Mathematically Formalizing the TOON Efficiency Revolution versus JSON (mateolafalce.github.io)

Mateo Lafalce published a mathematical formalization that quantifies why TOON — a compact, line-oriented data format — is more token- and byte-efficient than standard JSON. Using RFC 8259 JSON examples and TOON's key:value and class-array shorthand (e.g., users[2]{id,name,role} followed by CSV lines), he derives recursive character-count functions for both formats: a weight function for JSON values (handling objects, arrays, strings, primitives) and analogous expressions for TOON’s simple and complex objects. From those formulas he isolates a per-attribute overhead difference (roughly two characters saved per key-value line in TOON) and gives a percentage-efficiency expression showing savings grow linearly with number of attributes. This matters for the AI/ML community because fewer characters directly reduce token usage, bandwidth, latency, and API costs when transmitting structured data to LLMs or data pipelines at scale. The formalization clarifies how TOON’s removal of JSON punctuation (quotes, braces, commas) and its compact class/CSV encoding turns per-record savings into large cumulative reductions. Lafalce also shared a vanilla-JS JSON→TOON converter so practitioners can measure real-world savings on their payloads and validate the derived formulas.