Articles for people who
ship the stack
Rewrites on React, Node.js, TypeScript and AI. A translated body is shown when it exists. If it does not, you see the English text and a note that says so.
Tagged: Evaluation
Ontology Shape vs Model Size: What a Graph's Design Costs an AI Agent
How three lossless shapes of one insurance knowledge graph changed accuracy, tool calls, tokens and cost for small and large models on Azure Cosmos DB.
30 min readFifty Years of Retrieval Research Hiding Inside Your RAG Pipeline
RAG chunking, hybrid search, HyDE, reranking and evals all have decades-old ancestors in information retrieval. A tour of the research and where it falls short.
17 min readWhat a Model-Diffing Agent Finds: Style Shifts Overshadow a Narrow Fine-Tune
A model-diffing agent audited a finance LoRA fine-tune: 82% of runs caught the length and structure shift, while the intended reasoning change stayed hidden.
15 min readAuditing RAG Retrieval for Stale Chunks, Drift and Noisy Context
A Python audit for RAG retrieval: golden queries, cosine relevance gates, Kolmogorov-Smirnov drift checks and context efficiency, plus how to read the report.
12 min readDebugging RAG Retrieval: Recall@K, BM25, Hybrid Fusion and Reranking
A small experiment on a RAG pipeline that ranked the right chunk fourth, showing how chunking, BM25, rank fusion and reranking change Recall@3 and MRR.
11 min readRAG-Anything on 65 Wine Books: Where Graph Retrieval Wins and Fails
A hands-on test of RAG-Anything's knowledge graph on 65 wine PDFs: parser choice, costs, graph versus vector answers, and the sparkling wine blind spot.
12 min readFrom Prototype to Production RAG: Hybrid Search, Agents and Evaluation
A reference architecture for production RAG in .NET: chunking, hybrid and graph retrieval, re-ranking, query planning, typed output and ongoing evaluation.
8 min readWhy Production AI Agents Need More Than a Good Prompt
Production AI agents need more than a good prompt: context, tools, memory, failure handling, observability, testing and autonomy matched to the actual risk.
7 min readA Field Map of RAG: From First Principles to Production Failures
Two pipelines, three steps and a list of named failures: how retrieval-augmented generation works, where it breaks, and which fixes address each problem.
7 min read
About these articles
Request a 24h estimate
Need the same stack in a production operator layer? Send the brief — estimate within 24 hours.