<aside> <img src="/icons/light-bulb_gray.svg" alt="/icons/light-bulb_gray.svg" width="40px" /> For whenever someone needs a blub

</aside>

Pluralis is a research collective focused on Protocol Learning. It is our view that true decentralized training (multi-node, internet grade node-to-node bandwidth, heterogenous and elastic swarm), is much closer to feasibility than broadly thought. Pluralis combines decentralized training with Protocol Models (models that can be trained, used, and built on, but whose weight set never leaves the Protocol). This approach enables collaborative, democratic creation of foundation models, and provides a mechanism to enforce value flow from model use back to training participants. Pluralis was founded by Alexander Long in April and has raised pre-seed with CoinFund. Alexander is an ML PhD from Australia and spent 3 years in Amazons International Machine Learning group where he worked on multimodal foundation models.

Pluralis Research is pioneering Protocol Learning—an alternative to today’s closed AI models and economically unsustainable open-source initiatives. Protocol Learning enables collaborative model training by pooling computational resources across multiple participants, while ensuring no single entity can obtain the complete model. This approach encourages genuine open participation, incentivises contributions without upfront costs, and guarantees economic sustainability for creators. Protocol Learning relies, along other things, on solving the open research problem of low-bandwidth, model-parallel training. Almost no one believes this is solvable. We think it is. Pluralis aims to globally distribute AI innovation, preventing monopolies and empowering individuals to drive advancements based solely on merit, not capital.