TL;DR:
The ASI Alliance —a consortium formed by SingularityNET, Fetch.ai and CUDOS— announced a strategic partnership with Matterhorn, creator of a native integrated development environment (IDE) for “vibecoding” built around artificial intelligence. The goal is to bring decentralized application (dApp) development with AI capabilities and security safeguards to ASI:Chain, the alliance’s own network.
AI-assisted code generation is becoming entrenched in blockchain development. Most available tools allow developers to create smart contracts through natural language instructions, but few offer real protection against the vulnerabilities that code can introduce. Unlike a buggy web application, a compromised smart contract can result in irreversible loss of funds.
To address potential security issues, Matterhorn introduces Vibe-Audit, a system that combines an AI model trained specifically for security with human review in the process, enabling contracts to be audited before deployment. The solution also incorporates pre-validated templates and real-time concurrency testing designed for MeTTa, ASI Alliance’s native programming language and the language of its chain. That language features a concurrent execution model more powerful than Solidity, though also less tolerant of errors.
The initial integration operates with ASI:Cloud for decentralized AI inference, replacing centralized infrastructure. Subsequent phases will incorporate ASI:One and Fetch.ai’s Z.AI models for blockchain-specific code generation, along with integration of ASI Wallet and full smart contract support for MeTTa.
Abhinav, founder of Matterhorn, noted that “everyone else is racing to ship code faster” and that this is “the wrong race.” Khellar Crawford, Chief Innovation Officer of SingularityNET, argues that the partnership is “the beginning of the AGI-era software stack“, where blockchain security and consumer-grade usability converge through native AI inference.
Matterhorn’s roadmap also includes a fine-tuning pipeline that feeds real developer usage data back into ASI’s models, with the aim of progressively improving the specialization and security of the environment. Target metrics for the first quarter after launch include one million model calls and 500 active compute instances.
