Jaroslaw Wasowski argues that many companies are moving away from microservices and back toward larger, more organized systems. He says microservices were mainly created to help human teams handle complexity, not because they are always technically better.

In the age of AI coding agents, this changes because agents need clear context to understand and change code well. A modular monolith can help because it keeps the whole domain in one place while still having strict internal boundaries.

Wasowski explains that many companies actually have distributed monoliths, where services are separate but still depend heavily on each other. This creates high costs, slow releases, and complicated testing without giving the real benefits of microservices. He also points to examples like Prime Video, where moving back to a monolithic architecture reduced infrastructure costs strongly.

According to Wasowski, microservices still make sense for companies with mature developer platforms, independent domains, special scaling needs, and dedicated platform teams.
If these conditions are missing, a modular monolith is usually cheaper, safer, and easier for AI agents to work with.
Overall, he recommends starting with a modular monolith and only extracting microservices when there is a proven need.

Full article on Medium.com:
https://levelup.gitconnected.com/modular-monolith-instead-of-microservices-what-changed-when-the-ai-agent-started-reading-code-c586d9f63fd7