Seong-Heon Jung (정성헌)

38 Followers
47 Following
60 Posts
Pronounshe/him
PronunciationSung (sing/sang/*sung) Hun (*hon*ey)
@markusde do people consider aesop, grind, and other proof automation tools as autoformalization too?

The shrubbery grouping rules are very subtle. Namely, the relationship between newlines, indentation, separators, and opener-closer pairs get really crazy. (1,\n 2, 3), (\n1, 2, 3) (1:\n 2,\n 3) work, but (1,\n2,3) does not.

The documentation does not help too much either here because unlike the tokenization guide, the guide about grouping is mostly informal prose.

I'm a big fan of Rhombus' "macros without parentheses", and would love to implement it for my language. Unfortunately, shrubbery notation is just too damn hard to parse correctly, even with a fair number of simplifications. I would have really appreciated some pointers from the paper and documentation about some high-level strategies. Yes, I could technically just use #lang shrubbery, but I'd like to use it outside of racket too
@markusde learn the beautiful language of toronto
I wish I could get the QoL improvements of mathlib like `lemma` or `ring` without also having to download and compile a significant subset of all of the mechanized mathematics that exist in the world.

Pleased to live in a place where this is true (enough):

Eight states received “A” grades for LGBTQ+ safety based on their comprehensive pro-equality laws and low rates of hate crimes against LGBTQ+ people. **Only Rhode Island nabbed an A+ mark.**
https://www.safehome.org/data-lgbtq-state-safety-rankings/

2025 LGBTQ+ State Safety Report Cards | SafeHome.org

Eight states got A's for LGBTQ+ safety, with Rhode Island topping the list. Florida ranked lowest, with a failing grade due to discriminatory laws.

SafeHome.org
@markusde I didn't know this is what you meant by verifying randomized algorithms
I love that Lean supports Windows so well. I run windows on my personal laptop, and it's great to be able to program without WSL. WSL is good, but nowhere near as smooth as native. I compiled a kernel from souce for my WSL to use some additional modules not included with the default WSL kernel, and that has led to WSL shutting down my laptop without warning.

Another cool Lean exercise: a verified implementation of the Boyer-Moore majority voting algorithm.
It's an O(n) time O(1) space majority element detection algorithm.

The 1981 verification of took the authors 20 hours. In contrast, I spent no more than 4 hours (including building the CLI tool) thanks to proof automation techniques like grind and omega.

https://github.com/Forthoney/mjrty

GitHub - Forthoney/mjrty: Verified Boyer-Moore majority algorithm

Verified Boyer-Moore majority algorithm. Contribute to Forthoney/mjrty development by creating an account on GitHub.

GitHub
@markusde mastodon is your co-"lab notebook"