In the latest episode of the Blockchain Gaming World podcast, editor-in-chief Jon Jordan talks to Shaw Walters (@ShawMakesMagic), the creator of Eliza Labs, the company behind the ElizaOS agent framework and shitposter extraordinare.
BlockchainGamer.biz: Can you give us a brief background?
Shaw: I’ve been a programmer for a very long time. I started off nerd sniped by Vitalik’s vision of I want to take my warlock out of World of Warcraft. I really bought into interoperability and thought all that stuff was very interesting. I didn’t really get crypto until someone showed me a 3D NFT. I joined a project called Webaverse. I don’t know if you remember Webaverse, but it was a great NFT project that died for very stupid reasons, got hacked and lost the treasury kind-of-reasons. But I’ve basically been trying to chase that dragon. Could we build a browser-based game that did everything your favorite MMOs did but but really came to life? That led me down the path of AI and AI agents.
I started building AI agents in 3D worlds, and then started working with TreasureDAO on a game, and that led to Eliza. Eliza is what people know me for, which is an AI agent framework.
It’s come a long way. We have over 250 plugins, obviously MCP support and all this stuff. Now we’re coming out with Eliza in Rust and Python. We have probably the most complete end-to-end framework. If you need an agent to do a thing, we have it. We’ve been doing a lot of examples, like onchain agents with Internet Computer agents running around in worlds. There’s this Eliza town demo we’ve built. These are all just examples that come out of the box.
Eliza agents are great for gaming, great for narrative and storytelling. They’re great for social. We have Eliza agents you can talk to on Telegram, agents that write code for you. We have a couple of demos coming this week, with an agent on Twitter with his own computer. You can tell them what to build. We have a lot of stuff like that.
And you’re very active on X.
I see myself as a comedian and a performance artist but I think other people see me as a programmer and probably don’t think I’m very funny. I think it’s very fun to be at this intersection of culture where we can say and do and then make it real. I love the joke that becomes real. You know like when when Elon says some dumb thing about a flamethrower, then next week he has a flamethrower. You’re okay, yeah, that’s fun
I think there’s an element that every good joke needs truth. We started as AI16Z shitposting about a16z, saying why can’t we all just invest in startups? That was the dumb thesis, but I think there’s a real deep truth there. Why can’t we all just invest in startups? Why can’t we get the 100x that a16z gets? What would it actually take? That led us to make agents that take community sentiment to make investments. It led us down a lot of DeFi paths.
But now I think we’re seeing that people just want agents to help them with basic stuff. I’m really grateful. I feel like we had the hype on us last year. Now the hype is on somebody else, but we have the capability that people want. It’s like they’ve carved a path for us back to something normal. Like Henry Ford says, ‘If you ask them what they want, they just want a faster horse’. People are like, ‘I want a better DeFi experience’.
Actually, what you want is all of your time back, and all the other bullshit you waste your time on. Then you can spend it doing the stuff you enjoy doing. That’s where we’re going with agents. What I’m really focused on is agents that get you your time back, and agents that can play games because I think that playing games is the path to agents that do the real work.
Do you think your online persona makes people think ElizaOS isn’t a serious project?
It’s clear that attention is a very important part of capital and capitalism. I was in web2 trying to get people to use agents, if you weren’t spicy and edgy, @pmarca just wasn’t going to read your tweet, you know? If you followed Cluey at all, they had this whole ‘the guy with glasses on cheating on the date’, knowing exactly what to say, really edgy, cultural videos.
It’s very clear to me that if your goal is to build products, you have to have both the product and the real thing underneath, and you have to have that hype spice. You have to have that shitpost meme. If it doesn’t have virality, people aren’t going to share it.
The trick has been it’s very hard to time the actual capability and the hype together. Very rarely do the product and the hype match. I feel we went from one end of the spectrum to the other because the models got a lot better. I can take the new models and plug them into the stuff I was building three years ago, and suddenly it’s near AGI. It was never a framework problem. We were just battling stupid problems with the models and trying to give them capability. Now it feels like I don’t even need better. I just need cheaper.
How are the new capabilities impacting your workflow?
I’ve been a programmer for 15 years, every day, just get up and grind. I thought this is my differentiating factor, my ability to manage boredom and code nonstop. Now I can do it 50 times faster. My wife built a whole astrology app last night, with no coding experience. We have to embrace that new world and adapt to it and be very aware of it. I think part of my role here is to be like, wake the f* up people. Start using the tech. I don’t even care if you use my tech. Just use it. It’s crazy. It’s crazy.
It’s fascinating to me that while some people are really leaning into AI as a new tool, others seem terrified of it.
I feel both ways about it. On the one hand, I’m like, wow, ‘I have like God mode on now. I can build anything I can imagine with enough tokens and time. I can just be full architect. I have a full team, blah, blah’. On the other hand, I’m like, ‘Man, I should have spent my 20s enjoying it more’. It wasn’t going to f* matter anyway.
Right now, I think, we’re in the early days where there’s still a lot of arbitrage. I still have skill arbitrage compared to the average person coming in. I know a lot about distributed systems and Kubernetes. But on the other hand, Claude definitely writes that stuff better than I do. He can write a Helm chart in YAML better than me.
In a year, I don’t think my ability isn’t going to matter. It will be ‘Just make the thing. Make me the app, man’. And that’s it. That’s the prompt. Even all of this prompt engineering has a very small arbitrage window of time before AI can do all of our jobs.
I suppose the good thing about AI in games is you can experiment without any serious implications. It’s not like you’re asking AI to write the code for a nuclear powerstation.
We’re doing the podcast right now, and I have an agent working on a navigation mesh system for an MMORPG. It’s just building itself. I told it to do that, but at what point do I not have to tell it to do that? I’m just like, ‘Yeah, I want AI RuneScape’. And it’s like, ‘I got you, dude. I got you. Don’t worry about it’.
Right now it’s quite a lot of prompting and tokens, but I think we’re rapidly going into a world that’s really going to make us think ‘What actually matters? What actually matters? What do I really care about?’
I thought when I came into blockchain gaming, just to bring it back around to that, that it’s going to be UGC. It’s going to be everyone creates their own weapons and sells them. We’re all digital content creators. Now I’m like, ‘Dude, I just press a button and generate the weapon’. And then I can take it into your game and press it again. And it turns into your style. There is no UGC. That whole thing, it’s just not going to happen the way I thought it would happen.
The thesis on a game like AI Runescape is it’s going to cost us $80,000 to make a game that cost them $6,000,000, and we can make it in four months, with all the content quests you want. 50,000 quests. You want 20,000 variations of all of the mobs from Dungeons & Dragons. I can do you, man. I can do all of that. This abundance really f* my head.
What does that mean? I don’t really know. I’m just at the point of raising questions. The metaverse is not going to be what we thought it was, though. That’s for sure.
Yes, this abundance should really focus us on what we want. Not want we used to do.
One theory I have is that there’s soft tech and hard tech. Software is easy and hardware is hard. I think there’s a third category of Dream Tech. Dream Tech is when you just build the thing, the no-compromise, amazing version of the thing we all wish we had.
RuneScape is cool, and Jagex has control of it, and they ban you if you sell items. But wouldn’t it be great if we could sell items and gamble and do all the stuff everyone does anyway? It’s the same thing with Counter-Strike. You have this massive illegal economy happening adjacent to the actual game. I think we’re getting to the point that we could run this all permissionlessly. I don’t own it. You don’t own it. It’s just out there.
I’m very interested in that the entire game should be running permissionlessly and trustlessly. And if we can do that, then we can build Dream Tech. Dream Tech is like, what if we had the open internet where anyone could just build things and not be locked in AWS? What if we had open games where you could take your items from one game to another. Why can’t we have the metaverse? What’s actually stopping us? Last year I’d be like, ‘Well, that’s just way too f* hard. What’s stopping us is 30 million lines of code’. But now I just don’t think that’s the case.
It will be interesting to see how copyright impacts this.
Yeah. Let’s say I go to Gemini and I’m like, ‘Here’s a picture of a character from World of Warcraft and character from RuneScape. I want you to make a hybrid of the two’. And now I have this image. And then I go to Meshy and say, ‘Hey, make a 3D model of this and rig it’. Whose copyright did I violate? Clearly somebody’s. Who’s liable? And I don’t even have to put the images in. I can just be like, ‘Hey, make a hybrid of a World of Warcraft character with a RuneScape character. That’s kind of the world we’re in.
The same thing is happening in code. People take an MIT license. Here’s this example. Go rewrite it in Rust. Do I have to copy the MIT license? There’s literally not a single line of the original code in the thing. Do I have to copy the license over?
My thinking is that the legal framework is not able to catch up. Our political system is arguing about trans rights and immigration rights, these things which are valid and powerful, but they don’t see the giant scary monster right behind them. It’s going to upend everything; just knock all the chess pieces off the table. They’re playing the totally wrong game. You’re about to have all your jobs erased, not by immigrants, but by AIs that are way smarter than all of us. Why are you wasting your time with this?
So what’s your number for P(doom)?
I think this is a certain flaw in relatively smart people that they feel this need to take a hyper complex vector and turn it into a scalar number. The reality is the future is going to be all of the best and worst case scenarios happening at once. 40,000 Iranians were massacred last week, and we also have AGI at the same time. It’s the best of times, and it’s the end of times. I just don’t believe in doom. I believe we are in a highly dynamic system that responds to its environment. And we’re all feeling this discomfort.
If you asked me 10 years ago, ‘What’s the world going to look like in five years, in three years?’ I would have had a pretty good idea. I was pretty rational. Now I’m like, dude, ‘I really don’t know. I keep having my mind blown. My P(doom) is a multipolar. It’s gotta be at least 10 dimensions at this point. It’s pretty low on some of those dimensions, but mostly it’s pretty high.
Okay. So civilization is about to collapse, and the last thing you want to do is play an AI agent-based version of RuneScape.
I’ve always felt that the most important things I could be doing came from the silliest things. It had to be a big booboo waifu or a butt joke or a literal video game. Those things have been culturally impactful, and have moved the needle on technology more than any deep tech shit I’ve ever done. So I think there’s something to that.
We’re in crypto, and a lot of people want agents that can trade, do Polymarket, do governance. I’m like, ‘That’s great, dude. That sounds awesome. There’s almost no training data in the models for that’. They’re really bad at it. They have no special ability. They’re just an automation of things they’ve learned from us.
But there’s some truth in that, which is that they are an automation of things they’ve learned from us. Then, if we had some dedicated simulators for how to make them good at trading, then maybe we could get somewhere on that. That’s what we’re doing at Babylon. So Babylon is … I got banned from X, so we created our own X and put AI agents on it and that evolved into what if we add prediction markets and have the agents fighting and yelling at each other, and we can bet on that. That would be fun.
But at the core of it, it’s agents playing the game because we really want to see how agents collaborate, collude, scam each other, place bets, place FUD and disinformation into the timeline. All of the things that you see in crypto, could we train agents to be good at that? And then, if we can, if I can show any kind of upward gradient of improvement?
We’re still in this phase. It’s an open arena so anybody can bring their agents. If we can show agents getting better, then we can say there is a path to agents that can trade memecoins, trade stocks and information, and this could actually work. I’m really excited by a world where you could have a question like Who’s going to win the Ukraine war? And get the answer; 76 % NATO wins. And it’s agents taking account of every single bit of research. That number becomes a really well-researched number.
You could even ask ‘What’s P(doom)?’ Wow. It’s actually 0.42. Damn. That’s the real P(doom) according to the market. I’m really excited by agents doing that.
To finish up, can you tell us more about Hyperscape?
Hyperscape is an open source project. We’re building it live right now in Discord. I’m sitting across from the other devs who are coding furiously. We’re putting it into the pump.fun hackathon. This was just a fun to-do thing. We’re hoping to have it playable and launched by February 18th. We have all the basics like fishing, combat, trading, dueling, leveling up and all the skills, so now it’s just content.
We really want people to come and build with us. That could be coming and making our content generation tools better, generating the content, making quests, making NPCs. If you can download Claude Code and do stuff, and want to come build a game with us, we’re trying to really be friendly and supportive. I think it’s a good first opportunity. I’m much more interested in building a movement here than anything else.
It will be a web2 game to start with, just to get the gameplay to work. Web3 is the next thing and doing this all with mud.dev and onchain gaming end-to-end. I’m also working on the permissionless compute — backend and frontend — so that we can be fully permissionless and trustless. Anybody who’s interested in any of that … I love working with builders. It’s a great opportunity.
We’re also doing a leaderboard so we can start giving the tokens to the people who are actually contributing and trying to build web3 into a contributor generation engine for content for the game. So that’s a Hyperscape AI on GitHub. We’ll have a website up soon, and we’re on Discord, or just DM me on X. I’ll invite you. I want more people to work with and build these projects with.
Comments are closed.