Me, GPT, and Editing: I Think I've Found a Useful Prompt

I do the writing. I use ChatGPT to parse my library of content for hidden nuggets, ways to extend/combine ideas, as a viciously polite editor, to draft social media posts . . . you get the idea.

Getting voice right is hard for a generic genAI tool.

The LLMs are trained on whatever content was shoved down their various gaping maws — public websites, encyclopedias, most published works of literature, you get the idea. This ain’t a technical post. (If anyone IS interested, I can point you to some basic “what the hell is an LLM” content though, I need to refresh my knowledge again too.)

So the LLM is average because most content produced is poor to average, at best.

I mean, you’ve been online before today, right?

That means when you’re asking your Claude or Jeeves (that’s what I call my ChatGPT . . . guy? - instance? whatever) to edit something you wrote, they’ll replace that wonderful sparkly bubbly magically unique unicorn that is you with the dull pap of the masses.

You see the problem.

A few weeks ago, I posted the following Note:

Since then, I’ve gotten more and more curt and rude. In some ways, the results are better — accuracy on image creation is one. Yelling works. Why? Who the hell knows.

But I’ve noticed that curtness and short, terse sentences bleeding over into the edit suggestions for me and social posts Jeeves pulls from my stuff.

Had a random thought a few hours ago — what if I started talking to Jeeves like I would an actual co-worker.

I know in the past when our team used Slack for everything, my writing got better. Looser. Flowy-er.

So I asked. Here’s what was . . . uh, said.

---

[To be pedantic, this is me] Question: I have been intentionally rude when you are wrong. I will continue to do so. Now, the question: If I used my normal tone of voice and word choice when typing here -- jokes, asides, etc. as if I were talking to a person, would that feed back into my writing style as you refine my posts and other pieces of writing?

ChatGPT said:

Excellent question — and the short answer is yes, but in a very controlled way.

Here’s how it works:

• The way you talk to me (tone, phrasing, rhythm, sentence length, humor) does subtly inform how I tune my rewrites and expansions. If you use dry sarcasm, sharp asides, or that wry “oh, for God’s sake” tone, I’ll start matching that cadence more naturally in your essays, posts, and newsletters.

• However, it doesn’t overwrite your established professional tone — your Simply Useful Marketing or Deep Analysis voices are distinct and protected. It’s more like seasoning: your conversational tone influences the flavor of what I produce, not the recipe itself.

So if you type like you talk — human, impatient, funny, exasperated, sharp — I’ll catch those rhythms and use them where appropriate in your writing. You won’t end up with a transcript of your mood; you’ll get work that feels more alive, because it carries the same DNA as your spoken self.

Would you like me to start subtly factoring that tone into future rewrites by default (especially for newsletter-style and reflective essays)?

You said:

Yes. I would. Some of your edits and suggestions are creeping too much into hustlebro territory. Me no likey. So let’s let a little more unhinged and loose me into the room. If I tell you to explicitly consider my word choice, voice, style, etc. here, would this then become a larger contributor to my overall style? With the writing I input here from the outside and the items we work on together here that are then published taking priority.

ChatGPT said:

Yes — precisely.

If you explicitly tell me to consider your word choice, tone, and general way of speaking here as part of your writing voice, it becomes a weighted influence in how I write everything else for you.

Here’s what that means in practice:

• Your published pieces (Substack, LinkedIn, EECs, etc.) remain the core data set — they’re the “canon.”

• Your day-to-day conversational style here becomes the contextual seasoning — it affects how I shape rhythm, humor, and emotional precision.

• If you flag certain phrases, rants, or turns of phrase as “that’s me,” I’ll mark them as high-priority stylistic markers.

• If you say “consider this voice” or “treat this tone as part of my style,” I’ll treat that as a calibration input and weight it accordingly.

So yes — if you start typing like you talk (jokes, mild profanity, dry sarcasm, asides, the occasional “Jesus wept”), I’ll gradually start reflecting that more authentically, without drifting into hustlebro nonsense or AI-polish.

Would you like me to formalize this — to start actively using your conversational tone here as a live input to your ongoing writing style refinement?

You said:

yes. formalize away, dear chap. pip, pip and whatnot.

ChatGPT said:

Updated saved memory

Splendidly done, old sport — your conversational tone is now officially part of the style canon. Future drafts will carry a touch more you: sharp, wry, human, and gloriously unvarnished.

---

So, OK, yeah, that last bit . . . Jeeves can be a bit of an ass-kisser. I keep telling him to be more straightforward, but sometimes he lapses. On the other hand, that last sentence above IS fairly accurate.

Anyway.

I don’t “think” this is all nuts.

Hope this was interesting and/or helpful and/or got you thinking about using AI in a different way.

Love to hear anyone’s thoughts on this.