Conversation

There's a distinction that's probably only obvious to people with some grad-level math: Corner cases are not the same as rare cases. Neither implies the other. "Corners" are a measure-0 subset of a space where open-nbhd/non-singleton methods don't apply Ie "special treatment"
Not necessarily, when the majority is corners, we switch language and start talking about how it's a diverse/varied context. Like say the food scene in a cosmopolitan city with lots of food carts.
1
You assume abstractability is entirely a function of the intelligence of the agent acting on the context. I disagree with that premise. It's fundamentally a propert of the domain. Tax law is less abstractable than classical Newtonian mechanics.
That's not what I'm talking about. I'm talking about say writing tax software for existing taxes, not legislation. It's got high Kolmogorov complexity. You can't write elegant code for it because it's a bundle of special exemptions and stuff.
You're conflating normative concerns ("good" tax laws) with phenomenological. Alt example with no normative angle: special-order catalogs are harder to abstract than combinatorial product offerings like Starbucks menu.
Replying to
Also obvious to people who play competitive tabletop games. Manifest and double faced cards in the same Standard environment was quite a treat.
1
1