Fifteen years of the same click: what the agent era keeps rediscovering about distributed systems

In a recent reflection on modern software architecture, developer Pierre-Laurent Médori explores how current trends in AI agents and graph engineering often mirror foundational challenges in distributed systems. Through three specific examples—idempotent webhooks, reconciliation crons, and workflow orchestration—the author demonstrates that modern automation frequently encounters problems solved years ago in microservices architecture. Médori argues that while AI adds new layers of complexity, the underlying issues of state management, race conditions, and failure handling remain constant. He emphasizes that developers often spend significant time 'rediscovering' these mechanisms under new terminology. The article serves as a reminder that despite the rapid evolution of tools and the rise of autonomous agents, the core principles of robust system design, such as transactional integrity and independent verification, remain essential for building reliable, scalable software in any era.