Thoughts on
owning your stack.
AI agents, data sovereignty, workflow automation, and the tools you actually control.
lynox 2.0: the foundation goes live
The first major release. A unified automation core — workflows, triggers, and tasks in one store with a clear 'when vs. what' model — and an upgrade that migrates itself, with no action and no data loss.
2026-07-06 releaseWhat we shipped in June 2026
Nineteen releases in one month — provider-agnostic model tiers, an inbox and workflows that live in chat, readable documents, and the quiet foundation under everything that came next.
2026-06-30 releaseWhat we shipped in May 2026
One month, fifteen releases — Unified Inbox, passwordless login, an agent that stops making things up. The honest recap of lynox's busiest month.
2026-05-31 deep-diveSet-Bench: an agent benchmark for Anthropic & Mistral
What it actually costs to run an agent loop, measured across 8 axes for Anthropic and Mistral. Small models are 30–100× cheaper; caching swings it 2–5× more.
2026-05-30 philosophyWhy I built lynox
A small Swiss agency owner, 15+ years in the boring-but-important layer of digital ops, drowning in SaaS. The inverted bet that became lynox.
2026-05-27 deep-diveAnatomy of a lynox agent run
What actually happens between hitting send and lynox answering: the four systems behind every run — engine, session, knowledge graph, worker loop.
2026-05-15 guideGDPR and AI: How to run business AI in the EU
A practical guide to running AI-powered business automation without violating GDPR. Provider choices, data flows, and what your DPO needs to know.
2026-04-04 philosophySelf-hosted AI: Why it matters in 2026
Every AI tool wants your data in their cloud. Here's why running AI on your own server isn't paranoia — it's good business.
2026-04-04