Building user skills in the world of AI: Prompting fundamentals

We have ways to detect names without @mentions. For example, sharepoint search can match a name "lexically" -- documents and communication containing a specific text will be matched when searching for that specific text (e.g. "Aaron Halfaker" for me). So long as a name is unique, it can be found that way. But it's much easier to know who is being referred to when a @mention is used. There are, of course, secondary effects of @mentions in that they lead to notifications so you may choose not to @mention in some circumstances and that is fine. But the cost is that retrieving documents and communications without and @mention will be less consistent and predictable. One of the super powers of language models is that they can often work out who is being referred to from context. E.g. if someone sends an email to a group and refers to someone by their first name only, the language model can be smart enough to cross-reference the "To" and "CC" list on the email to disambiguate who exactly is being referred to. But we make the language model's job easier and thus it's outputs more consistent when we don't require it to cross reference in order to make sense of what we need it to understand for us.