WIP - Expression Type Hierarchy
manticore.dev
WIP - Expression Type Hierarchy
@manticore.dev
coming soon...
Yes but other way! Hurtboxes take the damage
December 22, 2025 at 5:42 AM
All hurtboxes are hitboxes, but not all hitboxes are hurtboxes.

Hurtboxes receive damage
December 22, 2025 at 5:41 AM
Someone will write a generalized NtoN+1 based on induction and we’ll solve Python
December 15, 2025 at 4:39 AM
Really cool
December 15, 2025 at 4:33 AM
One more mystery down, hopefully not much more than one to go..
September 19, 2025 at 10:49 PM
Aha, I got it :) you can perform such a shift arithmetically! An interesting thing is that right-shifting a transfinite expression reduces its order, so this operation is a rightshift on finites, but a rightshift plus a compensation on transfinites, to make up for the order decrement
September 19, 2025 at 10:48 PM
It takes a lot of mental effort to consider but it doesn’t seem insurmountable either. You just have to preserve
1. Order (ordinality)
2. Sequence of cardinalities under ceiling division

I know that none of this makes sense without context but oh well
September 17, 2025 at 2:55 AM
Take a sample of large, randomly generated Python programs with arbitrarily named variables, and translate them to Go, and back

Now do the same with English and Spanish

The Python-Go sample is practically guaranteed to contain more errors, because, like you said, it’s testable
July 21, 2025 at 8:39 PM
You’re of course correct that “yeet the doggo” will never be unambiguous, but that’s not a disadvantage for an LLM, it’s an advantage!

“Yeet the doggo” is error tolerant. Doggo could be replaced by a number of words

“a+b” is not error-tolerant at all
July 21, 2025 at 8:37 PM
I’m trying to illustrate the difference between natural and programming languages.

Obviously, a+b must have one meaning or the Python interpreter wouldn’t know what to do. All valid Python programs are unambiguous.

“a+b” in a vacuum is NOT a valid Python program. It gives an error
July 21, 2025 at 8:35 PM
That’s like saying “foo” can mean literally anything because you can set it equal to whatever you want. Still, it only ever means one thing at a time.

From our perspective, we don’t know whether or not “a+b” is overloaded. But from the interpreter’s perspective, the answer is a strict yes or no
July 21, 2025 at 6:46 PM
If you’re getting good results, I won’t argue with that, but a chunk of Python code certainly can’t mean multiple things!
July 21, 2025 at 5:31 PM
A sentence of English or Spanish can have multiple possible interpretations. A piece of Python or Go code only ever means one thing

An LLM has a lot more leeway when working with a natural language than a programming language
July 20, 2025 at 5:10 AM
Sometimes you do want to throw though
July 20, 2025 at 5:03 AM
Now THIS is contrarian
July 19, 2025 at 9:34 AM
I do it stinks
June 5, 2025 at 1:10 AM
Programming languages are much stricter, like algebra. It’s easier to create something meaningless by placing random algebraic symbols than by placing random words
June 5, 2025 at 12:53 AM
I know I’m not the only one sick of nominal typing
June 5, 2025 at 12:47 AM