AI Engineering

AI Engineering here is not a collection of model demos. It is the work of putting language models, tools and retrieval into systems that already have architecture, constraints and owners. Most interesting failures I have seen did not come from the model being “not smart enough”. They came from unclear tool contracts, missing evaluation, context that looked complete and was not, or an integration that nobody could operate after the prototype. That is engineering work. This hub collects writing on MCP and agents, context and retrieval, vector search, and the infrastructure around AI-enabled products. The goal is practical: how to design the interface between a model and a system so that the change remains reviewable, observable and reversible. I care about the parts that survive contact with production. Tool design. Boundaries of authority. What the model is allowed to do. How you know it did the right thing. How a team keeps ownership when generation gets faster than review. If you are looking for a starting point, begin with the materials on retrieval and tool-calling. They sit at the junction of “the model can talk” and “the system can change”.