An AI has formalized the proof of Fermat's Last Theorem in Lean, in 11 days according to Anthropic. This article explains Lean, the agent swarm and the limits of the announcement.
Highlights
More Posts
AI systems are increasingly improving their own development through recursive self-improvement (RSI), in which an AI system improves its own capabilities in order to improve itself. This leads to rapid progress, but there are challenges in control and safety, as AI systems still struggle to choose goals and assess results. Research is under way on how to steer RSI while still driving progress.
When you type a question into an AI chatbot, the moment feels almost intimate: you formulate a thought, press Enter, and a few seconds later a fitting answer appears. This apparent immediacy obscures how much happens in the background.
The KI-MIG, Germany's act on AI market surveillance and innovation promotion, governs the use of artificial intelligence. The Federal Network Agency is the central AI supervisory authority and offers companies support and advice. Specialised authorities such as BaFin supervise sectoral areas.
OpenAI's new language model GPT-6 Astra shows impressive progress in certain areas such as solving complex programming tasks and finding security vulnerabilities. In general intelligence and handling tasks across different professional fields, however, clear differences remain compared with Anthropic's top model, Claude Fable 5.1. The benchmark behind the "AGI era" claim is disputed, since OpenAI's own test methods may overstate the actual performance.
Security scan or pentest? This article shows why a specialized website check is usually the more important first step and where a pentest adds value.
Modern AI can help companies create reports quickly and efficiently by merging data from different sources, detecting patterns and turning them into understandable summaries. Classic reporting processes are often time-consuming, error-prone and not scalable. AI-powered reporting systems automate many of these processes and let professionals concentrate on analyzing results and making decisions.
Cloud AI services such as ChatGPT or Claude are poorly suited to querying large collections of company documents: context length, API costs and statistical analysis are limiting factors. Generic solutions need an extra RAG layer that still cannot cover every structured query, so a tailored solution is often the better choice.
The images from the World Defense Show in Riyadh in spring 2026 looked like science fiction: a four-legged combat robot, ready to buy. Prices are falling fast – this article shows when AI robots become part of everyday life and when they become dangerous.
AI is changing the labor market rapidly – but not all jobs equally. Find out which jobs benefit, which come under pressure, and why empathy and craftsmanship remain irreplaceable in the age of AI.
When does local AI pay off, how does it work with company knowledge, which language model fits and what should data protection officers check? The overview leads to basics, practical tests, guides and examples.