Articles

Articles, benchmarks, and evidence from the Mosaix Format project.

Versione italiana

2026-09-22 benchmark

MCP benchmark: seven models, one format

Controlled comparison of an MCP server implementing the Mosaix Format against a monolithic file baseline. Seven LLMs, four corpus sizes, nine test dimensions. MCP answers score 4.8/5 vs monolith 3.3/5 — with two instructive exceptions.

2026-09-04 evidence

Why one note at a time

Empirical measurements on atomic, self-describing notes: token cost stays flat as the corpus grows, and local models on consumer hardware produce correct output where cloud models without the format fail.