Business applications often coordinate multiple steps that need to run reliably or wait for extended periods, such as customer onboarding, payment processing, or orchestrating large language model inference. These critical processes require completion despite temporary disruptions or system failures. Developers currently spend significant time implementing mechanisms to track progress, handle failures, and manage resources when waiting for external events, shifting focus from business logic to undifferentiated tasks.
At re:Invent 2025, Amazon Web Services (AWS) launched AWS Lambda durable functions, a new capability extending Lambda’s event-driven programming model with built-in capabilities to build fault-tolerant multi-step applications and AI workflows using familiar programming languages. At its core, durable functions are regular Lambda functions, so your development and operational processes for Lambda continue to apply. However, when you create a Lambda…