The idea would be to focus on the areas of effective utility for end users (which are limited but not non-existent)
What percentage of consumer use of pay-per month or free LLM use could be replaced by this kind of smaller software?
Would a wider group of people be open to using LLMs if they had control over the training data and the system?
How big would such a system need to be to be useful in these limited ways?
2/