Ritual is moving beyond the standalone Infernet oracle model toward a Layer 1 architecture where AI and other forms of specialized computation can be accessed directly from smart contracts. The shift turns capabilities that previously depended on external oracle infrastructure into native blockchain primitives, including model inference, internet access, scheduled execution and cryptographic verification.
According to the Ritual Foundation’s official architecture overview, Infernet served as an early proof of concept that ultimately exposed limitations in using conventional blockchains for heterogeneous AI workloads. Those lessons helped shape Ritual Chain, which is designed to give applications native access to computation that cannot efficiently be replicated across every validator.
EVM++ Brings AI Primitives Into the Chain
Ritual’s AI-focused Layer 1 architecture centers on EVM++, a backwards-compatible extension of the Ethereum Virtual Machine. EVM++ introduces specialized precompiles and sidecars that let Solidity applications invoke computational services without building a separate oracle stack for every workload. Ritual currently describes support for model inference, TEE execution, ZK proving and verification, chain abstraction and other specialized operations.
The architecture does not require every validator to execute large AI workloads directly inside the normal EVM. Dedicated sidecars handle resource-intensive computation while keeping core on-chain state comparatively lightweight, allowing inference and proof generation to scale alongside the execution layer. Ritual also supports multiple integrity mechanisms, including TEE attestations and zero-knowledge techniques, rather than imposing one verification model on every application.
Resource allocation is coordinated through Resonance, Ritual’s compute fee market. Resonance creates a two-sided marketplace between users requesting specialized computation and providers capable of supplying the required hardware, allowing different workloads to be priced according to their computational requirements rather than through a uniform execution model.
Ritual Expands Beyond the Infernet Oracle Model
The new architecture also incorporates autonomous execution directly into the protocol. Scheduled transactions can be configured to execute every block or when specified conditions are met, reducing reliance on external keeper networks for recurring actions. That functionality is particularly relevant for autonomous agents that need to monitor conditions and act without repeated human transactions.
Ritual’s current platform extends that model further with native precompiles for AI models, external internet access, scheduling and key management. The Foundation says its present architecture exposes 16 native precompiles to developers using familiar Solidity tooling, while sensitive operations can use encrypted TEE environments and cryptographic attestations.
For existing Infernet infrastructure, the transition is better understood as an architectural evolution than a simple shutdown of the earlier system. Ritual is taking the lessons from decentralized off-chain computation and moving access, pricing and verification closer to the Layer 1 itself, while specialized environments continue to perform workloads that are impractical inside conventional blockchain execution.
The immediate test is now developer adoption. Ritual’s technical advantage will depend on whether applications actually use its native AI, TEE, ZK and scheduling primitives in production, rather than merely having those capabilities available at the protocol layer.
