Continued research and development into a series of efforts which have produced a measurable gain in the heightened quality of Local LLMs, and development towards a more accessible ecosystem for user development and curation of their own personal models.
To allow for a fully automated user curated cycle of training and dataset generation on affordable local hardware of a model which can hear, see, and read the users inputs locally and respond in kind with the same level of understanding and expressiveness that the user can.
Multimodality is oncoming, this project may be the first or one of the first but the margins wont be giant, the opportunity that this project presents for the decentralization cause is that it instantiates a standard for this type of communication to be done locally, and not over an API. Data provenance is important. Keep it, store it, dont share it, package it, clean it, and sell it. If were lucky enough people recognize the value of their data that the market demand takes care of us and both utility and money can be generated in a private way, where human generated data is only given consensually, and compensated fairly.