Since early 2022, we have engaged in on-and-off discussions about the idea of L3s, which were made possible through the concept of fractal scaling on zk-rollups. Although they were a pipedream back then, they have now become a reality. With networks like Katana and Madara, as well as open source provers coming online, we can truly build hyper-scaled games that allow for extremely low fees and fast transactions.
The path we envision involves a future of Starknet L2 and many L3s, applied where it makes sense. After all, the world we are building is not just a singular monolith but rather hundreds of shared stories and narratives, all woven together with a shared idea.
The first game released by the DAO will be "Loot Survivor," an extreme case of optimization to allow for deployment onto L2 immediately. "Loot Survivor" is very much a siloed game experience for now, built with an open and forkable codebase. It exists on the mainnet as it embodies the purest idea of an immutable game. Other games will be deployed on L2s when it makes sense, and their design does not require high-frequency transactions or limit the computational complexity. In fact, we envision a world contract built with Dojo being deployed onto mainnet Starknet very soon to allow anyone to build and extend the Realms.World. This is expected in Q4 of 2023.
Loot Survivor is undergoing final checks before deployment on mainnet - where it will run forever as an immutable arcade, played by inserting $Lords tokens.
Layer 3s can be defined as anything that settles a shared state back to L2, just as an L2 does to an L1. These could be bespoke Madara networks, bespoke Katanas, or even client-side Starknet state channels.
Realms.World is a network first, games and applications second. It's non-trivial to build onchain games, but now, with frameworks like Dojo, this has changed dramatically, as well as our understanding of how to design scalable experiences.
Depending on the game, different solutions are available and members of the DAO and third party builders are pursuing where they see fit.
NOTE: This is our interpretation of how we see Realms.World playing out using the Starknet Stack. We are not implying this is the only way to build onchain games.
Layer 2 will be used primarily for slower-paced games and as a Data Availability (DA) layer, much like how L1 is used as a DA layer for L2. Some high-level notes:
Thanks to recursive STARKs, we can run custom Starknets and settle their state to L2, achieving hyper-scalability. It's important to note that these networks can be entirely reconstructed from Starknet state.