Notes from the engine room.

How we design, build, and run software for startups, enterprises, and the Mittelstand. Written by the engineers who do the work.

• July 26, 2026AI strategy for the Mittelstand: the first ninety daysMost AI strategies are documents nobody implements. A concrete sequence for the first three months: which process to take first, how to judge it, and how to tell when to stop.aimittelstandstrategy6 min read
• July 26, 2026Choosing an AI agency for the Mittelstand: five questions that place any vendorThere are now more AI consultancies than AI systems actually running in production. Five questions that place a vendor quickly, realistic price ranges, and the criterion most buyers leave out.aimittelstandconsulting7 min read
• July 26, 2026Intelligent process automation: when you need AI and when RPA is enoughAI is not a better version of RPA, it is the tool for a different part of the problem. Where the boundary runs, how a real process splits across both methods, and what the AI portion adds to running costs.automationaiprocesses5 min read
• July 26, 2026Automating business processes: where it pays off and where it does notA calculation instead of a recommendation. How to judge candidates for business process automation, which four process types almost always pay off, and which four you should leave manual.automationprocessesmittelstand5 min read
• July 24, 2026Run LLMs offline on Windows: llama.cpp and Qwen3.6-27B on your GPUA beginner friendly guide to running a serious language model on your own PC. What models, GGUF files and quantizations actually are, and how to get Qwen3.6-27B answering on your GPU in under an hour.aiguideoffline6 min read
• July 24, 2026Built and operated: why one team should own your software end to endMost software projects fail in the handover, not in the build. Here is what changes when the team that writes your system also runs it in production.engineeringoperations2 min read