Show HN: ISON – Data format that uses 30-70% fewer tokens than JSON for LLMs (github.com)

The recent introduction of ISON (Interchange Simple Object Notation) marks a significant advancement in data formats tailored for Large Language Models (LLMs) and Agentic AI workflows. ISON is designed to be a minimal and token-efficient alternative to JSON, reducing token usage by 30-70% while maintaining human readability and compatibility with LLMs. This efficiency is crucial for applications where token constraints can hinder performance, such as multi-agent systems and retrieval-augmented generation (RAG) pipelines. The technical architecture of ISON leverages familiar tabular and relational schemas, making it intuitive for both machines and humans. It supports advanced features such as type annotations, references, and computed fields, facilitating complex data interactions without significantly increasing cognitive load. Moreover, ISON includes validation and schema/schema-compliance features across multiple programming languages, including JavaScript, Python, Rust, and C++. With support for ISONL streaming for large datasets, ISON presents a robust solution for optimizing data interchange in AI and LLM applications, promising enhanced context handling and processing efficiency in AI workflows.