Articles · ai
Taking the long view on technology: what is genuinely changing and how to decide with a clear head.
The EU AI Act now applies: four questions for a Swiss company
A Swiss address does not settle whether the EU AI Act applies. Start with the company’s role, the EU connection, the effect on people and the evidence needed before production.
Private AI for law firms: a risk-based architecture
Swiss law does not prescribe a particular machine. It requires firms to protect professional secrecy, govern providers and verify output. Here is a three-tier model.
A knowledge base the AI can read and a human can review
An organisation's knowledge lives in people's heads, in mailboxes and in a portal nobody opens any more. A folder of versioned Markdown files changes three things: review becomes possible again, authority becomes traceable, and the assistant stops guessing where one idea ends and the next begins.
Agent memory: one word, several mechanisms
Karpathy's LLM Wiki pattern, Anthropic's memory tool, context editing and managed stores do different things that the word memory conflates. What to write down, what never to, and what changes under Swiss law.
Reducing the IT bill: the levers that work, AI included
Caching, batch processing, model tiering, virtualisation, per-seat licences: what actually lowers an IT bill, the July 2026 figures, and the cases where free costs more.
RAG, loops, graphs: four decisions, not four products
The vocabulary of AI agents renews itself every quarter. Behind RAG, GraphRAG, context engineering, loops and graphs sit four distinct decisions: what the model sees, what verifies, who runs next, and where your signature goes.
Kimi K3, Soofi S and Apertus: an open-model snapshot
A dated snapshot, not a permanent ranking: release status, weight precision, internal benchmarks, total cost and deployment conditions.
An AI agent in practice: its work, and where it stops
Bank reconciliation in an accounting practice, incoming mail in a municipality, quotations in an SME, research monitoring in an institute: realistic AI agents, with their guardrails.
AI arrives in the classroom: the points to settle for the 2026 school year
Copilot for Fribourg lower-secondary pupils, a supervised tool in Vaud, the IB requiring AI to be cited: the new school year brings AI into the classroom. The real work is organisational: pupil data, approved tools, a policy and training.
Local AI: three machine profiles for several kinds of work
Mac mini, DGX Spark or a Blackwell workstation: three starting points for on-site AI, with the memory, context and operational assumptions that still need testing.
Internal AI without a US cloud: realistic in 2026?
Open models, Swiss hosting and local deployment: a dated method for selecting internal AI without reducing sovereignty to the server's location.
AI in the clinic and in the laboratory: two risk regimes, two architectures
A private clinic protects patient data under professional secrecy, the FADP and the Human Research Act. A biotech laboratory protects unpublished know-how and patent novelty. I build those two rooms differently.