AI & Automation

If you or your team is doing by hand what a system should be doing, this is where I write about fixing that. Build posts, learning notes, and real workflow patterns using n8n, Claude, and MCP.

Dive into the latest posts or explore my work on GitHub .

• RAG, Step by Step: What Each Part Does and Where It Broke

  6 Oct, 2026

  How each part of my small RAG works, from chunking and embeddings to BM25 and a reranker, and where each one broke, with the numbers.

• Same Bot, Same Questions, Different Answer

  28 Sep, 2026

  Six problems I hit red-teaming my own RAG bot, from a safety judge that contradicted itself to a guard beaten by a rephrase, each with the fix and what it taught me.

• What a RAG Eval Gate Actually Catches

  8 Sep, 2026

  Four real bugs a CI eval gate caught while I was building a RAG app, and what each one taught me about testing systems that answer in probabilities instead of exact strings.

• Deterministic Scripts Before Prompts: Cheaper, More Reliable AI Workflows

  9 Aug, 2026

  Why deterministic shell scripts, not the model, handle any step in my AI workflows that already has one correct answer, and where Claude's judgment actually earns its tokens.