Topology
January, 2025
The idea of the world computer is to create a single programmable medium where code and data live and operate correctly forever without centralized intermediaries, enabling trustless and censorship-resistant interaction at global scale.
For the world computer, blockchains excel as decentralized hard drives, great at global persistence, but limited by consensus overhead, lack of application privacy and isolation from global network activities. Topology introduces two key components to enhance the world computer: Distributed Real-time Programs, or DRP, a peer-to-peer Internet protocol that acts as world computer’s RAM by enabling ephemeral real-time peer-to-peer computation among local participants, and Mandu, an orchestration layer that directs bidirectional traffic between this RAM and decentralized hard drives.
These protocols complement the current infrastructure, working in harmony with blockchains. Their integration enables developers to build applications with real-time interaction, no single point of failure, safety against Sybil attacks, and end-to-end encryption. With the addition of RAM-like structures, the world computer can then power unstoppable AAA-level games, HFT-speed order books, real-time AI agent collaborations, and enable entirely new classes of real-time decentralized applications.
Through these innovations, Topology reimagines the architecture of the world computer, unlocks its full capabilities, and advances the cypherpunk vision.
The world computer vision was originated from Ethereum, which imagined itself as a “public decentralized shared hard drive” [1]. Indeed, a public programmable blockchain works like a stateless CPU running on a hard drive that allows only sequential writes.
A decentralized hard drive typically exhibits the following properties:
Persistence: Global persistence is guaranteed. Blockchains typically offer the guarantee that any data sent onchain is persisted indefinitely.
Lock-heavy block time: Transactions are total ordered and batched into blocks by global consensus, acting as a single global lock. As a result, transaction latency is dominated by the overhead of this lock.
A single shared transaction log. All transactions related to any part of its state are collectively recorded in a single shared log, typically in the form a merkle tree. To trustlessly recreate the latest state of any particular smart contract requires syncing and verifying the entirety of this log.
Difficult to scale horizontally: Horizontally scaling a distributed system means to increase the system’s performance by adding more nodes to it. Blockchain consensus inherently requires global coordination among all nodes. Instead of improving performance, adding more nodes to a blockchain system can negatively impact its speed and throughput. This makes high-performance geo-distributed computation impractical for blockchain systems.
The primary limitation of these hard drives is to scale in decentralized ways. To achieve higher throughput and lower latency, blockchains would require both a smaller number of participating machines and more powerful hardware to efficiently process large blocks. which creates a centralization vector.
Multiple base layer blockchains in the recent years have pushed the performance boundary. However, they do not address the fundamental limitation of blockchains: they are subject to the trilemma where at a given security level, design approaches to achieve higher scalability necessarily reduce decentralization.
Moreover, public blockchains operate on transparent global state where all data and computation are publicly visible, making privacy preservation difficult without additional cryptographic techniques.
The modular blockchain architecture, and layer-2 rollups in particular, offer a path forward by separating the system into specialized layers: