NanoLang drops: machine-native language for agents

NanoLang signals the arrival of "machine-native" development where AI, not humans, is the primary user. It is a bold bet that the future of coding belongs to autonomous agents that can verify their own output. Mandatory shadow blocks for testing ensure LLM-generated code actually works before compilation, while prefix notation and ARC memory management eliminate syntactic and memory footguns for coding agents. Specialized MEMORY.md and spec.json files provide an "instant learning" path for LLMs within their context window, and transpilation to C ensures it is more than just a toy by offering native performance. The project demonstrates how LLMs will significantly reduce the friction of launching new niche languages.