A local AWS re:Invent recap meetup in Montreal featured six ten-minute demos of concrete features, not marketing slides. Presenters showed how new tools change real system building, from AI-driven incident investigation to serverless orchestration. The format provided specific, practical insight without requiring deep AWS expertise, focusing on what each feature actually does.

AWS DevOps Agent, now in preview, acts as an AI operational assistant. It investigates incidents by parsing logs, mapping service dependencies, and suggesting root causes. AWS Transform offers a controlled AI migration pipeline, with built-in verification tests to ensure correctness. Air Canada and Twitch case studies saved thousands of developer-days, compressing years of technical debt cleanup into weeks.

SageMaker Studio is becoming a unified workspace for data and AI, integrating a built-in catalog, SQL querying, and serverless Airflow for ML pipelines. Durable Lambda adds native wait, checkpoint, and callback primitives, eliminating the need for external state stores like DynamoDB for long-running workflows. This simplifies architectures for AI tasks needing human approval or external callbacks.

Lambda now runs on AWS-managed EC2 capacity providers, eliminating cold starts and offering predictable performance with GPU support coming. For high-throughput workloads, this can slash costs dramatically compared to standard Lambda. Meanwhile, S3 Vectors bring affordable, durable vector storage to object-store scale, trading some latency for massive scalability in RAG and semantic search applications.