Equilibrium Infra Bulletin #21: Avail Trinity, AO Computer Initial Specs, and Inco's "Confidentiality as a Service"
Equilibrium Labs builds the state-of-the-art of decentralized infrastructure. We are a global team of ~30 people who tackle challenges around security, privacy, and scaling.
Research, Articles and Industry News:
📚 The Avail Vision: Accelerating the Unification of Web3 - recommended by Hannes:
Avail recently raised a $27m seed round and released details around Avail Trinity - the endgame for Avail which consists of:
Data Publication (Availability) - The data publication (DP) layer is the core of Avail’s offering and allows rollups to reduce costs by publishing data blobs to Avail rather than Ethereum. Avail leverages erasure coding, KZG commitments, and data availability sampling (DAS) to enable light clients to check probabilistically whether the data was published correctly.
Nexus - A proof aggregation and sequencer selection component which in itself is a ZK (validity) rollup on top of Avail that acts as a coordination hub between other rollups. Nexus enables cross-rollup and cross-ecosystem settlement, hence expanding on attempts of many L2s to create isolated interoperability within their ecosystem. The aggregated proof is submitted periodically to Ethereum - allowing all rollups participating in Nexus to share the verification costs on Ethereum by posting only a single aggregated proof.
Fusion - Shared security with a basket of high-quality tokens (ETH, BTC, SOL, etc.) alongside Avail’s native token that rollups can tap into. In the future, new rollup tokens get additional utility by playing a role in securing the Avail base layer (although limited to a small percentage of the total stake).
This announcement sets Avail apart from other data publication solutions and enables them to provide a more holistic offering with additional services for rollups leveraging the DP network. It also seems that Avail is doubling down on a future dominated by ZKPs, with validity proofs for correct erasure coding (relevant for DAS) and proof aggregation (relevant for cross-rollup interaction).
Key Takeaway: Avail is expanding from being just a data publication layer to a more comprehensive offering with shared security and rollup interoperability as well. Fusion draws inspiration from EigenLayer, IBC mesh security, and Babylon, whereas Nexus has similarities to Near’s fast finality layer, Polygon’s AggLayer, and zkSync Hyperchains. All this is tied together by the Avail token which is used to pay for short-term storage and bridging, as well as for helping secure the DP network, Nexus interoperability layer, and some rollups building on top of Avail.
📚 Protocol Specification of the ao Computer - Recommended by Teemu:
The recently announced AO Computer is a highly scalable, distributed computing environment built on top of the decentralized storage network Arweave. It aims to enable trust-minimized and verifiable compute with high performance by leveraging horizontal scaling, parallel processing, and message passing (“holographic state”) instead of a single shared state. AO Computer can support any VM as long as it's a deterministic computation device (EVM, SVM, WASM...).
The decentralized operating system built on top of this distributed computing network (AO) is called AOS, which is compatible with the Lua programming language. Lua is a high-level language that’s intuitive to learn and is already used in popular games such as Roblox and Minecraft.
The AO Computer consists of three unit types - Messenger, Scheduler, and Compute - which form the foundation of the network (on top of Arweave):
Messenger unit (MU) - Responsible for relaying messages around the AO network and where users’ messages are sent (entry point to the network). Also tasked with finding the right scheduler.
Scheduler Unit (SU) - Ensures messages are properly sequenced and stored on Arweave, maintaining order for consistent replay and verification of message evaluations.
Compute Unit (CU) - Nodes that users and messenger units can use to calculate the state of processes. Handle computation, loading binary modules, and managing memory to ensure processes run with current data. The evaluation results are then returned to the MU for further message handling.
Key Takeaway: AO Computer builds on top of the Arweave network and tries to enable a highly performant and distributed computing network, while still enabling verification (a core benefit of blockchain networks). It’s designed from the ground up to focus on parallel execution rather than shared state and makes different tradeoffs to existing blockchain networks, but there isn’t much detail on how the Holographic State will work in practice. A key difference to existing high-throughput integrated blockchains is that AO can support storing large amounts of data (thanks to its integration with Arweave), which opens the door to running more sophisticated applications that require extensive data handling and computational resources. Examples of these include machine learning tasks and high-compute autonomous agents.
📚 Introducing Inco: The Modular Confidential Computing Network - recommended by Hannes:
Most blockchains today are fully transparent, which means individual users share sensitive data (such as account balance and transaction history) and most enterprises can’t use blockchains due to competitive concerns, regulatory requirements, and lack of confidentiality. Inco aims to change this with a confidential L1 blockchain powered by fully homomorphic encryption (FHE), which enables computation directly on encrypted data.
There are two main ways that Inco enables confidential computing:
Natively - the L1 blockchain powered by Zama’s fhEVM abstracts away the complexity and enables developers to quickly get up and running building confidential dApps in Solidity. Popular tooling from Ethereum, such as Metamask, Remix, and Hardhat, will also be supported.
Outsourced (Modular Confidential Compute) - Existing networks, such as Ethereum and other L1s/L2s, can leverage Inco for confidential storage, compute, and access control (“confidentiality as a service”). This enables hybrid applications where most of the core logic lives on a public chain and only the parts that require storing or computing confidential information (e.g. cards, player stats, and hidden attacks) are done on Inco.
Bootstrapping an entirely new L1 blockchain with users, developers, and liquidity is a very different effort than integrating with existing networks. There are solutions, such as Elusiv and Namada, that enable bolt-on privacy for specific use cases (transactions and swaps). Inco takes a more general-purpose approach, which is interesting. The main risks for now seem to be around message-passing (bridging), which might leak information and is currently quite slow.
Key Takeaway: While we’re seeing more examples of blockchains that aim to combine programmability with privacy/confidentiality - both on the ZK side and FHE - we still believe more experimentation around granular information flow control is needed to enable several use cases to work on blockchains. Inco’s approach as a modular confidential computing network is interesting and a direction worth exploring, yet questions remain about how secure the message passing is and how much information it can leak.
Personal Recommendations From Our Team:
📚 Reading: Comfortable Feeling Uncomfortable: This blog post from the CEO of the Finnish game developer Supercell contains lots of interesting nuggets about how to inspire creativity in an organization and ways to create a process around game development. Many of these topics and questions are relevant to other fields as well.
🎧 Listening: Windows95man - No Rules: Finland’s entry to the Eurovision follows a similar pattern to last year - we’ll let you be the judge of whether it’s good or not 🫡
💡 Other: Intuitive Machines Lands on the Moon: Odysseus, a robotic lander built by Houston-based company Intuitive Machines, is the first private spacecraft to land on Earth's nearest neighbor. It’s also the first American vehicle on the moon since 1972.