How Programming Languages May Evolve in the AI Era

Summary

José Valim’s essay examines how programming languages and their communities may change if coding agents write most of the code. He argues that agents could lower the cost of building frameworks and help smaller ecosystems catch up, while also weakening the collaboration that causes shared ecosystems to form. Human-oriented syntax improvements may matter less to agents than stronger guarantees, so languages could put more emphasis on explicit types and intentions, static analysis, proofs, runtime enforcement, and empirical validation. Valim rejects the idea that agents will make programming languages and compilers unnecessary, arguing that architecture-independent representations and different computational models remain essential. For agentic tooling, he proposes exposing compiler information through queryable program databases rather than document- and position-oriented LSP interfaces. He also advocates runtime observability that lets agents inspect state, traces, and production behavior programmatically, with safe access to capabilities such as those provided by the Erlang VM. The essay presents these as evolving personal views, and notes that AI assisted with stylistic and grammatical editing.