Going Commando decomp reaches 1.79% byte-matched, and AI is in the workflow
The open PS2 decomp of Ratchet & Clank: Going Commando (USA v1.01) reports 1.79% of loaded code matched. Not a playable PC port yet.
The PlayStation 2 decompilation of Ratchet & Clank: Going Commando is now public on GitHub, and its maintainers say 1.79% of the loaded code is recovered as C that matches the original byte for byte. The target is the USA v1.01 release (SCUS_972.68). This is not a playable PC port, and the project says so right at the top.
The number to watch is the loaded-code figure: boot code plus all 27 level overlays, 874,768 matched bytes out of 48,788,176, validated on 10 October 2026. A stricter count of unique EE code sits at 0.5818%, because the project uses a conservative grouping policy. Both numbers are small, and that’s normal. Byte-matching decomp is slow, careful work. A function either compiles to the exact original bytes or it doesn’t count.
What’s new is the workflow. The repo says its research, coding and tooling use AI assistance under human direction. It also says an AI answer alone is not evidence. A match needs compiler output that passes a byte-for-byte comparison against the pinned executable. Nothing from the game is distributed, so you need your own legally obtained copy.
Native PC execution is still unverified. Emulator observations and their limits live in a separate PCSX2 validation doc in the repo. The PAL release is registered but not pinned, so there are no matching proofs for it yet.
This is contributor territory, not a player download. The project offers a guided setup: you point a coding assistant at the repo, it checks your ISO and toolchain, and you submit changes through a fork and a draft PR that a maintainer reviews.
- Start a coding assistant that can edit files and run commands.
- Verify your ISO and the full tool suite.
- Submit the tested change as a draft PR for review.
My take, clearly opinion: plenty of AI-assisted projects promise a lot and ship demos. This one treats the AI as a helper and the byte comparison as the judge. That’s the right rule, and other vibe-coded projects should copy it. Watch whether the percentage keeps climbing steadily or stalls once the easy functions run out.
