Agents as Software: A Programming Languages Agenda for Agent Reliability

2026-09-26

AI Summary

This study addresses the behavioral divergence of AI agents and the inadequacy of conventional testing in ensuring their reliability by proposing a programming-language perspective that treats agents as programmable artifacts, systematically restructuring their reliability framework. Methodologically, rather than pursuing deterministic execution, this work adopts a structured management paradigm. It employs trajectory- and state-based behavioral specifications to define expected outcomes, integrating pre-deployment static analysis with runtime dynamic monitoring for full-lifecycle governance. Consequently, this approach renders agent behavior both reason-able and controllable while supporting continuous iterative repair from operational failures. Ultimately, this research offers a novel pathway for constructing highly reliable AI agent systems.

Abstract

AI agents increasingly resemble software systems: they call tools, remember facts, follow policies, delegate work, and take actions with real consequences. % Yet the ``program''of an agent is scattered across prompts, tools, memories, workflows, and execution traces, making its behavior difficult to inspect through ordinary testing and debugging alone. % This essay argues that a programming-systems perspective offers a natural lens for making agents reliable. % We recast agents as programmable artifacts whose behavior can be specified over traces and state, checked before deployment, monitored during execution, and improved from observed failures. % The goal is not to make probabilistic agents behave like deterministic programs, but to give them enough structure that their behavior can be reasoned about, controlled, and repaired.

Problem

• AI agents
• reliability
• programming languages
• software systems
• behavior inspection

Innovation

• AI Agents
• Programming Languages
• Agent Reliability
• Formal Specification
• Programmable Artifacts