What are the limitations of using a file-based agent workflow? Why do massive context windows tend to collapse? This week on the show, Mikiko Bazeley from MongoDB joins us to discuss agentic architecture and context engineering.

The RPython blog offers insights, tutorials, and updates on RPython, a subset of the Python programming language designed for writing interpreters and virtual machines. Developers can learn about RPython features, optimization techniques, and applications in building programming language implementations. Additionally, the blog covers topics such as JIT compilation, garbage collection, and static analysis using RPython, providing resources for compiler and language developers.

Real Python

A podcast episode featuring Mikiko Bazeley from MongoDB discussing the limitations of file-based agent workflows and why databases are often a better choice for agentic architectures. Topics include context rot in large context windows, strategies for context engineering, multi-agent state sharing and coordination, dataflow patterns, human-in-the-loop design, and the model vs. harness debate. Resources on 12-factor agents, vector search, and MongoDB MCP server are also highlighted.

Episode #295: Agentic Architecture: Why Files Aren't Always Enough – The Real Python Podcast