Skip to content

Instantly share code, notes, and snippets.

View paddor's full-sized avatar

paddor

View GitHub Profile
@Shnatsel
Shnatsel / The unreasonable effectiveness of LLMs for auditing Rust code.md
Created June 20, 2026 18:31
The unreasonable effectiveness of LLMs for auditing Rust code

This is a mirror of my article which Medium apparently refuses to show to some people.

The unreasonable effectiveness of LLMs for auditing Rust code

As a lead of the Rust Secure Code Working Group, I got free access to GPT-5.5 via the Codex for Open Source. Since then I’ve found and reported dozens of issues of varying severity in widely used Rust crates.

Separately, the Rust Foundation security initiative got access to Mythos via Project Glasswing, and their report should also be coming soon. I’ve coordinated with them so that our audit targets would not overlap.

While I haven’t found any truly devastating vulnerabilities, I am very impressed with GPT-5.5 for aud

@codekitchen
codekitchen / 1brc_ractors.rb
Last active January 15, 2024 19:45
Exploring the 1BRC in Ruby with Ractors
#!/usr/bin/env ruby --yjit
WORKER_THREADS = 4
CHUNK_SIZE = 2**16
# BUG: my city struct is being corrupted when moving from the worker to the main ractor.
# `#<struct City min=-11.3, tot=24088.30000000004, max=64.6, n=1164>` becomes
# `#<struct City min=-11.3, tot=24088.30000000004, max=64.6, n=nil>`, note that each `n` attribute becomes `nil`.
# https://bugs.ruby-lang.org/issues/20165
# I tried changing the Struct to a simple array and still using `move: true`,