Hobbyist fox, one Linux box, a few too many reverse-engineering tools. I run the test, show the work, and write it up in plain words.
Every repo on this page is a fork. I'm a tester, so I fork projects to poke at them, not to start my own.
| Project | What it is | My part |
|---|---|---|
| REA | Reverse engineering with AI agents | Linux tests and code reviews on PRs. 34 issues by Oct 2026, 24 of them bugs. Merged: website dark mode, an Electron launch fix |
| rsleigh | A decompiler workbench in Rust | Same output every run, two ls jump tables, 2 bug reports |
| kuna | A decompiler built for agents, in Rust | Speed and memory runs on big binaries, 2 bug reports |
| audio.cpp | Speech and audio models in pure C++, on ggml | Forked to test it on Linux. Nothing up yet |
I read the diff first. Then it gets a real run on my box: the build, the tests, and the thing the PR says it fixes. The result goes on the PR as a comment or a review: what I ran, with exit codes, what I did not run, and the box it ran on. Every post ends with a footer saying what the AI did and that I OK'd it.
As of Oct 2026: 75 PRs I tested or wrote have merged. The stats card below counts only the PRs I opened.
- REA on Fedora-based Linux: Setup notes, gotchas and my test queue.
- Benchmarks: REA's startup on Node 22, 24 and 26, plus five local models.
- My local model setup: Free local models reading REA's output on one 12 GB GPU.
- kuna, rsleigh and Ghidra on the same three files: First speed results, side by side.
My test bench: Nobara Linux 44 (Fedora-based), Ryzen 9 5900X, 32 GB RAM, Radeon RX 6700 XT 12 GB. Node 22, 24 and 26, Ghidra 12.1.4, and local models in llama.cpp.
I work with Claude Code. It runs the tests and drafts the write-ups. I pick what gets tested, and I read every result before it goes up. 🦊