In this TED talk we discuss the current reasoning capabilities and ongoing developments both from a theoretical and practical perspective and comment on the role of Agentic AI platforms. In the theoretical part we discuss how the internal architectures foster or hamper reasoning capabilities and what measures can be taken to strengthen them with either training, prompt and context engineering. In the practical part we focus on cybersecurity application like AI-assisted software analysis and answer if AI can autonomously design complex cyberattacks that are not yet known.