Notes on building production AI
Agentic AI, MCP and the unglamorous engineering that keeps LLM systems running inside real, GDPR-bound businesses.
June 5, 2026 · 6 min read
Why most AI never leaves the demo — and how to ship agents to production
A flashy agent demo takes an afternoon. An agent that runs every day — monitored, GDPR-safe and cost-controlled — is a different discipline. Here is what actually separates the two.
ReadMay 30, 2026 · 7 min read
One MCP server, five scopes: how 27 agents share a single brain
Letting every agent wire its own bespoke tools is how you end up with 27 incompatible integrations. The Model Context Protocol let me collapse that into one well-guarded entry point. Here is the architecture.
ReadMay 22, 2026 · 5 min read
GDPR-compliant AI for German companies: a practical checklist
German clients do not ask "can it use GPT?" They ask where the data goes, who can see it, and whether you can prove it. Privacy is not a feature you bolt on — it is a design input. Here is how I build it in.
ReadWant this kind of AI running in your business?
Tell me your most painful manual process — I will show you where AI takes it over.