Prompting People

> The more surprising part is the unusual reactions of the other people getting a better picture and context of what I’m explaining without the usual back and forth - which has landed me my fair share of complaints of having to hear mini lectures, but not more than people appreciative of the fuller picture.

It’s not surprising to me at all. People don’t tend to appreciate being lectured at - especially in a conversational context. Moreover, people really don’t like being spoken to as if they’re robots (which is something I’ve started to notice happening more and more in my professional life).

The fact that the author considers these reactions surprising and “unusual” betrays a misunderstanding of (some of) the purposes of communication. Notably, the more “human” purposes.

From the article:

> I was midway through explaining a concept he hadn’t covered when he stopped me. He pointed out that my way of speaking had completely changed and how it was unusually structured and didn’t give him the opportunity to ask follow up questions.

IMHO this sounds like a bit of an exaggeration in service of a specific narrative for the blog post, but language convergence has been a topic of conversation ever since the earliest autocomplete features appeared on smartphones.

The feedback loop in this situation is (LLM trains on people <-> People then train on LLMs).

Agree it's not common to give unprompted background context within any given normal conversation. People usually default to the pull style which Id agree is ultimately less efficient.

All that said, even though AI prompting is forcing the issue, which is good, the takeaway should be that _intentionality_ is very high leverage. Less so that it's because of some given (prompt) structure.

I already tend to talk the way the author describes, which means I'm probably going to get accused of sounding like a prompter. -.-

(It's something I've been wanting to change, though.)

"prompt engineer". Good God. It's come to this now.

The medium shapes the message. People who use Twitter a lot tend to write in tweet form.

I just hope I don't start calling people dumb when they don't get what I'm trying to tell them do with a single prompt/request, like I do with Claude

That and also how many actors online in the threads are humans? That too is conditioning.

When you write for an LLM you write in order to get exactly the outcome you want. It is input to a machine.

Talking that way to a person is bound to make them feel weird for sure.

Back in the day we used to call it writing a brief, and using that brief for briefing teams and individuals on the project.