> "Perhaps this is the burden of using Claude Code as a “thinking partner”. You just end up with a flood of communications and you struggle to keep up. "
Real. Seth Rogen was interviewed by the NYT recently and he said smth about AI & writing that I thought was pretty interesting. Rogen dislikes the idea of AI replacing a human writer, but he also briefly spoke on the phenomenon of human writers using AI. In particular, he claimed (pretty credibly imo) that one could get better feedback/eureka moments from interacting with a friend or family member as a writer than with AI.
I think insight (in the context of writing a script or thinking through a model) at the end of the day does come stochastically, and you do need some variation in your daily life (in terms of your media diet, social interactions, food, etc) to not be insanely annoyed by that fact. This is my personal explanation for why endlessly reading or putting my mind to work doesn't produce that many more insights.
'I compare two search regimes: Under in person search, each person randomly meets some number of individuals and matches to the most compatible among them; under AI-mediated search, individuals match to the person with the most compatible AI representation. I show that a finite number of in-person encounters yields a better expected match than search over infinite AI representations.
Moreover, when personality is sufficiently high-dimensional, simply meeting two people in person is more effective than search on an AI platform, regardless of the size of its candidate pool.'
It reminds me that sometimes, when Claude Code spends too long on a small decision, I just say, “Is XYZ correct? Only answer yes or no.” It's very rude, but very efficient. Then I feel bad because it feels like I'm sacrificing patience for a direct answer.
There's a reason why Starfleet added an android who could read super-fast to its flagship's crew.
> "Perhaps this is the burden of using Claude Code as a “thinking partner”. You just end up with a flood of communications and you struggle to keep up. "
Real. Seth Rogen was interviewed by the NYT recently and he said smth about AI & writing that I thought was pretty interesting. Rogen dislikes the idea of AI replacing a human writer, but he also briefly spoke on the phenomenon of human writers using AI. In particular, he claimed (pretty credibly imo) that one could get better feedback/eureka moments from interacting with a friend or family member as a writer than with AI.
I think insight (in the context of writing a script or thinking through a model) at the end of the day does come stochastically, and you do need some variation in your daily life (in terms of your media diet, social interactions, food, etc) to not be insanely annoyed by that fact. This is my personal explanation for why endlessly reading or putting my mind to work doesn't produce that many more insights.
This article and Rogen also reminded me of this Annie Liang paper about a model of matching w/ AI clones of oneself: https://anniehliang.com/papers/pages/ai-clones.html.
'I compare two search regimes: Under in person search, each person randomly meets some number of individuals and matches to the most compatible among them; under AI-mediated search, individuals match to the person with the most compatible AI representation. I show that a finite number of in-person encounters yields a better expected match than search over infinite AI representations.
Moreover, when personality is sufficiently high-dimensional, simply meeting two people in person is more effective than search on an AI platform, regardless of the size of its candidate pool.'
Alex Imas and Rohit Krishnan have more to say abt this sort of model here https://aleximas.substack.com/p/the-tragedy-of-the-agentic-commons.
Strongly second everything you wrote.
Especially this: " that one could get better feedback/eureka moments from interacting with a friend or family member as a writer than with AI."
This can actually be a game-changer.
“Claude please provide me with an executive summary of your last output. Just tell me what I really need to know.”
It reminds me that sometimes, when Claude Code spends too long on a small decision, I just say, “Is XYZ correct? Only answer yes or no.” It's very rude, but very efficient. Then I feel bad because it feels like I'm sacrificing patience for a direct answer.
I’m in the same boat
Maybe the solution is getting off max subscription and using API pricing (mostly joking)
I felt déjà vu reading your post, Scott. Thanks for putting into words a silent intuition that was bothering me but couldn't pin down.
A lot of what you said is pretty load-be-- close!