Use Cases with LLMs

We are exploring some ideas within Pioneer Square Labs around using the growing list of open LLMs (large language models). What should we build for internal use and for us to open source to the community writ large?