We write about what we build daily — and what's changing right now for owner-led mid-market companies in the DACH region. No hot takes, no hype cycles. Insights from operational reality.
Fable 5 leaves Claude subscriptions on 7 July — temporarily, not for good. The chronicle of the 19-day forced shutdown and what the mid-market should learn.
Mistral, Aleph Alpha or EU hosting? Why “European alternative” means three different things — and how to find and test the right option per workload.
High-risk duties postponed, transparency duties still coming: the sober state of play for 2 August 2026 — with a deadline table and a five-point list for the mid-market.
The market is flooded with new AI agencies — many working from the same templates. Seven signals and five screening questions that reveal a real partner.
From PDFs, scans and forms to structured data — with bounding boxes, block types and confidence scores. Why the European, self-hostable document AI matters for the mid-market.
GPT-5.6, Gemini 3.5 and Claude in summer 2026: instead of “which is best?”, the criteria the mid-market should use to choose the right model.
Google Cloud names 5 AI agent trends for 2026: 70% of enterprises are in production. What employees, workflows, security and skills mean for the mid-market.
AI as a growth opportunity for Europe: why sovereignty, compute and trust are decided in the mid-market — and how to start concretely now.
Is the “AI employee” dead? Why one context-rich system beats many role-based agents — and when multiple AI agents still win for the mid-market.
Is China’s new AI really 6× more efficient than Claude? What’s behind DeepSeek & co. — MoE, sparse attention, data residency — and what it means for the mid-market.
What are agentic workflows — and is n8n really dead? Workflow vs. agent, the five key patterns, the role of tools and MCP — and when agentic AI pays off for the mid-market.
Anthropic’s most capable model — 1M token context, but twice the price of Opus 4.8. What the “blocked” narrative means, what Fable 5 costs and which workloads it really pays off for.
Eigenes LLM aufbauen oder kommerzielle KI-API einkaufen? Was die Zahlen zu Kosten und Modellqualität sagen, wo DSGVO und EU-Datenresidenz den Ausschlag geben — und wann sich Eigenbetrieb wirklich lohnt.
Der Klick stirbt, die Antwort entscheidet. Warum 50 % der Deutschen KI-Chats statt Google nutzen, was AEO von SEO unterscheidet und wie du in ChatGPT, Perplexity und Google AI Overviews zitiert wirst.
DeepSeek V4, GPT-5.5, Claude Opus 4.7 im Vergleich. Wann lokale Modelle für inhabergeführte Unternehmen Sinn machen — und wann nicht.
Why real GitHub issues matter more than toy tasks — and what companies should learn from SWE-bench.
The business comparison: assistance, agent work, limits and useful company workflows.
Demos are easy. Maintainable software needs structure, tests, architecture and ownership.
Our proof: not just talking about agents, but working daily with specialized roles.
The pillar article on benchmarks, SWE-bench and the difference between tools and managed agent systems.
MCP sounds like tech insider jargon but is probably the most important development of the year for anyone wanting to connect AI to business systems. What it is, and why it matters now.
When consultant decks throw around "Agentic AI", "Autonomous Workflows" and "Multi-Modal Systems", a reality check pays off. We translate the buzzwords into operational reality — showing where the hype holds, and where it tips over.