Notes from the build — applied AI, written plainly.
Practical, source-backed articles on voice AI, agentic systems, MCP, automation and applied generative AI — structured to be useful to humans and discoverable by AI search.
Browse by topic
All articles
Governed MCP servers: tools Claude can't misuse
TL;DR — Policy layers, per-call caps, allowlists and audit trails that make autonomous tool use safe by construction.
From idea to deployed AI MVP in a weekend
TL;DR — An n8n + Claude workflow that captures, qualifies and routes leads automatically — start to deploy.
RAG that cites its sources: FastAPI + vector search
TL;DR — How JobSignal ingests postings and answers skill queries with cited, verifiable sources.
Designing "AI employees" that work end to end
TL;DR — The architecture behind always-on agentic systems that handle real tasks without a human in the loop.
Local LLM deployment with Ollama & MSTY
TL;DR — Running capable models on your own hardware — when local beats the API, and how to set it up.
System & meta prompting for reliable agents
TL;DR — The prompting patterns that make agentic behaviour predictable instead of improvised.