Knowledge that moves you forward.
Clear thinking on websites, software and AI. For decisions that make a difference in everyday work.
Jev vs Laya vs Clef: Open Decision Models Tested on 45 Customer Requests
Jev vs Laya vs Clef: we ran Clef, Laya and Strands Decider on 45 customer requests in German, English and Russian. Who sorts correctly, who is confidently wrong, and where Jev stands.
Read article
AI chatbot for WhatsApp, Instagram, Telegram and your website: what our AI Platform does
An AI chatbot for WhatsApp, Instagram, Telegram and your website without code: set it up by chatting, answers from your own documents, human handover and real use cases.
Obsidian second brain with AI: how an LLM maintains your wiki, with a setup guide and our test
Build an Obsidian second brain with an LLM using Karpathy's LLM wiki pattern: folder layout, a rules file, our own test with Codex and what to check for privacy.
Is AI overhyped? Why a free chatbot is the wrong yardstick for your decision
Is AI overhyped? Karpathy's funnel shows why executives often know AI only from a free chatbot, why checking is becoming the real work and how to test AI on a real task.
AI chatbot requirements document: what to include, with an example and acceptance test
An AI chatbot requirements document in eight sections, a filled-in example for a heating contractor and acceptance criteria with thresholds you agree before signing the contract.
Jev vs. LLMs: when a decision model beats Claude or GPT-6
Jev, Clef, Decision-1 or Claude? How decision models differ from LLMs, what they cost and when to use each, with a worked cost calculation for 100,000 requests.
Running AI in the EU: EU hosting and the 7 questions SMEs should ask now
EU hosting for AI: what data residency does and does not deliver in 2026, and 7 questions SMEs should ask before choosing a cloud and model provider.
Five layers of an AI agent: from a prompt to a working system
Context, Loop, Jev, Harness and Evals: how information, action loops, routing, execution environments and evaluation turn a language model into an AI agent.
Opus 5.5, GPT-6 Sol and Luna: when a pricier model earns its keep
Opus 5.5 leads the benchmark; Luna charges far less per token. Compare the scores, API prices and integration traps before switching models.
When does your business need a customer portal on its website?
Use recurring customer requests to choose between a form, an existing service and a custom portal. Estimate the workload and define a useful first version.