The Evolution of Microservices: Agents, Monoliths, and the Patterns That Never Die

We recorded this conversation at the AWS Summit London with Matheus Guimaraes Matheus is a Senior Developer Advocate at AWS with 25 years in the industry — from programming Game Boy ports in Z80 Assembly to CTO, to leading digital transformations at scale. He's one of the sharpest minds on distributed systems I know, and in this episode, we go deep on how microservices are evolving in 2026. Matheus introduces his concept of "agentic microservices" — what happens when you embed a foundation model inside an API — and makes a compelling case that the same architectural debates we've had for over a decade (monolith vs. microservices, single responsibility, coupling, service discovery) are now replaying at the agent level. Context dilution, the "lost in the middle" problem, monolithic agents vs. specialized agents... sound familiar? We also cover modular monoliths (Spring Modulith), how AI coding assistants like Kiro enforce architectural governance, the serverless angle (Agent Core runtime vs. Lambda for agentic workloads), durable functions vs. Step Functions, and why the fundamental patterns of software architecture simply refuse to die. Whether you're building agentic applications, running microservices at scale, or debating your next architecture decision — this one's for you. ⏱ Timestamps 00:00 — Intro at AWS Summit London 02:30 — Who is Matheus Guimaraes? 05:00 — How to think about microservices in 2026 07:30 — Smart APIs: putting a foundation model in your microservice 09:00 — MCP, A2A, and the new integration layers 11:00 — Context dilution and the "lost in the middle" problem 13:00 — Monolithic vs. specialized agentic microservices 15:00 — The monolith vs. microservices debate in 2026 18:00 — Modular monoliths and Spring Modulith 21:00 — AI coding assistants (Kiro) and the architect's evolving role 25:00 — Serverless for agents: AgentCore Runtime vs. Lambda 28:00 — Lambda Durable Functions vs. Step Functions 31:00 — AWS DevOps Agent and autonomous operations 34:00 — Microservices are inevitable (in concept) 37:00 — "Is It Agent?" — when NOT to use agents 41:00 — Closing: critical thinking about AI adoption 43:00 — Book recommendations & who to interview next 🔗 RESOURCES & LINKS 📄 Martin Fowler — Microservices (2014): https://martinfowler.com/articles/mic... 📦 Spring Modulith: https://spring.io/projects/spring-mod... 📖 Sam Newman — Building Microservices: https://www.oreilly.com/library/view/... 🧠 AWS Bedrock Agent Core: https://aws.amazon.com/bedrock/agent-... ⚡ AWS Lambda: https://aws.amazon.com/lambda/ 🔄 AWS Step Functions: https://aws.amazon.com/step-functions/ 🔧 AWS Durable Functions: https://aws.amazon.com/blogs/aws/intr... 🔌 Model Context Protocol (MCP): https://modelcontextprotocol.io/ 🤝 Agent2Agent Protocol (A2A): https://google.github.io/A2A/ 💻 Amazon Q Developer: https://aws.amazon.com/q/developer/ 🛠️ Amazon Kiro: https://kiro.dev/ 🌐 Werner Vogels — All Things Distributed: https://www.allthingsdistributed.com/ 🎤 Morgan Willis — Architecting AI Applications    • We Need to Talk About AI Agent Architectures   👤 GUEST Matheus Guimaraes — Senior Developer Advocate, AWS LinkedIn:   / intmain