The AI Alliance, formed in late 2023 to advance AI, has been working on projects ranging from Dana, an agent-native language and runtime, to Semiont, a knowledge base for human/agent collaboration.
Some projects emerging from the alliance were covered in a presentation September 17 by Anthony Annunziata, AI Alliance director of AI open innovation, at The AI Conference in San Francisco. With Dana (Domain-Aware Neurosymbolic Agent), formally announced in late-June, users get AI-powered programming with intent-driven development, where developers can describe what they want to build and the language will handle the implementation. Dana features native support for agent workflows, memory grounding, and concurrency. Designed to run locally or in the cloud, the language is designed around domain-specific knowledge from human expertise and industry-specific workflows. It leverages large language models (LLMs) with symbolic grounding for deterministic, reliable outputs.
Semiont, meanwhile, enables humans and agents to co-create shared knowledge. Offering an AI-native wiki, Semiont features high-accuracy context retrieval, according to AI Alliance. Integrating via Model Context Protocol (MCP), Semiont enables locally owned knowledge bases, which are deployable on demand. Other projects from the alliance include Open Trusted Data for AI models and AI agents, as well as the Deep Research agent architecture with MCP. The Open Trusted Data effort involves a metadata specification for provenance, lineage, and utility, as well as a catalog of existing open data sets based on a trust score. The efforts features the curation and creation of new data sets for agentic AI use cases. The goal of Deep Research is to explore the challenges of building production-quality agents accessing data and tools only through MCP servers. The project also intends to release a reference implementation.
