I keep my own version of taste.md, which I call an influence roster. It's essentially a detailed list of people who have inspired me, taught me, or shaped me.
To offer up another context anchor, I'd also suggest a State of the Union. It's what's true right now. While your audience and your influencers will likely stay long-term, your current situation in life and/or business will change by the month, the quarter, or the year. If you want your AI to keep up with you, you have to keep it updated.
I found having a synthetic persona of my ICP very powerful, rather than only an audience MD. I build it on first-party data as well as scraping a lot of LinkedIn, and it really helps me create highly converting and highly engaging content.
The models already know how to write a LinkedIn post. What they don't know is what you'd cut.
Same thing in ghostwriting. The useful brief isn't "be professional but conversational." It's the list of phrases that sound fine and still aren't you.
Context stops AI averaging you out.
Taste is what decides what never gets near the page.
Taste.md is the one I'd fight hardest for. We built something similar for my own writing, and the part doing the most work is the list of phrases that sound fine but aren't how I talk. The model needs to know what to throw away as much as what to aim for.
Totally agree with this. An article without a POV lacks depth and won't interest readers. The key takeaway is how you can use AI to outline and sharpen your thinking.
I keep my own version of taste.md, which I call an influence roster. It's essentially a detailed list of people who have inspired me, taught me, or shaped me.
To offer up another context anchor, I'd also suggest a State of the Union. It's what's true right now. While your audience and your influencers will likely stay long-term, your current situation in life and/or business will change by the month, the quarter, or the year. If you want your AI to keep up with you, you have to keep it updated.
I found having a synthetic persona of my ICP very powerful, rather than only an audience MD. I build it on first-party data as well as scraping a lot of LinkedIn, and it really helps me create highly converting and highly engaging content.
Taste.md is the file I'd fight hardest for.
The models already know how to write a LinkedIn post. What they don't know is what you'd cut.
Same thing in ghostwriting. The useful brief isn't "be professional but conversational." It's the list of phrases that sound fine and still aren't you.
Context stops AI averaging you out.
Taste is what decides what never gets near the page.
Taste.md is the one I'd fight hardest for. We built something similar for my own writing, and the part doing the most work is the list of phrases that sound fine but aren't how I talk. The model needs to know what to throw away as much as what to aim for.
This is so good, Kieran. Saving it to read again.
"I think people are overcomplicating AI writing"
Totally agree with this. An article without a POV lacks depth and won't interest readers. The key takeaway is how you can use AI to outline and sharpen your thinking.
Thanks for the insights!