This article discusses how AI coding tools struggle with legacy code due to missing context and institutional knowledge. It highlights the productivity challenges faced by engineers when using AI on outdated systems compared to new projects. The piece also outlines strategies for improving AI effectiveness through better documentation and restructuring.

This article discusses how AI will reshape engineering by enhancing prototype development, improving documentation quality, and increasing compliance focus. It emphasizes the need for strong data practices as engineers leverage AI to streamline workflows and tackle complex challenges more efficiently.

The article discusses why large software systems are often poorly understood, even by their creators. It highlights the challenges of documenting complex features and the reliance on engineers' tacit knowledge to answer basic questions about the software's functionality. As software evolves, keeping track of these details becomes increasingly difficult.

MCP (Model Context Protocol) has gained significant attention as a standard for LLMs to interact with the world, but the author criticizes its implementation for lacking mature engineering practices, poor documentation, and questionable design choices. The article argues that the transport methods, particularly HTTP and SSE, are problematic and suggests that a more straightforward approach using WebSockets would be preferable.