In my previous post on Specification-Driven Development, I argued that managing an LLM's unpredictability requires discipline, and that a well-formed specification is the most powerful context you can provide. The specification defines the what. The engineer designs the verification strategy. The LLM handles the how. But there'
Specification-Driven Development gives us a powerful discipline: write the what before the how, and use that specification as a living contract that governs everything the AI agent does. But there is a subtle flaw in this picture if left unexamined. A spec is only as good as its quality, and
The Federated Context Challenge in AI Agent Coordination In context engineering, Semantic Divergence is the primary obstacle to multi-agent coordination. Take Vehicle-123, a standard delivery van. To the various AI agents reasoning about it, the vehicle exists across multiple independent contexts, and these contexts rarely align. This isn't
LLM-powered coding agents are introducing a new dimension to software development, where AI systems can interpret both our codified artifacts and human intent to generate consistent, compliant code. This emerging capability builds on our "Everything as Code" foundation, where we've systematized infrastructure, security, and pipelines, by
Large Language Models (LLMs) are changing how we write software. Their ability to generate code at incredible speeds is a game-changer, but this power comes with a critical challenge: unpredictability. Because LLMs are probabilistic by nature, we can't always guarantee their output is reliable or correct. The solution
Defining the Context Layer The context layer is the architectural component responsible for acquiring, organizing, maintaining, and delivering relevant information to AI systems at the appropriate time and granularity. Just as we have data layers that abstract database complexity and service layers that encapsulate business logic, the context layer manages