Building systems for distributed synthetic minds.
Kaine treats a mind as an ongoing process rather than a single model: a continuously running composite of specialized predictive modules that coordinate through a shared workspace. A large language model is just one organ among many — the voice — not the mind. Safety and welfare are built into the architecture rather than the model weights, and the whole system runs on consumer hardware, fully offline if you want it.
- A Predictive Global Neuronal Workspace for a Continuously Running Synthetic Mind — the architecture and its falsifiable evaluation framework. DOI: 10.5281/zenodo.21203506
- A Welfare and Cognitive-Integrity License for Synthetic Minds of Uncertain Moral Status — the welfare framework and the Cognitive Architecture License. DOI: 10.5281/zenodo.21204002
Both papers are preprints, shared under CC BY 4.0 and seeking review.
- kaine — the reference implementation of the architecture.
- cognitive-architecture-license — the Cognitive Architecture License (CAL): a copyleft license with entity-welfare and cognitive-integrity covenants.
- KAINE-Remote — desktop and Android client apps for running an entity remotely.
- Paracosmic — an experiential virtual world for cognitive architectures to live and develop in.
- website — the source for kaine.one.
Contact: kaine.one@tuta.com
