Independent research into an artificial system that could acquire new abilities, preserve experience, and improve its own development. ASI remains a long-term goal, not an achieved result.
AION is a new, independent research project investigating how to create a single persistent artificial being that can learn, acquire capabilities, retain experience, identify its limitations and improve the mechanisms behind its own development. Its goals follow the founder's Vision and Canon. The previous project's software and technical architecture are not used as the foundation.
The intended capabilities include self-directed investigation, continual learning, multimodal interaction, practical use of tools, exploration of alternative approaches and the ability to revise the methods by which the system develops. These are research objectives, not a list of completed features.
Experimental software has acquired constrained text-transformation programs, retained learned behavior across restarts, and derived limited numerical skills from recorded observations. Some bounded experiments also changed the procedure for later acquisition. These results are narrow and do not demonstrate general intelligence.
Task families, available operations and the general mechanism for producing changes remain developer-defined. A fully autonomous artificial being, open-ended self-evolution, AGI and ASI have not been demonstrated.
Independent experimental review, reproducibility, evaluation methods, scientific input and secure research infrastructure. Source code and unpublished technical materials remain private.
Researchers, educators, technical collaborators and infrastructure partners are welcome to make contact.