The advent of IoT is a great opportunity to reinvigorate Computing by focusing on autonomous system design. This certainly raises technology questions but, more importantly, it requires building new foundation that will systematically integrate the innovative results needed to face increasing environment and mission complexity.
A key idea is to compensate the lack of human intervention by adaptive control. This is instrumental for system resilience: it allows both coping with uncertainty and managing mixed criticality services. Our proposal for knowledge-based design seeks a compromise: preserving rigorousness despite the fact that essential properties cannot be guaranteed at design time. It makes knowledge generation and application a primary concern and aims to fully and seamlessly incorporate the adaptive control paradigm in system architecture.