know, greater rewards. So you have to get something that doesn't, like, it's not easy to game. But then further, if you take people out of the equation in the choosing part at the very top, like, all of those feedback All that feedback always propagates all the way t…

AI assessment note: “I think it's fine to feed in human intervention along the way.”

Q Was that when they broke, broke apart the monopoly?

A That's right. So, so breaking Bell apart effectively stifled and killed Bell Labs. So a few things happened. One was the rise of Silicon Valley and the great invention, or like not invention, but like the great use of stock options, um, or just giving stock to, to everyone in the company. Working on something? Cause a ton of people, um, working on very, you know, research oriented things at the time to become quite wealthy, right? Or, or like get, you know, very significant personal returns. And that coupled with the excitement around all of the stuff that was happening in Silicon Valley in the fifties and sixties with, you know, a number of people kind of moving out and then coming back and, and, you know, talking about all the great and exciting things that were happening in the West started to drain a lot of people out of Bell Labs and out of Boston. And so, you know, it's known as this, like, brain drain. And, uh, part of that, what, what happened there was not only were people leaving and going and creating other research organizations that had different funding models, um, but Bell also started getting broken up. And, um, this is more like the eighties, nineties. I forget the exact date on this. Um, but when, when Bell got broken up, Bell Labs had to find a way to, like, charge the, you know, new separate entities for all of its work, and it just became infeasible to fund…

AI assessment note: “That's right. So, so breaking Bell apart effectively stifled and killed Bell Labs.”

Q How do you do it over 10 years? Like, let's talk about, let's like drill down to, because that's a great point. We were just talking about it. What's the tenure? Like, how do you, do you have to keep selling bits? Not you personally, but if you're one of these folks, do you have to keep constantly reissuing tokens to keep feeding yourself?

A I mean, if it depends on whether another token appreciates, right? So if the token appreciates enough, then you're, You're gonna have to sell less and less of it over time. Yeah. So you saw this happen with Bitcoin. I mean, there were people, there were some people that were early to Bitcoin that are, are now, you know, they have their, their personal wealth at a point where, um, you know, unless there was a major crash in their assets, like they don't have to work again. Yeah. Um, and you know, Bitcoin is 10 years old now, almost. Right? So it started in 2008, dozen nine? Roughly. Yeah. Um, and so like, you know, it's roughly 10 years old. And yeah, I mean, I think, I think maybe you could claim that the origins of Bitcoin happened through the Cyberpunk mailing list, and Mojo Nation, and all these other things, and all those discussions, and so that was like long-term innovation that happened, uh, and then only was getting funded afterwards. Um, so it's like a, you know, very different approach than, say, the Bell Labs, you know, centralized perspective. Um, but, uh, yeah, I think, I think the funding of these things is gonna, uh, Depend entirely on whether these things are, are continuing to be useful, right? So if Ethereum continues to be useful five, 10 years from now, you're going to have, and continues to accrue, um, continues to grow, right? So if Ethereum becomes more a…

AI assessment note: “if the token appreciates enough, then you're, You're gonna have to sell less”