The Reverse Engineer's Guide to Mechanistic Interpretability
We reverse engineer a neural network the only way we know how: by writing the world's worst interpretability tool from scratch in Go, and pointing it at GPT-2's memory.
fuzzing.science
mutation, selection, and the bugs that emerge
Latest posts
Babysitting an Army of Monkeys 2: Planet of the Agents
What happens when Charlie Miller's dumb, honest monkeys learn to reason, and to lie. Babysitting used to mean counting crashes; with LLM agents it means refereeing liars and rebuilding the free oracle the kernel used to give you.
Chronometry - a transparent and cryptographically verifiable proof-of-hack signature store
In this blog post, I will introduce Chronometry, a transparent and cryptographically verifiable proof-of-hack signature store.
Rudroid - Writing the World's worst Android Emulator in Rust 🦀
In this blog post, we'll write an emulator that can run a 'Hello World' Android ELF binary. While doing this, we will learn how to go about writing our own emulators.
Google Project Zero Issue Downloader
In this blog post, I will be talking about a small script to download/monitor issues from @ProjectZeroBugs issues 🐞.
ManuFuzzer - Binary code-coverage fuzzer for macOS, based on libFuzzer and LLVM
In this blog post, I will be talking about ManuFuzzer, a binary code-coverage fuzzer for macOS, based on libFuzzer and LLVM.
Fuzzing Android Native libraries with libFuzzer + QEMU 🦥
In this blog post, I will go through the process of why and how I built a new Android fuzzing framework called `Sloth` 🦥
Crashmon - A LLDB Based replacement for CrashWrangler
In this blog post, I will be talking about Crashmon, a LLDB Based replacement for CrashWrangler.