A Language For Agents
Summary
Armin Ronacher argues that the rise of agentic engineering will drive interest in new programming languages designed for AI agents, highlighting factors like tooling, model weight representation, and code generation aspects. He discusses desirable language traits (LSP-agnostic, readable syntax, greppable code, local reasoning) and what agents hate (macros, barrel files, flaky tests), concluding that the number of languages will grow as we optimize for agent-friendly design and measurable agent performance.