This is very much a work in progress. Let me just get that out of the way. This serves to replace the current how-to guide with something that has more long-term maintainability.
This aims to demonstrate the official and best methods for interacting with the bot. It is a complex system, and things don't always work out. If this AI was a human, it would be a mere toddler that still eats glue and draws on the walls with crayons. We're working consistently to see how people want to use the bot and ensure it learns accordingly. There is still a long way to go but this public testing phase has helped a great deal.
All examples that you'll find here were used as training models for the bot. They are not just random theoretical samples, they have been used extensively in training and testing. They're trusted.