Technical writing on AI coding agents, context engineering, and developer productivity.
MCP (Model Context Protocol) servers extend AI coding agents with new tools and data sources. They're the plugin system that makes agents truly powerful.
Three approaches to code indexing for AI: embeddings, dependency graphs, and RAG. Each has trade-offs in accuracy, token efficiency, and maintenance cost.
RAG retrieves relevant code from your codebase before the AI generates a response. But vector-based RAG misses structural relationships that matter for coding.
Loading more files into the context window doesn't improve AI output — it degrades it. Quality context with 5 relevant files beats 50 random ones every time.
Copilot's wrong suggestions aren't a model problem — it's a context problem. The AI can't see enough of your codebase to make accurate recommendations.
At $39/user/month, Copilot Enterprise costs add up fast. Most organizations waste 60-70% of their token budget on irrelevant context per developer.
Copilot Agent Mode turns autocomplete into autonomous coding. It works best with focused context — here's how to set it up and optimize it.
Copilot wastes tokens by sending unfocused context with every request. Optimize what it sees and you get better suggestions using fewer premium requests.
Windsurf slows down on large projects as its context engine struggles with scale. External dependency graphs solve this by pre-computing code relationships.