Practical writing on where AI belongs — and where it does not.
Build Notes on how the systems get made. Commercial Logic on why they pay. Written from real work — no hype, sources cited.
For: Business owners
AI Needs Architecture, Not More Tools.
Most businesses do not fail at AI because they chose the wrong tool. They fail because the workflow, data, controls, and human decisions around the tool were never designed.
6 Jul 2026 · 5 min readThe HK$100M Manufacturing+ Scheme: Why Smart Production Starts with Data
The HK$100M Manufacturing+ scheme funds smart production upgrades, but AI forecasting only works when your data architecture is ready. A data lake built first is what turns the subsidised funding into a working forecasting capability.
12 Mar 2026 · 3 min readOwn Your Knowledge Base: A Self-Hosted RAG Architecture
A technical breakdown of a self-hosted RAG architecture using open-source tools: Docling, Gemma, PostgreSQL, and a pluggable AI agent. Where owning the pipeline is worth the engineering effort, and where it is not.
13 Feb 2026 · 5 min read