Insights
Worth reading before you decide
Short, concrete pieces on AI in practice: what works, what does not, and what it asks of you. Written for the person making the decision, not the person writing the code.
10 September 2026
5 min readRead the articleWhy the AI is so confidently wrong
A language model does not look things up, it predicts text. Here is what that means in practice, where the errors cluster, and what actually helps.
How AI worksTraining10 September 2026
5 min readRead the articleWhy the AI gets it wrong: it is almost always context
The model only knows what it sees at that moment. Four ways context goes wrong, and why context work matters more than the choice of model.
ContextPractice10 September 2026
4 min readRead the articleWhat should you automate first?
Three questions that sort the candidates, where the line between manual and automatic should sit, and why the big first project is the wrong instinct.
WorkflowsStrategy10 September 2026
5 min readRead the articleWrite down how you work, once
Tribal knowledge costs more than it looks like it does. What a skill actually contains, and why writing it down pays off even without AI.
SkillsWays of working10 September 2026
5 min readRead the articleChatbot or agent: what is the difference, and when do you need which?
A chatbot answers, an agent acts. What changes when the model gets tools, where the line should sit, and which one you actually need.
AgentsAutomation10 September 2026
5 min readRead the articleFrom demo to production: what is missing when the AI has to do real work
The demo was never the hard part. Five questions the harness around the model has to answer, and why they get skipped.
HarnessProduction8 September 2026
5 min readRead the articleWhat is sovereign AI, and why are Swedish companies talking about it now?
Sovereign AI means you own the model, the hardware and the data. Here is what the term actually means, what is driving the interest, and what it is not.
Sovereign AIStrategy8 September 2026
7 min readRead the articleLocal LLMs for companies: what hardware do you actually need?
GPU memory decides which models you can run and how many people can use them at once. A practical guide to sizing an AI server without overbuying.
HardwareGPU8 September 2026
6 min readRead the articleGDPR and AI: why running locally is the simplest answer
Processor agreements, third-country transfers and the CLOUD Act. A local model removes most of the questions before they are asked.
GDPRCompliance8 September 2026
6 min readRead the articleOpen models versus the cloud: is local good enough?
For most internal tasks, yes. Here is where open models stand, where the cloud still wins, and how to decide which applies to you.
ModelsStrategy8 September 2026
5 min readRead the articleTailscale or SSO? Giving the right people access to the AI server
An AI server nobody can reach is useless. One everybody can reach is dangerous. Two proven ways to get it right, depending on what your IT looks like.
NetworkSecurity