How to build AI systems that survive production.
Field notes on AI automation — agents built with n8n and LangGraph, RAG systems, evals, and LLMOps. No hype, no demos: just architecture decisions, tradeoffs, and lessons learned from real projects.
What I write about
AI · AI · AIAgents & Orchestration
Building reliable multi-step agents with n8n and LangGraph — control flow, tool use, retries, and human-in-the-loop.
RAG & Retrieval
Chunking, embeddings, hybrid search, and grounding LLM answers in sources that actually hold up in production.
Evals & LLMOps
Measuring quality, catching regressions, cost/latency tradeoffs, and shipping AI systems you can trust and maintain.
The archive
All essays
First issue in progress
The first essays are being written.
This journal is written entirely about AI automation — deep, practical, and grounded in sources. The moment an essay ships, you'll find it here, in the feed, and in the sitemap.
- 01
Designing agents that survive production
Guardrails, retries, and observability for n8n / LangGraph workflows.
- 02
RAG that actually grounds answers
Retrieval quality, citations, and evaluating hallucination rates.
- 03
Evals before you ship
A lightweight harness to catch LLM regressions before your users do.