Applied Engineering Desk
The engineering desk builds with these tools daily — agents, RAG pipelines, MCP servers and AI coding assistants — and writes up what survives contact with a production codebase.
7 articles published
The Model Context Protocol turns bespoke AI integrations into a standard interface. Here is the architecture, the three primitives, the handshake, the security model, and a working server.
Cursor and Windsurf are both VS Code forks with agents built in, and they solve the same problem with opposite philosophies. Here is which one fits which developer.
Everything between your first API call and a service you can put in front of users: key handling, streaming, structured outputs, tool calling, retries and token accounting.
A vector database is an approximate nearest neighbor index with a storage engine attached. Understanding the index type and the filtering strategy explains almost every performance difference between them.
RAG puts the right documents in the model's context at query time. The generation half is mostly solved; the retrieval half is where nearly every production system breaks.
Claude Code puts an agent in your terminal that can read, edit and test a real codebase. Here is how it works, how to configure it safely, and how to use it on real work.
Prompt engineering is specification writing, not incantation. This guide covers the structure that works, the techniques that are overrated, and how to test prompts like code.
The AI signal without the hype — new models, tools worth your time and what actually shipped. One email, no vendor pitches.