The search for a digital promised land
The search for a digital promised land
Video games are about to get really weird
Happy Friday! Are you seeking a relationship with a human-like AI video game character? Or perhaps considering an overdue vacation to a digital planet? Look no further. Glitch is back.
In this issue:
Stuff that has us like 👀
NPCs that “remember”
Visa did an interesting blockchain thing
Can on-chain games ever be as good as their mainstream counterparts? Could they be more fun?
1. Two things that have us like 👀
Open source “AI agents”
An AI research project known as “Stanford Smallville,” which made waves in April with its surprisingly real-feeling digital world full of walking and talking “generative agents,” has open-sourced its code. Anyone can now copy or tweak the software, which used generative AI to make the video game-style characters seem creepily human.
Modeled on the EA Games classic The Sims, the dummy world tested the relationships and actions of characters, who were able to “remember” events and “reflect” on those experiences. Their “complex” intellects were fed by ChatGPT’s large language model. During the experiment, one autonomous character hosted a Valentine’s Day party for 24 others. An associated research paper delves into the “emergent social dynamics” of the party, which the authors say were reflected in its believable awkwardness.
Alongside arguments for potential practical applications of this software—such as the ability to rehearse a job interview, or a “difficult, conflict-laden communication”—the paper also acknowledges that such a powerful kind of simulation presents ethical challenges that have long plagued scientists and inspired sci-fi writers. The researchers write:
”We argue that these agents should be tuned to mitigate the risk of users forming parasocial relationships, logged to mitigate risks stemming from deepfakes and tailored persuasion, and applied in ways that complement rather than replace human stakeholders in design processes.”
This intervention seems optimistic, given how much closer this software brings us to a world where Joaquin Phoenix’s character in Her is no longer a futuristic dream. At the very least, video games are about to get really weird. —Lucy Harley-McKeown
Visa’s trip into the crypto weeds
Crypto is known for coming up with names for new things that can be, well, abstract. It starts with a familiar pattern: Somebody writes a paper about some new form of tech, (say, a novel smart contract), they name it and send it to their friends (other influential crypto nerds), and soon, if it’s useful enough, their friends start talking about it as if it's a well established concept. It spawns conference papers and maybe even new startups. Eventually, people act like it’s real for long enough that it becomes that way.
And then, naturally, Visa begins to use your new tech.
In a nutshell, that seems to be what happened with Ethereum creator Vitalik Buterin’s latest pet protocol: ERC-4337, a standard that outlines how people can use “smart contract wallets.”
Historically, Ethereum has had two types of accounts: smart contract accounts, which are controlled by decentralized computer programs, and “externally-owned” accounts, which are managed by people. The new standard outlines how to abstract away the distinction, and that process has a trendy new name: “account abstraction.”
So, WTF is account abstraction?
Briefly: it allows people to turn a crypto wallet into a smart contract. And why would you want to do that? Because then you can program it to do different things for you, including some of the things that can make today’s crypto user experience annoyingly challenging.
If you are still confused, you’re in luck, because Visa just made it this all less … abstract. (Sorry.)
Earlier this week, Visa’s crypto team announced that it had developed a way for cardholders to pay their gas fees, or the cost of using a blockchain, with their credit cards. That may sound pretty simple, but it is actually pretty complicated—and potentially useful, at least if the goal is to get more people using crypto networks.
The reason it’s complicated is that it can be technically challenging to get crypto rails and the traditional financial system to play nicely together. In part, that’s because crypto prices fluctuate all the time, so the value of a given transaction can change in the time that it takes for it to be processed.
Visa used account abstraction to address this issue. The main job of a crypto wallet is to manage the user’s cryptographic private keys, which represent their money and other assets. A smart contract can be programmed to handle those keys in specific ways, given specific circumstances.
What Visa did is make it possible to initiate “pull” transactions with crypto—think: your hydro provider automatically pulling money from your account—without the wallet holder needing to manually approve every transaction. The end result, says Visa, is that “the gas fee experience on the user’s end is as simple as a regular card payment.” You can read about the technical details here.
While this setup involves the risk that the price of crypto will drop in the time it takes for a transaction to be processed—a problem it is trying to hedge against by processing transactions in USDC, a “stablecoin” designed to keep a consistent price—Visa seems to be betting that if it makes it easier to do blockchain stuff using its credit cards, it will make money because people will transact more. Whether or not that happens, at least the logic is fairly concrete (ha ha!). —Sam Venis
2. In search of on-chain worlds—and a new species of fun
Once upon a time, there was a game called Dark Forest. And it was unlike any other game that had come before it.
It began as an audacious experiment hatched by three friends who wanted to see if they could build a game that existed entirely on a blockchain. Defying conventional wisdom (Blockchains aren’t good for complicated applications!), they pulled it off.
A 2-D multiplayer strategy game that takes place in space, Dark Forest’s gameplay was inspired by Galcon, a classic online multiplayer game in which the object is to conquer the galaxy, planet by planet.
Besides building the first complicated game that lived completely “on-chain”—encoded in smart contracts, or decentralized computer programs—the creators of Dark Forest also used advanced cryptography to implement the classic “fog of war” mechanic that keeps unexplored portions of a game’s map invisible until the player discovers them.
It didn’t take long after Dark Forest’s launch in 2020 for it to become a phenomenon within the crypto-sphere. The game was played in one-to-two-week rounds—the most recent of which took place in February of last year. More than 10,000 players competed in at least one round.
The game was crude, slow-moving, and required technical skills to play. But for a certain sect of nerds it was a revelation: it was not just a crypto game. It was fun—in a way they had never experienced before. On top of all that, the fact that this new world lived on the blockchain meant that no one owned it. No one could block people from using the game’s code to make new things. And because it lived on a blockchain, no one person or company could kill it (to a community of people who had run-ins with mainstream gaming companies in the past, this was a major selling point).
Some of these new converts became convinced they had discovered the future of online gaming. Now they are determined to bring more players into their on-chain world—if only they can make a game fun enough to reveal it in its true glory.
A “digital planet”
Will Robinson showed up so late to his first round that he had no chance of winning. He also realized that he was “terrible” because he didn’t have the right technical skills. So he began recruiting teammates, cold DMing people on Twitter that were near the top of Dark Forest’s public leaderboard.
Technical skills were required, says Robinson (who in addition to working for crypto startup accelerator Alliance holds a PhD in game studies), because in order to win you needed to be able to build and deploy new software inside the game. This quality, which he and other Dark Forest enthusiasts call “composability,” is part of what made the game feel so novel.
A normal multiplayer online game consists of a “backend” server, usually owned by a company like Blizzard or EA, and “frontend” software clients. The server is God, essentially. It’s in charge of maintaining the game’s history and providing each connected client—the software you get when you download a new game—with enough information to display a view of the world at any given moment to their respective players.
In Dark Forest, the blockchain took the place of the server. That meant that all of the game’s rules, history, and state were publicly available and tamper-resistant thanks to the security of the blockchain. Anyone who wanted to could write new software that used that information to gain an edge.
For example, although the creators released a client, once it was live players started building “plugins” that provided advantages, from useful new ways of viewing the map to automated attack mechanisms to in-game marketplaces. Some even went so far as to create whole new clients. “It’s the first game anybody’s ever seen that’s frontend agnostic,” says Robinson. This by itself explains why Dark Forest was like catnip for crypto developers.
There’s more to it than that, though. User-generated “mods” that make tweaks and sometimes even overhauls to established games have long been a fixture in gaming culture. A modern version is a custom Minecraft server. But Mojang, the company that runs Minecraft, has the power to shut these down, or make unilateral changes that effectively ruin the world (a power it exercised last year when it banned NFTs after one of those servers was found to have set up a successful NFT marketplace). When the rules for a world live on a public blockchain instead of a private server, it’s possible to design it to be “autonomous”—meaning no one can shut it down, says Justin Glibert, co-founder of the crypto research organization 0xPARC.
That would be more like a “digital planet” than a game, Glibert says: “more like Mars than World of Warcraft.”
Sure, but does it scale?
Glibert founded 0xPARC along with one of Dark Forest’s creators, who goes by the pseudonym Gubsheep. The organization, which is funded by grants and donations, aims to “push the boundaries of what is technically possible with crypto applications, both on and off blockchains.” A division called Lattice, which Glibert helms, is focused specifically on blockchain gaming.
Lattice is best known for MUD, a software architecture it’s developing that helps game developers build on the Ethereum blockchain. Words3, a Scrabble-like game that recently had a round on Coinbase’s rollup network, was built using MUD. So was Primodium, a factory-themed game that’s in the midst of a public playtest (whose creators happen to be the same people who were behind the custom Minecraft-with-NFTs world that Mojang killed). Dozens of other developers are hacking away at game prototypes using MUD.
But while Lattice’s vision for “autonomous worlds” is a cool idea—one that might even have mainstream appeal—we’re a long way from blockchain-based games as advanced in their gameplay as traditional massively multiplayer online (MMO) games.
One of the most glaring issues is that blockchains are simply not good at handling the volume of moves typically associated with MMOs. Emerging “layer 2” approaches like rollups, which use various technologies to process a batch of transactions before sending the bundle to the blockchain, can help. But the computing infrastructure still pales in comparison to that in the traditional game world.
Dark Forest could handle as many as 800 players before all their moves started “clogging” the layer 2 network it was using, says Robinson. (Today’s successful MMOs tend to have at least tens of thousands of concurrent players.) Plus, on the blockchain, when space becomes scarce it becomes expensive. So even the hundreds of players it could support had to pay a pretty penny to make moves.
Perhaps future decentralized computers will be able to handle larger throughputs. In the meantime, it might make sense to narrow the design space, argues Robinson, who has written a series of blog posts focused on how on-chain game designers and developers might overcome technical limits to scaling. Game makers should consider capping the number of player moves, he suggests, or perhaps design games around competitions that involve a small number of professional or high-stakes players. Or he says blockchain game designers could take inspiration from complicated board games where players expect there to be long pauses in between turns.
On the other hand, in some cases it may not be necessary that every single move be a blockchain transaction. For example, while millions of people can play World of Warcraft concurrently, Robinson points out that much of the core gameplay involves smaller groups of players fighting monsters in caves and dungeons, using their own server instances. The shared part of the world is slower-paced. It’s where players return from the dungeons to develop their characters and show off and trade the items they won fighting.
A World of Warcraft-inspired on-chain game could be designed so that combat occurs off-chain, Robinson says, and then players could return to the shared space of the game by sending a zero-knowledge proof to the blockchain to verify, for example, that they beat a dungeon boss and earned a powerful new shield—and did so in compliance with the game’s rules. The team behind Dojo, a MUD-like framework for developers to build on a rollup called Starknet, is working on making something like this practical.
A new kind of fun
Scott Sunarto, one of the original creators of Dark Forest, has a different idea. Sunarto, who also worked for Lattice, has founded a company called Argus to develop a “custom rollup” that he says is designed specifically for games. Sunarto says he decided to focus on this goal after experiencing the “pain points” of encoding a game in Ethereum smart contracts.
Thanks to computation that Argus’s rollup executes on its own, games running on it can scale “horizontally,” similar to the way games like World of Warcraft do, Sunarto says. He says this also makes possible other things we take for granted in traditional games but which will be very technically challenging to emulate in the blockchain context—like non-player characters and continuous, real-time action (as opposed to turn-based games like Dark Forest). That expands the space for designers to break free of the Dark Forest mold and create more complex and exciting games, says Sunarto.
Argus recently raised $10 million in a seed funding round. Another startup, called Curio, is taking a similar approach, and has raised nearly $3 million.
But if certain aspects of the game are not encoded in the blockchain, is that still the same thing as Lattice’s autonomous world? If not, does that matter?
Without any clear examples of the ultimate prize, that’s still up for technical and philosophical debate. “It’s hard to put your finger on how much decentralization is necessary,” says Tarrence van As, cofounder and CEO of gaming startup Cartridge and a lead developer of Dojo.
It’s also not necessarily true that the killer app is a complicated game that looks and feels like what we’re used to. Simple games can be just as fun as complicated ones. “Just make a single-player game to start,” suggests Robinson. “Everybody’s decided to make the hardest possible game first.”
Dark Forest’s gameplay mattered. But what made the game so fun was what its true believers call the “emergent” behavior possible only because it was built on the blockchain.
Though their approaches vary, everyone now trying to build on-chain games is hoping to set the stage for a similar sort of phenomena, convinced that if they succeed it will be so much fun that even “normies” will see the light. —Mike Orcutt