Ideas for an Agent-Oriented Programming Language (davi.sh)

A new programming language idea, named Markov, has emerged to optimize software development by leveraging the unique characteristics of large language models (LLMs). As LLMs like Claude Code demonstrate impressive capabilities in navigating large codebases, the Markov language aims to capitalize on their strengths by offering a syntax that is both human-readable and LLM-friendly. This alignment seeks to streamline the coding process, allowing agents to generate boilerplate code while humans focus on core business logic, enhancing productivity and code quality. Key features of Markov include strong static typing and exhaustive pattern matching, inspired by successful components from languages like Rust. This design ensures that errors are less likely to propagate through the system, a crucial factor as coding agents increasingly manage end-to-end system development. Moreover, token-optimized syntax will make the code cheaper and easier for LLMs to process, potentially enhancing both performance and readability. The initiative reflects a growing recognition of the need for programming environments that naturally integrate human insights while empowering AI-driven coding agents, paving the way for more efficient and transparent software development processes.