Nerd – An intermediate language for machines, not humans (github.com)

A new programming language called NERD has been introduced, designed specifically for machine-to-machine interactions and optimized for Large Language Models (LLMs). In a landscape where approximately 40% of code is now composed by LLMs and that figure is on the rise, NERD aims to enhance efficiency by employing a dense, human-observable, yet non-editable syntax. Unlike traditional languages, which prioritize human readability with verbose syntax, NERD utilizes simple English words to minimize token usage, achieving 50-70% fewer tokens for the same logic, thereby reducing costs and increasing execution speed. NERD's architecture allows machines to write and read code, compiling to native binaries through LLVM while maintaining human oversight for auditing purposes. It consciously avoids features like symbols and complex formatting, demanding no effort from humans to handle code directly—changes can be proposed in natural language, with machines adjusting the NERD codebase. This innovative approach rethinks programming language design in the AI era, positioning NERD as an experimental foundation that invites community contributions and engagement from those interested in the future of coding, compiler development, and language design.