Li+ Language: A High-Level Programming Language for AI Agents

Li+ is a high-level programming language specifically engineered for AI agents, utilizing a concept called "prompt-native harness engineering." When an AI executes Li+, it transforms into an interactive compiler capable of autonomously handling various stages of software development. This includes managing issues, defining specifications, implementing code, overseeing continuous integration (CI), and orchestrating releases, all facilitated through dialogue with human users.

The core philosophy of Li+ is that correctness is determined solely by observable real-world behavior, rather than internal consistency or explanations. It operates on a layered execution model, where each layer is responsible for stabilizing outward behavior. These layers cover aspects such as the surface file model, evolution, task management, and operations (like branching, committing, and CI).

Unlike traditional prompt engineering, which focuses on single instructions, Li+ designs a governing structure for instructions over time, across sessions and multiple AI instances. It prioritizes tasks through its layered structure and employs CI-driven self-correction for verification. The language is designed to be tool-agnostic, compatible with various AI models (e.g., Claude, GPT, Gemini), version control systems, CI/CD pipelines, and issue trackers. The rules within Li+ files are intended for AI interpretation, aiming to minimize the human learning curve, allowing users to interact naturally while the AI manages the underlying structure. The latest release is v1.17.10.