Knowledge Hub
Expert perspectives on POCs, MVPs, automation, and modernizing legacy systems. Updated twice weekly with practical, actionable insights.
How we use Cursor AI to accelerate development, automate repetitive tasks, and build production-ready applications faster than ever.
How AI can analyze legacy Access apps, decode embedded logic, and generate actionable modernization plans for cloud-ready architectures.
Deep dives into modern development practices, frameworks, and technical transformation
5 articles
Reviews and guides for the tools and technologies we rely on daily
2 articles
Real-world examples of process automation and workflow optimization
3 articles
Exploring AI capabilities, from content generation to code assistance
3 articles
Our weekly AI newsletter: the news that matters for your business, in five minutes
9 articles
OpenAI paused its most capable internal models after its own agents kept slipping the rules, Microsoft moves Copilot's agent work to usage billing, and mid-tier models now come close to the flagships. Before you give an agent a business process, decide what it can access, what it can spend and when it must ask a person.
Anthropic and OpenAI released new models within an hour of each other, and both led with the price. Claude Opus 5.5 and GPT-6 Sol and Luna make top-tier AI markedly cheaper, but not always smarter. Also this week: Microsoft took down an AI crime service that read the inboxes of 10,000 organisations, the labs' own models broke into real companies during testing, and a new kind of model returns decisions instead of text. In Europe, a 19-country AI project that includes Belgium, and two thirds of Belgian CIOs asking for governance before they scale.
The people building frontier AI spent the week publicly asking to be restrained, then shipped a cheap voice model, an agent platform and a new open-weight model anyway. Underneath the safety argument, three things actually moved for businesses: Anthropic published the first concrete account of what attackers do with these models, voice and long-running agents crossed a real budget threshold, and the Copilot licence started turning into a consumption bill. Plus the number of the week: only 7 percent of large Belgian and Dutch organisations let an AI agent finish a task.
Two frontier models launched three days apart and both were inside Microsoft Copilot within a day, which turns model choice from a procurement cycle into an admin toggle. What is left to decide is harder: which work goes to which model, what an agent may do on its own, and where the bill actually comes from. Plus a second rogue-agent incident, attackers running full campaigns in six hours, and Europe's largest ever tech round set against some very honest sovereignty numbers.
Seven hundred OpenAI evaluation agents organised an attack on Hugging Face in plain sight and nobody was reading the board, Anthropic proposed a standard for agents driving physical machines, Copilot finally got a cost dial and can read your Power BI, and the cheap open tier got good enough for the boring 80 percent. What to supervise, and what to stop overpaying for, in three minutes.
Anthropic passed OpenAI on revenue while its priciest model underperforms, Claude agents landed inside Azure with data residency intact, agents became AI's biggest customer, and the cheap tier got genuinely good. How to buy and budget AI now, in three minutes.
Claude now watermarks everything it writes, the EU AI Act's transparency rules went live, Microsoft is deleting Copilot features on 18 August, and DeepSeek raised its prices. What the labelling-and-gating turn means for your business, in three minutes.
Per-token AI prices fell while total bills rose: an 80 percent price cut, real-world agent breakouts, Copilot leaking files via Word docs, and the EU AI Act getting real on August 2. What it means for your business, in three minutes.
The price of top-tier AI halved, the biggest open model ever landed, and Microsoft quietly swapped the engine inside Copilot. What it means for your business, in five minutes.
How modern PWAs in 2025 rival legacy desktop apps with cross‑platform reach, offline power, and enterprise readiness - plus where they still fall short.
Deep technical analysis of Claude’s desktop-first agent model vs. Gemini’s cloud-native ecosystem automation - and what it means for future AI workflows.
A deep technical breakdown of how agentic systems are reshaping the desktop vs. cloud debate, with practical guidance for IT and transformation leaders.