A Horizon Europe consortium completed and demonstrated an open cognitive architecture that helps autonomous robots maintain explicit models of themselves and their surroundings, detect unexpected situations and adapt their behaviour.
Q.ANT released an open-source SDK for programming its photonic processors, with C and Python APIs, documentation, sample applications and a CPU simulation backend. The simulator lets developers explore computing with light on an ordinary laptop, removing the need to secure scarce photonic hardware before writing or validating an application.
An open-weight cybersecurity AI model designed for autonomous penetration testing and fully local or air-gapped deployment. A locally deployable cybersecurity model is particularly relevant for organizations that cannot send sensitive source code or security data to external AI providers.
A production deployment of 127 Skypod warehouse robots at GXO's Venlo fulfillment facility for Guess, with 60,000 rack locations, eight picking stations and approximately 200 meters of conveyors. This is technology operating at meaningful physical scale rather than a prototype, funding announcement or pilot.
An open-source demonstrator in which AI coding agents modify high-integrity Ada and SPARK software while deterministic tooling independently checks proofs, requirements traceability, tests and structural coverage.
A major full-stack development platform release adding hot reload, agentic-development workflows, embedded Postgres, experimental offline sync, cloud deployment and an environment designed to work with external coding agents. The architecture treats coding agents as interchangeable clients of the development environment instead of tying developers to an IDE-specific agent.
An open 3.4B-parameter document OCR and page-to-Markdown model with approximately 570M active parameters per token, optimized for inexpensive GPUs. Efficient document parsing is increasingly important infrastructure for RAG and agent systems, while the relatively small active model could make high-quality local processing cheaper.