Articles

A RAG chatbot that answers questions about your Obsidian vault in 2.5 seconds with source attribution and one-click navigation to source notes.

1,280 chaotic tags, three different frontmatter formats, fixed in 30 minutes for $1.50 using AI-powered batch processing.

1,024 notes, zero manual links, 2,757 bidirectional connections discovered automatically using vector search and semantic similarity.

782 files, 8 batches, 25 minutes. Building a dual-mode API architecture that automatically chooses between real-time and batch processing for 50% cost savings.

1,000+ videos, 2,757 auto-generated links, $1.50 in API costs. How we built an AI-powered pipeline to transform YouTube videos into interconnected Obsidian notes.

Claude Code's Agent Teams is a coordination layer that lets multiple Claude Code agents work together on shared codebases with explicit task managment, dependency tracking, and inter-agent communication. This feature is the difference between a productive parallel workforce and a chaotic swarm of agents overwriting each other's code.

Claude Code Hooks are user-defined shell commands that execute at specific points in Claude Code's lifecycle. They are not prompts; they are system-level interceptors that guarantee certain actions always happen (safety checks, quality validation, observability logging, and workflow enforcement) regardless of what the agent's reasoning chain looks like.

A project with 8 skills and 10,000 lines of domain documentation loads just 500 tokens at startup instead of 70,000, because progressive disclosure means agents pay for knowledge only when they use it.

The most effective AI coding agents aren't the ones with the cleverest prompts. They're the ones with the best-designed environments. Here's how to build agents that reliably ship production software over extended sessions.

An orchestrator breaks a task into pieces. Specialized agents pick up work items, each carrying skills that define what they know and hooks that enforce how they behave. Context flows from session start to task completion through a deterministic pipeline. Here is how the pieces fit together.

Traditional RAG finds documents that mention your search terms. GraphRAG follows the relationships between entities to answer questions that flat retrieval cannot.

Graph databases represent a shift in how we store and query interconnected data

Instead of relying solely on parameters learned during training, RAG-enabled systems dynamically fetch relevant information from external sources, incorporating this context into their responses.

Vector databases have become essential AI infrastructure, enabling everything from advanced semantic search to personalized recommendation systems and multimodal AI applications.