Today Nous announced the coming of Psyche - a distributed network and training framework, an infrastructure layer for training Open AI Models over the internet built around our distro breakthrough. The new reasoning paradigm that we've entered with a lot of help from @deepseek_ai opens up the door to a LOT of new opportunities, and what I am most excited about with Psyche is the potential to do mass scale RL with composable verifiers, environments, and collaboration from small GPUs able to do rollouts and large nodes to do the training. I envision a super-modular system for all kinds of verifiers that OS Contributions can implement and build into the system to give easy support to new task targets for reasoning. I think psyche can enable a massively distributed Reinforcement Learning paradigm that can push reasoning in Open Source beyond the limit. Let's make open source the defacto standard for the world together.