LLMs as the new high level language (federicopereiro.com)

Recent discussions in the AI/ML community propose that LLM agents might emerge as a new high-level programming language, performing a role similar to that of C or Java in relation to lower-level languages. The hypothesis suggests that developers could significantly increase their output—potentially tenfold—by utilizing multiple autonomous agents that require minimal human intervention for feedback. This shift could redefine how coding is approached, offering a more dynamic interaction with programming tools that combine documentation, implementation, and agent-based execution to streamline the development process. The implications of this hypothesis are substantial. LLMs are expected to introduce new efficiencies and workflows, fundamentally restructuring programming paradigms. While concerns exist regarding the quality and understandability of code generated by LLMs, proponents argue that, if managed correctly, these agents could facilitate a more productive code generation environment. By implementing standards like MCP for tool interactions, developers can access functionality from various applications, potentially breaking down existing silos. This vision not only promises to enhance the collaborative aspects of software development but also encourages a shift toward more efficient and user-friendly programming methods, making the future landscape of coding more interconnected and dynamic.