I've had incredibly productive discussions with Claude about category theory. (I prefer Claude because it's the most pleasant to talk with; I think they optimized for that.)
The ability to explain what I know already, hand-wave at what I think I understand about my question, and then get a description that meets me where I'm at is invaluable.
Sure, occasionally Claude will tell me (incorrectly) that a CRDT's lattice operation needs an identity function: you absolutely have to go back and forth with wikipedia.
LLMs are not a magic genie or oracle. But if you use them for what they're good at, they're amazing.
I do too, but sometimes I read a blog post that makes me wonder about something and I don't want to schedule a 1:1 with them and then wait to chat about it. We're mostly remote now…
The ability to explain what I know already, hand-wave at what I think I understand about my question, and then get a description that meets me where I'm at is invaluable.
Sure, occasionally Claude will tell me (incorrectly) that a CRDT's lattice operation needs an identity function: you absolutely have to go back and forth with wikipedia.
LLMs are not a magic genie or oracle. But if you use them for what they're good at, they're amazing.