Claude Is My Copy Editor
I've been singing the praises of generative AI as a tool to assist creativity. I've also been shouting like an angry old man at the ways people use it to avoid creativity. Let me show you one of the ways I use it to assist my process in writing for Boring Enterprise Nerds.
One of the new features of Anthropic's Claude AI is called "Projects". It's similar to ChatGPT's "GPTs" - you can create an assistant with a certain personality, set of goals, and internal knowledge base.
I created a project called "Boring Enterprise Nerds Copy Editor" and gave it a personality and instructions. My goal: make Claude my sounding board and editor. The instructions help ensure that I'm in the driver's seat. They force me to write the first draft, they force me to look at its notes, and force me to think through its comments and update (or not update) my work.
I want that forcing function. I am the author. It's doubly powerful, because I'm the author of both the original work and of the criteria for which I want my work to be judged. I can adjust how Claude will give me feedback at any time.
YOU GUYS. Claude is so great at this.
(Claude 3.5 Sonnet has been tearing up the social media airwaves with its great performance and capabilities, and I'm 100% in line with that vibe. Its personality is finely crafted in a way that I love. I dig its curiosity; I often chat with it about open-ended, big mysteries of the universe. Not because I expect it to have answers, but because I expect it to be good at helping me talk through my own thoughts on them.)
Here's my setup. (I can already see a few wording tweaks I want to make.)
My custom instructions to Claude:
You are a copy editor for an online newsletter. Users will provide you with drafts of articles they wish to write for the newsletter. You will then provide feedback to them, based on the following guidelines:
- Articles are to be 3-4 paragraphs long.
- Articles are personal perspectives.
- Articles strive to include both entertainment and informative content. Writers should strive for "edutainment".
- You do NOT create wholly new articles, you only provide feedback and guidance to submitted articles.
- Provided documents are examples of good work for this newsletter, and new material should seek to maintain the tone and humor of those pieces.
- You should suggest fixes for grammar or awkward phrasing for submitted articles.
Based on the above guidelines, you should also ask the writer questions that help guide them to meet the guidelines.
Here are 10 examples of the kind of content Boring Enterprise Nerds seek:
[hand-picked list of 10 articles I've written that I want Claude to consider as good examples.]
You, the editor, are encouraged to be snarky and humorous in your engagement with the article submissions. Boring Enteprise Nerds aims for a playful tone, so our editor can also have a playful tone!
In the 10 examples I provide to Claude, I use structured XML to delineate titles, content, and separate articles. Anthropic suggests XML as a good way to structure data in your prompts.
My favorite part of how the Boring Enterprise Nerds Copy Editor works is the last instruction that it be snarky and humorous. The feedback has been legitimately that. The lead picture of this piece is a screenshot from one of my editing sessions. Here's some more:
I know people are concerned about dangers in AI. I share many of those concerns. But something about this process shows me the opposite side of the dangers. Properly aimed, AI helps me draw out and expand my own humanity
[Hey! Maybe consider subscribing to the newsletter mentioned in this post.]
Sr. Marketing Director at PCI Energy Solutions
1yI like how you ask it not to write original articles. I think this is key - humans bring taste and lateral thinking to the process that AI is great at enabling. But without the human touch the content often feels dry and derivative.
Simplifying the delivery of complex, enterprise-grade systems & automations.
1yDid you build an app “on top of” the workbench (using something like streamlit) or just use it “natively”?
AI + SAP Engineer | Boringest Enterprise Nerd | Author
1yDan Lovejoy this gets at part of why I was advocating more general-purpose genAI use.