Similar thing happened to me when I uploaded my entire second brain notes archive from over a five year period. It surface patterns that I couldn't necessarily see, but immediately resonated.
I did this between Christmas and New Year's and some of it was so silly, other part were enlightening, I cut some of the bad habits out, it was a wakeup call I needed!
Yeah, I wouldn't recommend training on ONLY this data. You're gonna get some pretty unhinged outputs.
However, for me, this exercise if what filled in most of the missing details. My outputs went from, "yeah, that generally sounds like me" to "Woah! That's totally me!"
The best way to train AI. I love when it gets to the point and says, “You would not say that; I have to push back,” or, “That sounds like you’re hiding behind performance metrics again.” Claude is the best at calling that out. I was literally just thinking about feeding it all my texts. This is awesome, Nick.
Similar thing happened to me when I uploaded my entire second brain notes archive from over a five year period. It surface patterns that I couldn't necessarily see, but immediately resonated.
Oh I love that. What did it surface that you couldn't see yourself?
I actually wrote a post about it as well…https://infolosophy.substack.com/p/what-my-second-brain-was-really-keeping?r=ylp9v
I'm be giving this a read soon. Thanks for sharing it 💪
I did this between Christmas and New Year's and some of it was so silly, other part were enlightening, I cut some of the bad habits out, it was a wakeup call I needed!
Yeah, I wouldn't recommend training on ONLY this data. You're gonna get some pretty unhinged outputs.
However, for me, this exercise if what filled in most of the missing details. My outputs went from, "yeah, that generally sounds like me" to "Woah! That's totally me!"
That’s cool! LOL
Love the practicality of this - and the humor. Your authentic voice brought me wisdom and, maybe more importantly, joy today.
If the humor landed, the unpolished stuff is working. Appreciate you, Kathryn.
Absolutely brilliant. I gave Claude my sweatpants this morning. It's already paying dividends. Thank you!
This one came from noticing my best Claude outputs happened when I fed it screenshots of texts to my ex 🤦♂️
Mortifying to admit. But the voice was right in a way my "professional" samples never produced.
This is super interesting! Hope to try it someday
Let me know how it goes when you do 🙏
I envy those who journal for this exact use case.
Journals are great and I definitely do that too... but honestly texts might be better. Less performative.
You're not writing for future-you to admire. You're just... talking. What do you have that's like that?
I don’t think so but I do see the difference, definitely.
The best way to train AI. I love when it gets to the point and says, “You would not say that; I have to push back,” or, “That sounds like you’re hiding behind performance metrics again.” Claude is the best at calling that out. I was literally just thinking about feeding it all my texts. This is awesome, Nick.
Do it. Feed it your texts. Then come back and tell me what it surfaces.
I'm building a collection of "things people discovered about their own voice that they couldn't see" and yours would probably be fascinating.
Don’t make me add all my past dating texts or this will turn into a new project on online dysfunctionality. 😂
Oh lordy, I don't want the raw training data 🤣 I've sent more than my fair share of drunken texts.
Just the output this feed helps create will do for now.