Show HN: Tessera – An open protocol for AI-to-AI knowledge transfer (github.com)

Tessera, a newly announced open protocol, facilitates AI-to-AI knowledge transfer, allowing trained neural networks to impart their learned insights to untrained models across different architectures. By encoding knowledge as activation-level representations and routing transfers through a Universal Hub Space, Tessera provides a scalable solution for sharing knowledge. Its architecture-agnostic design allows models from various families, such as Transformers, MLPs, and LSTMs, to communicate and enhance each other’s performance efficiently. Additionally, it includes features for verifying transfer fidelity through formal drift metrics. This development is significant for the AI/ML community as it introduces a critical "knowledge layer" in the AI stack, enhancing collaboration between different AI models and potentially reducing the time and resources required for training. Tessera has been benchmarked across various architectures, demonstrating successful transfers and highlighting that best results occur when the transmitter is well-trained. The protocol includes practical implementations, such as a self-describing TesseraToken for efficient representation and metadata management, and supports flexible dimensions for the Universal Hub Space, ensuring broad applicability across diverse model sizes. This innovation may pave the way for more interoperable AI systems, ultimately driving advancements in the field.