Vibeleaderboard
← Back to Vibers
Native
Company

Native

github.com/withnative

1 Intel

Native builds an open-source ChatGPT Space powered by a user's Markdown vault, currently in research preview, with a Rust engine and MCP server published on GitHub. Its research post credited here, by Richard Ng, reports that agents querying a Markdown vault wrapped in SQLite used 40 times fewer tokens per fully correct answer than agents reading an ordinary Markdown wiki.

Intel

Markdown is 40× more token-efficient with SQLite(www.withnative.ai)

Native benchmarks agents querying 11,000 records stored as Markdown folders versus SQLite, finding roughly 40x fewer tokens per correct answer with SQLite, and notes strongly typed frontmatter helps Markdown.

AI Agents