Posts

Diagram of a split agent architecture with a planner process and an executor process connected by a control channel.

Split Agent Architecture in Harness: Separating Reasoning from Execution

Harness separates its AI coding agent into a lightweight planner and a sandboxed executor. Here’s why that split matters for latency, security, and observability.

Abstract network diagram showing an API endpoint branching into multiple tool calls and looping back to itself.

Building Agentic APIs: When Your Endpoint Thinks for Itself

Agentic APIs trade the predictable request-response contract for an autonomous, tool-using endpoint that plans, retries, and adapts. This post covers the architecture, observability, and security patterns needed to ship them reliably.

Terminal screenshot of a fine-tuning training run showing loss curves.

Fine-Tuning a Small LLM: A Hands-On Experiment

A working engineer’s journal of fine-tuning a small open-source LLM with LoRA on a custom dataset — the full pipeline, the surprises, and the measurable results.

Diagram of a RAG ingestion pipeline showing document parsing, chunking, embedding, and indexing stages.

The Hardest RAG Ingestion Problems and How Engineers Actually Solve Them

Most RAG failures start long before the LLM is ever called. This post walks through the five hardest ingestion problems and the patterns teams use to fix them in production.

A network of connected nodes representing a knowledge graph replacing a flat vector index.

Graphify Replaces the RAG Vector Store: Why Graph-Based Retrieval Is the Next Leap

Graph-based retrieval is displacing the pure vector store in production RAG stacks. Here is how Graphify works, why it outperforms flat embeddings on multi-hop questions, and what it takes to migrate.