More on Mechanical Writing

Here’s some more somewhat-mechanical writing about mechanical writing.

This is a follow-up to “In Defense of Mechanical Writing”

A colleague alerted me to Claude Output Styles, which change how Claude composes its responses.

Claude ships with six styles: Normal (aka Default), Concise, Explanatory, Learning, Formal, and Proactive.

So I repeated the experiment from the original article: take the same source text, run it through each configuration, and see what changes — and, more importantly, how well each version communicates.

The results are here.

Along the way, I discovered something more interesting: Claude lets you create your own custom output styles. So the next step is to turn what I’ve learned into something you (and I!) can actually use.

Can we turn communication frameworks into reusable output styles?

Rather than repeatedly prompting for BLUF, Pyramid Principle, SCQA, STE-100, or PREP, can we make those mechanics part of the environment itself?

Using AI well means figuring out where the friction is useful, and sanding away the friction that isn’t.

More semi-mechanical writing about mechanical writing — and What Good Looks Like — to come. (em dashes my own)

Share