Notes on building
SaaS & AI products
Practical writing on RAG, natural-language data, choosing models and shipping AI to production — from the team that builds these systems for a living.
Graph engineering explained simply: the tutorial for anyone who already knows RAG
Vector search finds things that sound like your question; a graph finds things connected to your answer. A beginner-friendly walkthrough of knowledge graphs, typed edges, GraphRAG, what the benchmarks really say — and how to build your first graph in a weekend.
Read article →Loop engineering: stop prompting your AI, start designing the loop
Stop prompting your agent by hand and design the loop that prompts it for you. A full tutorial: the five parts of a loop, the four loop patterns compared, the three exits you must define before you start, why loops burn tokens — and how to build one that converges instead of spinning.
Read article →The agent harness: why the model is the smallest part of an AI agent
Swap the model in a working agent and little changes; swap the harness and everything does. A full tutorial: the five layers of a harness, the control plane between intention and action, context strategies, sandboxing and permissions — and the evidence that the harness beats the model.
Read article →From prototype to production: shipping a reliable AI feature
Every AI demo works on stage. The gap to production is evals, guardrails, observability and graceful failure — the unglamorous work that makes a feature trustworthy.
Read article →Choosing the right LLM for your product: a practical guide
The biggest model is rarely the right default. Picking by task, latency, cost and data-privacy usually lands you on a mix of models — each doing what it is best at.
Read article →Building an AI knowledge assistant your support team will actually use
Support teams answer the same questions endlessly. A knowledge assistant over your own docs can handle the repetition — but only if it cites sources and knows when to stay quiet.
Read article →Ask your database in plain English: how natural-language-to-SQL works
Most people who need answers from a database can’t write SQL. A natural-language layer turns "how many orders shipped late last month?" into a query, runs it safely, and hands back the answer.
Read article →What is RAG, and why your business data belongs in your AI
An LLM on its own only knows what it was trained on. RAG gives it your documents, your policies, your data — so the answers are grounded in what is actually true for your business.
Read article →Have an AI project in mind?
Tell us what you're trying to build and we'll map the fastest path to a real, production-ready product. Free consultation — no obligation.