A DAO Shipped an Aircraft: What It Takes to Coordinate Atoms with Tokens

Transcript

Well, thank you. Um, thank you all for uh for being here and and having me at the conference. Uh, I'm Thomas and uh I'm not a I'm not a fancy one dev or smart contract developer, but it's really great to be here. Uh I've spent the past five years of my life working on arrow uh and most of that has been uh thinking through coordination problems and trying to solve all the various coordination problems that have have come up for us as we uh as we try to build aircraft with people all over the world. Um yeah, so we're uh I'll kind of go through this talk and tell you a bit about the history of Arrow uh our experience like how we got started uh the idea of building hardware as a DAO

and all the lessons uh we've encountered along the way and then uh I'll tell you how I think coordination is changing and why I think Dows are about to make a huge comeback. Uh so this is this is what Arrow started with uh back in 2021. Uh this is a longtime personal dream of mine that I've wanted to build uh what what are called EV tall aircraft. So for those of you who don't know that's uh aircraft that can take off vertically like a helicopter and then shift into forward flight like a plane. Uh and it's not an uncommon dream. There's there's lots of startups out there that are all trying to build this. Uh back in 21 there was a lot it felt like about 50 startups that had all raised money. Uh and they kind of play the game the same way like raise

a bunch of VC money. um you know hire engineers, build in secret, guard your trade secrets um and then many of them end up selling to the military. Um so I wanted to do this but like none of it was was like feeling exciting to me and at the same time I'm watching Dows emerge and all these like DeFi Dows and other Dows and uh wonderful governance experiments in the space and and it was all very inspiring to me. I'd been in the Ethereum ecosystem as well. So I kind of I kind of merged these two ideas. It's like what if we what if we took this like aircraft vision and um built it uh built it in public with with kind of the Ethereum open source ethos where we invite anyone to come in and uh

and work on this together with us and you know we've got the open source side and then the the Ethereum side to kind of handle the expensive parts of hardware uh where open source normally can't fund like workshops and manufacturing and that sort of thing. Uh at the same time I was watching kind of these network society concepts come together people talking about that uh and if if you're going to build a society of people that meet online and manifest in the real world like what better pairing than uh than affordable pointto-oint travel. So that that was kind of the original idea and we started out um kind of having some various community calls with my various friends in the Ethereum ecosystem. Um, and then I had this this very viral tweet thread in early 22, um, about these these scammers

that were trying to social engineer me into signing a malicious token approval. Uh, so that was kind of an early community that that all came out of of, you know, various pockets of the Ethereum ecosystem. And it's kind of funny because crypto people, many of them don't seem to get hardware and and it's it's like not a natural thing. And then aerospace people when I pitched it to them they're they're like wait crypto is just a scam. Like what are you doing here? Um so we we got this really really wonderful like small but passionate community as we got started. Um and we had just various like engineering calls on Discord where we uh we worked on designing our first aircraft that you saw on the last slide uh Spearhead and it all uh ended in this wonderful first meet up where the

community went from being virtual to actually manifesting in the real world. So that was our first meet up there uh in Turkey. And we uh we attempted to fly spearhead and it was horrible. It was it was super fun, great build, but the aircraft was not structurally sound. The wing came flying off the minute we turned it on. And it's like ever since then we were just playing coordination games trying to trying to get this stuff working. And there's this like kind of latent like doubt imposter syndrome as we're doing it that's you know can this online community of distributed engineers like actually be as good as a team of people sitting in the same office where they can touch the same hardware and work in the same way.

And that was like kind of just the problem we were facing the whole time. And you know up until very recently that's that's been there. And in then the past few months uh it's really just just gone away. You know, over the years, we've we've just tried all sorts of different manifestations on coordinates on coordination technology. Um, you know, we we kind of kept the token dormant, just trusted our insiders as we try to figure things out. Um, lots of different iterations. We tried like different pod structures where we like divided our community into into subdows. Uh, we did coordinate and other other types of peer rewards uh circles where we like try to assign value to other people's contributions. Uh we tried various different like runs at bounty boards and

uh when we realized it's actually hard to get people to write bounties. We formed a committee of people to write grants and bounties. Uh we tried a more like centralized uh structure with like named project leads and budgets. Uh and now we're moving on to continuous governance apps which which is really exciting. And then on on our hardware front, we started building uh big ambitious aircraft. And aircraft uh as you can imagine have a tendency to crash very destructively. And it's it's quite unfortunate because you can spend lots of time and money building something big and and it just crashes in a moment and you're like reset. Uh and the the iteration cycle is directly tied to that. So uh we stepped back into building from our large like

one seat men uh project to a 20 25 kilogram multi-purpose drone that we've found a lot of success with and uh building in workshops all over the world now. Um so the main thing that that we've come to realize in the past few months is that AI assisted coordination is just this new wonderful wave uh for online distributed communities. So things that just were awkward and like no one really knew what we going was going on before uh suddenly are starting to work. So like every meeting we have in Discord gets like a transcript of who said what.

It goes into the corpus of knowledge that AI agents can can query. So anyone can kind of like use our agent or point their agent at at all the notes and uh the documentation is just getting updated in real time. So anyone can just kind of pull up and and see what's going on and look for something they can bite off and contribute. Uh, another great example of that is is CAD CAD tools. Uh, CAD doesn't natively play nicely for 3D files with like softwarebased version control that's just looking at text. Uh, there are some wonderful like codec CAD libraries, but uh, what I found is that most hardware engineers are kind of adverse to switching from their like nice GUI tool into writing Python code to design hardware parts. And you know what we've seen now just in the past few months is

these uh AI LLM models are actually good enough to just like kind of interpretate and and convert things between different formats. So no matter what software you're using for your CAD design, like you can kind of just merge it into the same repo. Uh and it seems to work. Uh so yeah, we've just been building lots of wonderful coordination tools on the fly and it's it's like I don't even want to call it vibe coding because vibe coding implies that you're developing a software product where this is like using uh you know internet coordination tools in in a very new way where they're just getting spun up on the fly to solve specific problems and there's like these throwaway apps where we can throw them away.

Uh this is a great a great new example. Uh we had a poll request for uh some some like hardware changes on the drone and you know I was looking at this poll request and it's like okay you got to go like spin up a deployment or spin up a local a local environment and like actually go and look at this thing and it's all just like kind of random CAD files that's not easy to review. This was one prompt. So we just sent this to one of our agents and said put up on a new subdomain an app to review this pull request. and it gave us this beautiful like 3D uh diff viewer where you can go and see like in the PR anything that was changed in the in the 3D files. Uh so this is just kind of us like starting to

intelligently like weave in AI and intelligence throughout the organization you know anytime we feel friction or coordination problems or governance problems like how can we just like stitch in a little bit of help here and and build an app to to make this easier for us. And it's it's been really mind-blowing at at the way it's increased our pace and the way we can build and and coordinate with each other. Another fun thing we've seen is uh we we've spent years trying to get good at writing bounties and posting bounties and sending them out to the community. And and now we're finding that if you spend the effort to write a good bounty and you know, if you're giving a bounty out to someone that uh that you don't necessarily trust and and may not communicate well with you, like you need to have it like well scoped uh so

there's a clear deliverable. Otherwise, like they're going to build something and you're going to come back and say, "Oh, that's not what we wanted. We can't pay you." and then just nobody's happy. Uh, but what we're finding is is we write a good bounty and then people end up just like copyping that bounty into an AI tool and we're we're like prompting by proxy where where someone's just kind of like taking our bounty, feeding it to AI, send it back to us and then we get feedback and and it's just this whole loop and and many of these AI models are getting good enough that our previous bounties can just be dropped straight in anyways. So, so what's the point of putting a human on the other end?

So, so what we find in is the is the uh the challenging part of this isn't actually doing the work. It's deciding what work ought to be done. Um, and that's that's kind of the judgment like governance layer. And this is where I think Dows get quite interesting. Um, because now it's more of decision problem than it is like actually doing the work. So we're we're experimenting with uh different apps and ways to like stitch in discussion and governance and voting like anywhere that they can come up in the organization. So uh so yeah here you can see like a super app we're working on iterating now and we haven't solved all this but like this is this is just our attempts at solving the problem. Um so yeah it's like how do you crowdsource judgment and

that's that's kind of the game we're playing now. So, as we as we work towards this, we're bringing engineers all into the picture and and finding ways for them to like collaborate on good decisions together. And there's lots of decisions when you build aircraft, you know, u down to PCB specs and big things like what does your wing look like? Uh, but we're getting like reasonably good at actually doing the design side of things. And then of course the hard part is is going from online to the real world. And this is kind of what we're working towards now uh with Quiver. kilogram drone there. Uh we we're building kind of a network of manufacturers and workshops in our ecosystem. So we've been uh this is here in Texas. We're building building some

we've been building them in uh Germany and Turkey as well. And kind of what we're working towards is this idea where uh all the design happens online within the DAO ecosystem and then people can come in and join as as a manufacturing node and actually build the whole aircraft. Um and then the DAO will work with them and bring them into the world and commission flows back in fun more design work. And that's that's hopefully the flywheel that we're working towards. We haven't done it yet, but that's that's the idea. And we really hope that that other people like kind of see this and want to join the idea space and uh figure out how open communities can build hardware.

Uh so we've done it in some sense, but there's there's still a lot to figure out. Uh so lessons I' I'd want to like put forward to anyone that's trying to build Dows doing hardware. Um for us, I think one of the things we did right was taking it slow and figuring out what works. Um, you know, there there were times where we could have like really tried to blitz scale the community and like throw more throw more funding at the problem and I think that would have been been a very bad thing had we done that in the past. Um, don't build a large aircraft that crashes. That's that's an important lesson. Um, start start small. Start with something that you can iterate on and find an iteration cycle. Hardware is slow. Uh, so for us that was that was like a big thing that

we learned. Uh and then just like try to weave intelligence to the organization. Um you know the internet ultimately is just like for coordination. You know any app you can build is just going to hopefully improve your coordination skills. Uh so we're finding that things like people using personal agents to interface with organization. Uh little custom apps to make decisions and and try to like solve through engineering problems are a lot more useful than just like getting on a call and putting on a PowerPoint. Um you know we're finding a lot of a lot of value there. Uh, so yeah, that's that's ultimately the conclusion that I'm coming to is is that um this is a warm audience for Dows, but many people have have given up on Dows and pivoted into other things.

But I'm I'm truly very optimistic on Dows that we're going to hit a resurgence. Uh many Dows, I think, kind of fall victim to something where either it's infighting and everyone just like gets angry and the whole thing falls apart. And I think more commonly it's just like fatigue. people. You get a few people on the core team that are like pushing governance along and everyone else just like can't show up to vote constantly and and it just kind of like it falls off. Uh but I'm really optimistic that you know all these little governance tools we can create uh just like we are here at Arrow are going to make it a lot easier for a Dow to come alive.

Uh so I'm very bullish on on how that goes. Uh I know for us we're we're kind of building into this very optimistic future where um we're starting with kind of local first problems. So we have uh a company in Texas in our ecosystem that found uh an agricultural kind of land conservation use case where we're using our drones to uh dispense herbicide on overpop populated plants. Uh, and then we're kind of ringing the community around things like that and uh, and everyone comes in and like sees it and and maybe see something in their own community or their own region that we can do something there too and there's like some slight changes and the community comes in and really helps out.

Uh, so I'm just really optimistic on this idea that that we start with like wonderful local economies that are solving local problems and then bring internet scale into it, internet coordination into it. Uh so yeah, design design globally, build locally uh is is really what we're working towards here. Uh very human first, AI being being the substrate of the organization. Um and then building something together that's that's just greater than the the sum of its parts. Uh so yeah, that's Arrow. This this is what we're working towards. Uh getting into bigger aircraft um you know, really just trying to surf the human spirit. Uh Ethereum and all of this, it's like it's all just uh about coordination in the end. It's all coordination in the end. uh economics are about allocating scarce resources

and for us that's um that's you know engineer time and hardware cost and and trying to figure out uh how one person with a small workshop can find other people with small workshops and build something that none of them could do on their own. Uh so that's what we're doing. Thank you all for being here. A warm round of applause for Thomas. Quite honestly, as an expoing aerospace engineer myself, this was one of the talks I was really looking forward to.

There are a couple of questions. So, let's redide let's head right in. What is harder to coordinate? A community around code or community around an aircraft that has to fly safely? Um, I mean, I'm biased, but I would certainly argue a community around aircraft that has to fly safely. Um, you know, one one small mistake can crash an aircraft where one small mistake in code will just kind of give you an error and you can fix it. So, uh, that's certainly a big issue. >> Well, you have to handle both. So, it's a couple of balls that you have to keep in the air, >> of course.

>> Who would be accountable with something built by the DAO goes wrong? And how do regulators and insurers see that? So this is something that I didn't get into in the talk but we actually spent a lot of time in our whole iteration game looking at different legal structures and how we can structure the DAO. So so first attempt was like we wanted the DAO to just hold everything and have DAO like kind of owned workshops around the world and we quickly realized that that was a bit impractical. Uh and we've shifted into this thing where we have companies within our ecosystem. So the Dow is mostly kind of living online in the metaverse doing doing design and and living in Git repositories and things and then companies in the ecosystem that actually go and manufacture the the products and the aircraft are responsible for compliance and insurance

and you know making sure the product that actually ships off the assembly line uh is built safely with quality. So that means the actual parts of the airplane air aircraft have been manufactured by licensed companies and they have like all of the stickers that they need but the system itself being put together all of the different parts are still based on the design by the DAO. Does that make a difference? Because it's not just parts, it's a flying machine in the end. >> Exactly. Yeah. It's not clear where the money came from and why tokens cost anything and how liquid they are. Could you elaborate on this? So, if I understand the question correctly is whether the tokens buy

assets of the company or are just a gift or a grant to the company. >> Yeah. So, I won't get too much into the tokens uh right now, but ultimately they're they are like distributed to our members as they contribute to the DAO. Uh and it's it's useful for participating in governance, making good decisions together. uh the initial funding was I came in from a donation uh for myself when we started um and then we we want that to be part of the ecosystem where those manufacturers they're participating in the network are bonding tokens to kind of guarantee that they're aligned with the DAO and you know going to go through all the effort to actually actually sell this thing and and deliver.

>> So the tokens are not funding tokens they are participation tokens and funding is separate from that. Correct. >> Correct. >> Thank you. How expensive it was it to carry out all the DAO proposal voting process on chain? >> So I mean if you're actually talking about the mechanism of of like how those token votes go, we we've been using snapshots so far instead of pure onchain voting. So uh so far it's been been free financially and then quite expensive and and the time it takes just to like wrangle people and get them to like sign their wallets and you know multi transactions and the like.

>> Thank you. Ukraine is a huge drone company. So both in designing and manufacturing right now. Do you exchange engineer solutions with them or do you learn from them? >> So there there's actually like quite a lot different in civil drones compared to military drones and and we haven't done anything closely with the Ukrainians and as a global organization it's it's like kind of in our interest to not get too involved in in conflicts like that. Um, but I'm very I'm very much looking forward to Ukraine, especially like, you know, God willing, once the war is over and bringing all their their skills and expertise into the the civilian market.

>> Definitely. And I can second that. It's a beautiful country. I lived there for two years of my life. So once the war is over, it's a place to visit. >> Absolutely. >> What are your bottlenecks? What do you need to know? Uh what do you now need more than anything else to make progress as fast as possible? That's a really good question. Um, you know, if you asked me that a few months ago, I'd probably say like like skilled engineers uh to come in and just like help us figure things out. Uh, but it's reaching the point where like even just like individual contributors in a workshop are like building entire aircraft uh in our community, which is which is quite impressive. Uh, so it it's starting to feel like what we really want is people to come in and take our designs and find local problems

and and join our ecosystem and actually actually use what we're building. I think that's that's what we really need for progress. Now, >> let's try one last question. Is it possible to have a Boeing type of organization at a large scale through this st? >> I mean, it's it's not going to look the same. Uh Boeing is is a very large corporation. They have, you know, tons of engineers and big manufacturing and um I don't know if a Dow could build a 747. That's that's just like huge, massive, expensive project. uh we're doing something that's much more decentralized. Like I think you can't have many different factories throughout the world that build like jet ainers, but I think you can have many factories that can build, you know, like a four- seat air taxi type aircraft and that's

that's what we want to do. >> Thank you for that answer. Thank you for the great talk and as you see there are a lot of more questions. So people will probably come come to you to ask more questions right after the show. So thank you very much >> Thomas. Thank you also for being here today. This was the last talk of the day.