Within the next few years and decades, major technological advancements will bring massive scale changes that will fundamentally alter what it means to be a human. Artificial general intelligence, brain-computer interfaces, augmented and virtual realities. These developments will shape how we live, work, play, interact — exist. Conversely, if we aren’t careful enough, these technologies could also cause dire unintended consequences, including mass extinction or basically, enable us to accidentally wipe out our entire species.
The future is a tightrope, and it’s more vital than ever that as we develop these groundbreaking technologies, we are intentional in baking in good outcomes during the R&D process.
On September 28, 2023 at the PL Leads Summit in Iceland, Juan Benet, founder and CEO of Protocol Labs, shared his vision for how PL is responding to these changes by creating an ecosystem for projects, teams, and companies to build technology for today and the future. It will require a new way of thinking, an ambitious vision that includes building an innovation network with a comprehensive research and development pipeline, fostering world-class teams, promoting open source technologies, and utilizing the crypto and venture capital model to support and grow projects. The goal is to reshape economies and governance systems globally while ensuring safe outcomes and addressing challenges, such as securing the internet and digital human rights.
For a lot of people, this section will be very familiar. A lot of people have heard this hundreds of times before. But just based on a lot of the questions that we've been getting, I figured it would be good to go through and kind of talk about the overall vision for PL from the founding to now and kind of carrying on in the future.
So we're going to do this in four sections. First off, talk about the mission and vision of PL — kind of the why, the what, and the how we do that. We're going to take a brief look at PL’s past, like the first 10 years, we're going to look at how the network operates now in the present, and then we're going to look ahead to the future.
So, PL’s about driving breakthroughs in computing to push humanity forward. That is a broad statement. It’s intentionally broad, and it comes from thinking about the massive-scale changes that are going to happen, have already been happening, and are going to be accelerating for humanity in the next 100 years and so on.
Things will be built, implemented, deployed, that will change what it means to be human, even more than it already has. So when you think about your personal life, and what you are as a human with these kind of supercomputers in your pocket and your capabilities — that being able to interact with humans all over the world — you're just a different kind of thing than humans 100 years ago, 500 years ago, 1,000 years ago, 10,000 years ago. And so that kind of shift will accelerate through a whole range of technologies that will arrive in the next 100 years, the next 10, 20, 30 years. And so the goal of the network is to be able to both accelerate some of those changes. But more important than that — to navigate towards safe outcomes through those changes, because those changes can create massive upheavals, and you want to direct the development towards good, safe outcomes. There's a lot embedded in that. And that could be like, a whole conference just about, it's like how to do that.
So in general, the kind of broader vision of PL is to build this innovation network. This used to say, an R&D company, and that shifted in the last three years. We'll talk through why.
But the rest of the picture remains the same, which is to build a full research and development pipeline across the entire thing, from early research all the way into product building, and deployment and scaling. We want a world class set of teams and network. We are very strongly into open source and open technologies. And we lean on the crypto and VC business model to create support and grow lots of projects.
So at the last LabWeek, I gave this PL Vision talk. I'm going to plug it again here because a lot of the questions that I saw coming through are just straight up in that talk. So I'm gonna give a summary of it now, but it's gonna be very compressed. If you want to unpack it more, I recommend you just look at the summit talk (opens new window), especially after kind of thinking through a lot of the stuff today. And I'm also very happy to have a problem solving breakout session on digging into any of this that is still kind of unclear for people.
So the broad picture of pushing humanity forward comes from looking at the past and looking at the present and the future. We've had this massive scale global improvement over the last few hundred years. We've also gotten ourselves into a position where we have technologies that can wipe ourselves out. And we're having this massive extinction event across the whole planet where we have massive challenges, we've become aware that, hey, we can accidentally wipe ourselves out. But, either by our own doing or by a mission of not solving some problem, that might happen. At the same time, we have this super crazy phase transition, where computing is just a super different kind of thing. And we're developing at a faster rate. And if things weren't more volatile, we just have inadequate macro systems to handle this kind of change at the rate that it's going in. And so there are tough problems there to solve, in terms of how to orient the global landscape towards good outcomes.
I'm going to hold questions for the Q&A, because we could probably be — if I start taking questions, we're going to really be in trouble in terms of time. But write it down and then let's make sure that we cover it.
So now kind of that's kind of the broad why. I want to talk about the kind of concrete what, so get very specific for a moment and then back into how we can do those specific things.
So broadly, there's a lot of sub-missions in this broader picture that a lot of us have been working on for many years, that a lot of us really care very deeply about, that are interconnected in ways.
And this is not comprehensive. There's a lot of people working across the network on a whole range of things. But it's broadly related to this kind of swath of things. First and foremost, PL started with IPFS and Filecoin and libp2p and projects about securing the internet, and establishing digital human rights. I'll go into that in a moment.
We also discovered through crypto and smart contracts and blockchains, and so on, that you could actually deploy some new structures that could shift how economies work, globally, and how governance systems work, globally. And that's a very powerful technology that I certainly didn't anticipate when I was first starting PL. And though it was talked about in a lot of CS literature, and broad economics literature and so on, it wasn't clear just how straightforward and easy it has been. Like we've built this massive scale, new financial rails infrastructure, and you can deploy new incentive structures into the world.
And that has been a relatively easy thing relative to many other buildouts of other industries and infrastructure. So that is very promising as an opportunity. And then broadly, there's a whole range of other shifts in computing technology that are going to be arriving — shifts in simple interfaces, like VR and AR, where those kind are symbol translations of what are used today, but just in that kind of more immersive environment, to then major shifts, like the latest super-advanced AI models, eventually getting to things like AGI, robotics, brain-computer interfaces, whole-brain emulations, and so on. So this swath of technologies is going to come with all kinds of super difficult challenges. And it's one of those things where innovation like this can't really be stopped, in a sense. It can only be delayed or accelerated. And you want to think carefully about the sequencing of these things, to aim for good, good, safe outcomes.
So you know, concretely in each one of these, and I’ve condensed here what normally could be like whole talks into a short segment, but a lot of you are very familiar with this. The securing the internet and establishing digital human rights thing comes from observing that the internet and the computing infrastructure that we have now has connected the species in a really amazing, open sort of way, and has given us the ability to grant superpowers to each other by just writing some software applications and deploying them worldwide. And at the same time, that same infrastructure can be used for all kinds of bad outcomes. And in particular, there's a lot of opportunity for certain kinds of central control failure modes, where the kinds of like, digital totalitarian kind of states that you could build it — now or in the near future — just kind of eclipse whatever we've imagined in the past, in terms of how bad things could get. So it's pretty important to bake in human rights and our notion of human rights into the internet now, before those kinds of things emerge to prevent those things from emerging in the first place. And there's lots of ways to do this, and lots of approaches and so on.
One way that the broad Web3 community has been following this is by thinking of specific human rights, thinking of systems that can give you those rights, and then implementing and deploying those systems worldwide, and to try and build those so they don't depend on corporations or governments because those are coercible and can change fairly quickly, and instead lean on large scale incentive structures that can coordinate lots of groups around the world. Again, this is a huge guess and bet that this could lead to a whole set of better outcomes. But none of this is at all guaranteed in any way. The Web3 community in general has been pushing for this better safety model.
There's a lot of people across the network that have been working on this particular mission. This is where PL got its start. Projects like IPFS and Filecoin. IPLD. Ethereum, itself. We were very involved — we've been very involved in the Ethereum community from the beginning, things like drand, libp2p, Bacalhau, and a ton of others. And this is a showing of a bunch of the teams across the network that are just working on this. This probably misses a bunch of others. I just kind of did a quick filter. But you know, lots of humans, and lots of organizations and communities are working towards this particular goal set.
Now, kind of in terms of operating economies and governance systems, this comes from thinking about that kind of inadequate macro systems problem space. Thinking about messes in the landscape of funding, when you think about the last 100 years or so. We've advanced our understanding of what's important to do and we've advanced our understanding of how to develop really breakthrough, broadly globally improving technologies, products, services, systems, and so on, but the incentive structures that we have, globally, don't quite lead to the best outcomes. And you see all kinds of really messed up situations all over the place. Things like, the R&D budget of the world remaining roughly the same as 50-60 years ago in actual amounts of value in dollars or any other currency, even though the world has inflated those currencies dramatically more in that time period.
And it's very difficult to motivate the addition of new, whole scale, new kinds of funding agencies for new things. There was this very interesting period, in kind of post World War 2, where the U.S. especially but also many other countries, worked pretty hard to set up a whole new R&D infrastructure to carry on the kind of the speed of innovation that the war had created. And then aim it towards solving a broad-scale set of problems. And the Cold War gave this ability to kind of create new agencies and so on, out of which we've got things like NASA, things like DARPA that ended up building the internet, and so on. And yet that kind of innovation has sort of stalled out.
There's still a lot of that kind of innovation, but it's not yet proportionally the same scale that it was before. However, crypto is a super interesting set of systems where, when you look at the scales of funding that are available to these crypto networks, and you think of them scaling, I don't know, one, two, maybe three orders of magnitude more in value. At that scale, crypto networks can actually start funding R&D at the scale of nation states, which is a super crazy realization. I kind of realized this a few years ago and it was kind of mindblowing, that you could actually build and fund things like DARPA, things like all the different kinds of funding agencies, out of crypto networks. Now crypto has to scale to be able to get there, and we have to figure out the mechanisms to be able to do that, but that kind of thing is possible.
So that thinking plus thinking about a bunch of other problems, like climate change and many other kinds of more local classic economic failures, led us and many other groups to develop a bunch of the systems in kind of public goods funding and this kind of structuring of coordinating broad groups to develop, fund, maintain, sustain and grow this infrastructure that is vital to all of us.
There's a lot of teams working on this across the network. This is of course much smaller than the prior one because it's newer. But this is just a snapshot of a few of the teams and projects and groups working on this. And there's many more across the broad Web3 ecosystem. I think this kind of thing will, if we can have some good short term successes, we could probably scale this quite a bit.
Now kind of zooming out for a moment, going back to this broader technology picture. The ordering of these technologies is going to be very important to get right and what kind of capabilities get unlocked and become enabled globally. And what they afford to various groups is going to significantly shift the outcomes for all of us globally. And certainly shift the outcomes locally in a bunch of places. And kind of our thinking around R&D, in general, is that this happens in this broad larger scale pipeline, going from early research and things like you know, you can think of academic research and so on, and you develop new ideas, you try out various different possibilities. You develop the conceptual theoretical framework for things. Then you start applying that into devices or concepts or structures. Some of those might work out and pan out in the near term. Some of those will sit on shelves for decades even. Wireless technology, for example, sat on shell in a shell for a decade, before it was broadly useful.
And then over time, you need a lot of effort and teams and projects, working on developing those things to find the concrete applications that will get you there. And so that requires not just a single team, it requires multiple different teams pursuing different lines of thinking, different lines of development in that large fan-out space. And one of the hardest parts about this R&D pipeline is that usually startups start very late in this pipeline. They start towards the later half of it. And startups are currently the best way to kind of capture some amount of the financial return from those technology developments, to then be able to kind of cycle back the ROI to the beginning of the pipeline. And so you really need to have a network oriented structure to be able to develop these kinds of things in the long term and at scale, and then be able to leverage the returns of startups and so on.
So we both want to accelerate certain lines of these kinds of technologies, and we want to slow down others or at least kind of sequence these to some extent, to try and maximize the outcomes for safety.
There's a number of organizations in the network already working on this kind of thing. Many of these organizations were started around the same time as PL, or a little bit before, or a little bit after. Foresight of course is pretty old. But FLI and FHI are fairly new. Convergent is newer than PL. And then there's now even startups that are working on some of these technologies. And so, very proud to say that two of the most important projects in AI safety and BCI are part of the PL network. And so even though it might not be a large scale, in terms of numbers of teams, we've been successful in working with some of the groups that are having the most impact.
So there's a lot here. And the question is kind of how do we drive these breakthroughs in computing, orienting towards safe outcomes across this kind of network oriented structure?
How can we kind of accelerate that good R&D that can get us into good spots, and that's where the innovation network model comes from. I started going into this a little bit in terms of the R&D pipeline, but zooming out for a moment, at the end of the day, is just about science and technology, and how you couple those together and build products that can diffuse innovation around the world. It's about how humanity learns more and figures out what will actually work. How those learnings can be translated into concrete artifacts or systems, whether they're physical or digital, that you can leverage, or social, like there are social technologies, like language and cities and so on.
And broadly, in the last 20, 40 years or so, we've kind of created this very large innovation chasm between the early academic works that are primarily driven by academic credit, and so on, and then the broad technology development incentive field, around building products, building services, and scaling them for diffusion. And that in-between spot gets not well funded, not well incentived, and so on, which is why tons of really important innovations are stuck in the academic pipeline, or are stuck in heads of people or in papers or in discussion forums, or even as practical implementations that got built to some extent that got stuck somewhere along the way, and never actually diffused all the way.
And so we've been very Bell Labs inspired from the very beginning. And I’ll go into that a little bit more later, but the way that Bell Labs oriented this is to think of the entire pipeline and structure programs and processes across this whole range of operations, and then was able to drive that kind of innovation that we can kind of haven't seen since, in terms of the scale and predictability, in one kind of broad organization. In reality, that was kind of a system of organizations, when you peek under the hood, it wasn't just a single company, it was a whole system of things. But it's not quite the same thing as an innovation network as we're talking about it now.
I've seen a lot of questions about the differences between single companies and innovation networks and kind of why a network, and so on. I'll share some of this briefly, and I'm also very happy to kind of have one of the problem sessions be about digging into this in detail and see how they're different.
But the broad thinking around this is usually in kind of a single company, you have one corporation or a set of related corporations. You have primarily one dominant business that then ends up starving and stifling others, and usually tends to lead to that culture in that organization, decaying over time. You can see this in all of the Silicon Valley tech companies that have scaled and kind of gotten to the top tend to kind of decay down to some sort of steady state.
You tend to have highly hierarchical decision-making. You tend to be able to create great, exceptional company cultures. But those degrade with scale and time, usually somewhere between five and 10 years. If you're very, very good, you can stretch that out to maybe 20 years. The big question mark was on Google this last 30 years, and it has not for my accounting has not really panned out as maybe they thought it was going to
Single companies are good for sharing resources and establishing a different internal environment. But when you do that, you tend to increase the difference, and you tend to create very strict boundaries, and it becomes harder and harder and harder to collaborate across those boundaries. Usually, IP becomes a big question. And the broad innovation pipeline gets stifled by the weight of the single organization.
It tends to also be very capital intensive because you don't have good forms of sharing the upside with many other groups or investors and so on. It's hard to do very significant R&D outside of the core domains. Some organizations managed to do this, but it's extremely difficult. Most organizations never even get to R&D in the first place. And then you tend to have a lot of coupled risks that end up inhibiting the broad scale R&D. So it's really good for tightly related businesses, for products and R&D that is just kind of around one swath of things. If you're extremely good and lucky, you can maybe open up that aperture. But in general, there haven't been good examples in the last 20-30 years of doing this very well across a very large swath of different technologies.
Or at least, it's sort of relative. There's some successes, but you can look at other models that are actually better. And in this sense, it would be bad for a large span of independent projects, different businesses, products, R&D lines.
So now let's think of some kind of innovation networks. I picked three examples to go against these other ones. First off, think of Silicon Valley, itself, like the concept of Silicon Valley is an innovation network. It's not a very concrete one. But it is. People have written about this for many decades of what was special about Silicon Valley and why it ended up manifesting the R&D that I did, there are others before and others in other locations. You can think of universities, like Stanford as one of the more successful at this. There are many, many across the world that do this pretty well, where they're able to create a kind of innovation network, but they usually don't route the capital flow back into the R&D pipeline. So they have — they’re concrete and you can point to them, and you can see kind of research programs that are articulated. And you can kind of see the boundary of a network. You don't have a way of cycling the ROI because of how universities are structured. And then you think of something like YC as another kind of innovation network that is kind of tightly coupled to a specific phase in the pipeline, which is, again, not in the early research part, but more in the production and station and scaling side.
So this kind of broad concept of innovation networks, you can think of it as a network of independent organizations or corporations, each one pursuing its own business, making its own decisions, but you can create the kind of internal environment, sort of things that you can get in single companies happening as a broad cultural structure, or broad systems and services that are shared across those organizational boundaries. So think of it kind of like if a single company is one cell with a bunch of kind of a different internal environment than the harmful outside. An innovation network is trying to create that kind of structure without the boundary cell wall, but having enough momentum inside that you get internal cohesion. It's closer to a colony of cells than a single cell.
Part of why they're extremely good is they tend to create a resource sharing structure that isn't dependent on that single business. So if that single business succeeds or fails or degrades, it doesn't really affect the whole outcome of the network. And so that can keep growing with many different structures. You get the coupled incentives. You get amortized failures, meaning groups can move across different efforts without kind of like blowing out all of the work that you've done for a decade or two building up those structures. This tends to be dramatically more capital efficient, you end up sharing a lot of the risks and the upside with many other investors. So you can create structures that are just kind of astonishingly more capital efficient than single companies.
And it tends to be very good for large spans of independent businesses, products, R&D lines. It tends to be very bad for super tight coordination. Or it's not that it's bad, but it's not as good as a single company for tight coordination on one business, one product, one R&D line.
So this is another kind of way of looking at this, which is kind of taking some of those things and comparing them on a table. I didn't color-code this one because color-coding it orients it to a specific approach. I'm gonna do this at some point for the kind of broader R&D question, but if your goal is to kind of get somewhere really fast with one product, the network feels like a distraction, it feels like something different. If you want to go really fast on one single product, usually a single company is better. But that's not what the mission of PL is, right, like the mission of PL is a much broader, larger scale thing.
It also happens that crypto networks by design and definition need innovation networks, not single corporations. And if you have a single corporation, driving most of the product development in a crypto network, the thing is likely going to fail because of how we've structured what it means to be a crypto network and what decentralization for a crypto network means. You need a structure that has resource sharing, and independent risk decoupled structures across those things. So you can go through each of the crypto networks that are deployed out there today and try to classify them on the spectrum of how much — are they driven by a very few organizations or a very large amount of organizations that are more decoupled? You can scale that by capital, by the way. You can normalize it to just kind of think about the same capital structure. And you'll tend to find that the crypto networks that do the innovation network part right and are much more successful.
In another, this is kind of an old slide, it's another way of looking at it, which is Alphabet set out to do this kind of longer term thing inside of a single corporation, and yet, it was not nearly as successful at doing this as it hoped to be. And when you compare it to the outcome of YC in that same time period, even though YC did not set out to do what Alphabet did, it is just kind of astonishingly better to be in the network model — much stronger output and much more capital efficient.
Now, this suggests that you can create a crypto network like this to give you this kind of innovation structure, and that's what PL is going to be experimenting with and trying to figure out. Can you create crypto network structures to do this kind of innovation? But that's kind of like a sub-goal; that's not the main broad goal. The broad goal is to just get it right, whether with crypto structures or normal, regular structures. But there's kind of like an open opportunity there for being able to do this. And, and, you know, build like the crypto network version of Silicon Valley, like, that's kind of what the potential is.
So this is a diagram of the pipeline that I’ve shared in a bunch of places. Here, I’ve also included the structure of Blue Funds and Green Funds, and the yellow service teams, because all of that segment is required and critical in enabling this broad, longer term R&D, especially with a network orientation. And that set of boxes in the middle is about like different instruments for doing the funding at different scales, which is what leads to these different funds. Green Funds, and Blue Funds tend to be different because the instruments on the return profile are different. And again, having to go into this into a lot of detail when people want to dig more into this in a session. (Here's like a zoomed in, zoomed in view into each of the sections.)
By the way, this kind of shared services component was so fundamental for the development of Silicon Valley and the development of things like YC. You were able to build an enormous scale of professional quality development that you didn't have to kind of build from scratch in one organization. And it's one of the things that led to Silicon Valley overtaking other innovation environments. And so being able to kind of build very high quality, professional services that are not coupled to the single business, but that are able to work across them. It's kind of a huge advantage.
Another way of looking at this is — where we localize some of the things that we've been talking about in the network and differentiating or distinguishing PL from these other things — think of YC as a network that's very oriented into this particular stage in the pipeline, when you're going from something that like, the concepts are broadly figured out, and you're trying to productionize something and trying to find product market fit and trying to then build a business and scale it. So this is kind of where YC sits. You can think of specific VC funds, like a16z and others, as kind of sitting downstream of that on a larger scale. It's worth noting that, by my accounting, this a16z model is more profitable than the YC model, because you're able to look at specific areas, but it is not successful at causing the startups to emerge in the first place.
And so something like YC will kind of output many more potential opportunities, which is interesting to see that now a16z is creating its own version of YC, at least for crypto.
So now localizing PL in this picture, this is where it's different from these other things. It's not looking at a specific set of stages and trying to out-compete the other groups in that environment. It's about looking at the entire end-to-end pipeline and building some structures along the way that can speed up that part of the pipeline. And you can look at sort of the conversion rates along the way, and you focus your effort and energy to increase some of these areas. And you of course have to do that with an eye towards ROI along the way and so you have to couple this to the investing model because that's currently the easiest way to have a strong ROI here. The other way that's also potentially interesting is the venture studio model, which, if we're to overlay it here would be kind of in the development part and like that chasm area. And so that's something that — where things like the big bets that we're exploring or like, the various bets that we're exploring, sort of fit in.
So that's the broad mission and vision of PL. Went very in depth here because a lot of the questions people had were kind of around this, and kind of the why, the what, and the how. And now I’ll just speed through these next sections — that was kind of like 80% of the content. So I'll kind of just go through PL, to just kind of zoom out, look at the whole picture, talk about the now, and the near future. So, you know, before PL got started, the things that led to the beginning of PL and the beginning of IPFS, where a whole bunch of learning across many different fields, looking at the internet stack across the board, learning how to build companies in the first place, and learning how to build products, and then so there's a whole track around internet startups and so on, and the kind of developing that. There's a whole different track, which is about noticing the innovation failures that led to wanting to create another Bell Labs type of environment. It was astonishing for me to read about the past and the kind of R&D that we used to do as a species that we no longer do, and why certain things are totally stuck and dead in the water now, and kind of why that is.
So my broad plan even before PL was to kind of find a good structure and a good business, and then orient it towards solving this broader problem around the pipeline.
This is kind of a bird's eye view of the first five years of field history. Getting started with very little money from YC, like 120k was like the very beginning of this. In fact, I worked on IPFS for like many months for like six to eight months with less than $0. I was actually like completely, absolutely broke — I can potentially go into this more in detail tomorrow when I talk about the founder journey. But that enabled PL to get its start, and the big initial thing was IPFS and Filecoin together.
This was a period where we were coming out of a crypto summer and into winter. And so that's why IPFS got developed first as opposed to Filecoin because the winter made people have less attention on Filecoin and more attention on IPFS. And so I just kind of followed and pursued the path of you know product market fit.
That led to kind of the beginning of go-IPFS, launching the alpha, socializing in a bunch of different places. The community started growing from early days and contributing a ton. We started building a larger company. Then we were able to raise a funding round in 2016. So know that like, up until 2016, for like two years, $600k supported about 12 people for that entire time period.
Then that led to — then go-IPFS continued with a ton of traction, tons of scale, lots of contributors. Then we started refactoring the stack to add things like libp2p and IPLD, multi formats and so on. We then turned our attention back towards Filecoin to develop it. There's so many people around in the room today that were here for that period, when people were like, why are we doing this Filecoin thing? That seems like a huge distraction over IPFS. And even, why are we doing this libp2p thing? It seems like a huge distraction over IPFS. But it's really fundamental and kind of the model of how do you build the financial structure to develop all of these systems?
We built the SAFT. Then we put out the Filecoin white paper, round two — you know, this two year lag, where we didn't do much in Filecoin during this period. What was neat about this is that Ethereum developed alongside, and we were very involved in the Ethereum community because of IPFS and the applications there, and more because of the social structures. And so we were able to kind of learn a ton of lessons from Ethereum along the way, and then use a lot of the tech that they built. Then we had the broad announcements of Filecoin. We ran the Filecoin SAFT sale. We also had the first official nucleation of PL, which was CoinList. We built this entire platform to do this fundraising and then we nucleated that as an entity, and so its own company.
We weren't using the term nucleation at the time, but that's kind of like the first example of creating another organization. That led to a pretty large fundraise, and now at that point, now that we had larger scale funding, we were able to then start funding projects at larger scale, not just our own individual people working on our core projects, but we started going into grant funding, and we had RFPs, we built an RFP program, we worked with a bunch of different centers. And for those of you around at the time, we started kind of socializing this concept of PL as a stellar nursery, of creating different projects and different businesses and so on, and enabling them to scale. And then at that point we scaled up to about 100 people. I might be getting the counts off here because I just did this quickly. And then second, there's a massive scale of contributors around this smaller core set. And in terms of scales, we're talking about thousands of people at this point, working across all our projects.
Then kind of the five years after that, we had to focus super hard on Filecoin in 2019, 2020 — a super difficult time period for a bunch of reasons — but we built it. We built the beginnings of Filecoin, and we got to the alphanet, if people remember that, and then the testnet, and then from there into the Filecoin Space Race. And then the mainnet launch.
There was other stuff going on during this period and a bunch of other projects, but my memory is a blur at this point (and when making this slide yesterday, just kind of very focused on that). And then from there and kind of 2020, late-2020, early-2021, by looking at the whole pipeline and looking at crypto networks and looking at kind of like the endgames, or not endgames, but like midgames and endgames of these types of systems, that's when the decision to transition to the network came into view, and I made the decision to go in this direction. And we rolled that out into LabWeek21. We talked about this kind of broad transition into the network. We introduced the concept of nucleations. We formed teams to do the Builders Funnel at the time, funnel network infrastructure, Network Goods, and so on.
Then we went into ‘22. Lots of things happened in ‘22. People's memories here will be really good and better than mine. We then, in terms of Filecoin, we focused really hard on scaling data onboarding and driving that as a big piece. Then we kind of built out the accelerator program and the VC program. We had already been doing a lot of investment by this point, so the first network investments were starting to happen in 2017. That's when we invested in things like StarkWare (opens new window) and other things that are now pretty successful.
And then we just kind of scaled up a program to enable it to turn into the network. We had LabWeek22. Then from there, we've been working on various projects like the launch of FVM and the formation of big bets. We have LabWeek23 coming up. And this is just only a small sampling of the massive scale of stuff that is happening across the entire network.
So now we're kind of thinking about PL, you know, the present state, kind of setting aside the — there's all of the projects that all the individual teams are working on, but just kind of looking at the network infrastructure. Think of the broad, the notion of like, sort of, like what PL is, is this broad network of teams, lots of different activities across it. And then we have a set of infrastructure teams that are supporting the network. And this is where this picture of building a network, helping teams start connecting into capital and talent and knowledge and services comes from. This is all in other decks that you can find, but this is the sort of the scale that we're dealing with. We now have hundreds of companies in the network, thousands of people in the network. We have systems that help groups coordinate, in terms of finding each other as teams, finding each other as people, office hours or other kinds of support structures. We enable a lot of teams to get started through things like the accelerator programs and early hackathons and grant programs. We then help groups connect to larger scales of capital — everything from angel to seed to larger VC funds, and so on. We have this broad model of different funds structures for different sizes and different scales. We have this broader portfolio.
And we also want to connect all of the teams with really good talent, so that's where the talent network funnel or Talent Network, as we're calling it now, came from, to help organize hundreds of teams and thousands of people. And this broad swath of services that we're developing — so all kinds of things to help coordinate across a network from local sorts of things, like local dinners, and LabDay and things short in time, or kind of updates, to this broad thing, supporting the entire network, tell its stories, and its messages and broadly disseminate what they're working on to the broad landscape, and then do things like LabWeek to help bring the network together and help form and help coordinate.
So in terms of the concrete goals of the concrete projects, there's an enormous amount of people in teams working across this whole swath of individual goals. And this is just a kind of sampling of these.
Now, talking about the future. It's extremely difficult to predict what's going to happen in the next 10 years, but I'll just give some broad strokes of this. So first off, the most important thing that we're going to be facing, as a group, as a species, is the rollout of some of these technologies that are going to fundamentally shift and change what it means to be human.
It's unclear if we'll get any big results of this kind in this decade. If I had to bet today, I think the bet would be on ML systems getting pretty close to things like AGI. It's extremely difficult to estimate timelines, but you know, things like OpenAI are targeting 2026, 2028, which is not very far from now. And that would be transcendental if we get there.
Also very risky. Not gonna go into that.
But then BCI. From my perspective, the prioritization right now should be in slowing down AI and speeding up BCI as much as we can. And so that's where we're very underserved in terms of the scale of effort there. I think companies like Science (opens new window) will have a pretty significant impact. But we need kind of like 10 to 100 of these not not, you know, one or two.
But hopefully, we will kind of steer the pathway well here. The broad thing that we can do is create an environment where starting those kinds of things can be dramatically easier. And starting those things looks fundamentally different than starting crypto projects or crypto networks, right? It's just a very different type of environment if you're doing a kind of neurotechnology-type of work. And so in order to build the support structures for those systems, that looks different. But access to capital is one of the core things in all of this, which is why we were even successfully involved in that space.
Then kind of thinking about individual kind of goal sets, or projects, unclear to say where we'll be in this particular set of missions. I think we have a pretty good chance of baking some of these rights into the network in the next 10 years. Things like, at least kind of the secure publishing and distribution of content, I think we can get there. It'll be difficult to get there in terms of private read, read or write privacy, I think we can make significant improvements. But my prediction is that we won't yet be at full read or write privacy across the entire internet for all content within 10 years. It'd be great if we did, and we should try. But that seems kind of like a little harder.
Now, individual projects and individual missions will hopefully be at a much larger scale. If IPFS keeps growing at the rate it has been, it'll be hitting billions of humans at that point, which will be pretty, pretty awesome. If Filecoin keeps growing at the rate it does, then — I need to do better math with this — I think we could be hitting in the zettabytes by the end of the decade. Not sure, we will have to figure it out. It might be like multiple zettabytes, I've consistently underestimated the scale and speed at which crypto networks can accumulate hardware and accumulate resources. So, I sort of was thinking or predicting, like tens, to hundreds of petabytes for the beginning of Filecoin. And we had an exabyte very fast. So this could be much faster.
In terms of economic systems, I think this is where there's a lot more question marks. There’s a lot less people working on this kind of stuff than there should be. And so this is an area of a lot of work on investment that should happen across the space, especially as we go out of the crypto winter into a crypto summer. We should kind of be riding a lot of the kind of development and so on to systems like these. And then in terms of the R&D, broader R&D pipeline, hopefully we can get to a spot where it becomes a lot easier to find these breakthrough big bet-type of ideas, and then start them within the network. Today, like if you were to start something with a lot of experience, you tend to not go to YC. And that's something that's an incentive structure that's to the detriment of YC. And we have to figure out an incentive structure for us that way, like you just cleanly always make sense to come work with a network.
Here are like some of the kinds of things that we were talking about. This is now much more zooming in to the next year, two years, three years. We need some form of PL network membership structure that we can formalize, so we can see what teams and groups are involved, and what kind of access they get to different shared network resources.
We want to create this like a broad unifying platform across the network. Spaceport (opens new window) has been pioneering that with the directory (opens new window). We want to kind of expand that tool in that system to create a pretty important way of collaborating and being pretty effective across the network. We want to build ways of better connecting the talent across the network and helping support companies with advice along the way and with helping them find the people they need. And we want to create, like, we want to have a very clear picture of the conversion rates along the way and in the entire pipeline, to see kind of where the gaps are in what's working, what's not working, where we should focus on and so on.
Now, in terms of kind of like the funding, or the teams or funding structures, we want to hit like we're currently at hundreds of Blue to Green teams. We want to hit thousands in kind of like the short two to three to four-year timescale. We want to hopefully be in the tens of thousands by the end of the decade. That would be a pretty great place to be. That would probably back into thousands of service providers, or hundreds to thousands of service providers, depending on the scale. If you're supporting 10,000 teams, that usually kind of like an order of magnitude lower is probably a reasonable kind of upper bound. And so this might be like low tens of service providers until maybe, which is kind of where we are now, for the next few years, but then maybe scaling up afterwards.
In terms of Green Funds, this is not a single fund type thing. It's like you want a whole network of funds. Here on the green side, we can just tap into the broad investment capital of the world, we can lean into tens of thousands of angels out there, the hundreds of large scale venture capital funds and so on.
And then in terms of Blue Funds, this is where it's harder because the Blue Fund world is much sparser and it tends to be very top heavy. It would be amazing to create the structure in crypto that enables things like tens of thousands of angel-sized impact funds to happen, where you're like deploying small amounts of capital, like $5,000 to $50,000, or $200,000-level of scale, as single individuals being able to make those decisions and routing that capital often is not there. Angels tend to invest their own capital, but recently, there's a lot of structures that enable angels to route capital. And finding a good way of doing that on the impact side could be just enormous.
And it would be really great to get to hundreds of VC-sized impact funds by the end of the decade. This is going to be Blue Funds our way that's — on the Green Fund side, we're lucky to rely on the broad world having really good investment structures. On the Blue Fund side, we're dealing with a pretty broken system today that, in my view, for certain parts used to work a lot better in the past. So we'll have to think through sort of how to get there.
And one big thing that we're gonna be working on is — and we've made a lot of progress on it — to figure out some kind of asset structure, whether it's one or multiple types of assets that can create incentive structure for the people in the teams and companies across the network to collaborate and to create value together. And that's a really key point in making the network sustainable and growing over time — and, I think, much more successful than other things.