LASRecomp: AI-Powered Static Recompilation of Zelda Link's Awakening from Switch to Native PC

Key Takeaways

• LASRecomp is a static recompilation of The Legend of Zelda: Link's Awakening (Nintendo Switch, 2019) that runs as a native application on Windows, macOS, and Linux — not an emulator.
• The project was made almost exclusively with Anthropic’s Claude (via Claude Code). Nearly all code, tooling, reverse engineering, and documentation were written by the AI; the human developer mainly directed, tested, and reviewed.
• No Nintendo game files are included. Users must supply their own legally obtained v1.0.0 extracted files. Recompilation happens locally on first launch.
• Features include unlocked aspect ratios (up to ultrawide), a built-in randomizer, Metal (macOS) and Vulkan (Windows/Linux) backends, controller remapping, and more.
• It is still in beta. Crashes and graphics issues remain, especially with certain actions and wide-aspect modes.

What Is LASRecomp?

LASRecomp (also called LAS Recomp) is an open-source project that converts the 2019 Nintendo Switch remake of The Legend of Zelda: Link’s Awakening into a native desktop application through static recompilation.

Unlike traditional Switch emulators, it does not rely on a full software emulation layer. The game’s executable code is recompiled into native host code, while graphics calls are translated to Metal on macOS (with MetalFX) and Vulkan on Windows and Linux.

The project’s README states clearly:

This project was made almost exclusively with AI. Nearly all of the code, tooling, reverse engineering and documentation in this repository was written by and will be maintained by an AI (Anthropic’s Claude, through Claude Code), directed, tested and reviewed by a human.

Developer peiikai handled task decomposition, prompting, testing, and iteration. This human-directed, AI-executed workflow has drawn significant community attention as an early example of AI-assisted reverse engineering and porting.

How Static Recompilation Works

Static recompilation analyzes the original Switch AArch64 machine code and translates it into C that can be compiled and executed natively on the target platform. A custom runtime layer then handles system services, graphics, input, and resource management.

Key technical details include:

• After the user provides their own v1.0.0 extracted game files, the first launch runs a local generator (using Unicorn and Capstone) plus a bundled Zig compiler to produce a shared library of the game code.
• The graphics layer translates Switch NVN commands into Metal or Vulkan.
• High-level emulation (HLE) replaces Switch OS services, while the core game logic runs natively.
• The repository and release builds contain no Nintendo assets. All game files must be supplied by the user.

Supported platforms (as of the current beta):

• macOS: Apple Silicon only, Metal + MetalFX
• Windows: x86-64, Vulkan (requires AVX2 CPU and Vulkan 1.3)
• Linux: x86-64 AppImage (Vulkan); aarch64 support is experimental

Features and Enhancements

Beyond native execution, LASRecomp adds several desktop-oriented improvements:

• Unlocked aspect ratio: Wider or taller windows reveal more of the island while houses and dungeons retain their original framing. The classic 16:9 mode remains available.
• Built-in randomizer: Item shuffle, trade quests, entrance randomization, dungeon room shuffler (multiple modes), Potsanity, Bushsanity, and more, with seed IDs and per-seed saves.
• Graphics options: Adjustable internal resolution, dynamic resolution, blur removal, and MetalFX upscaling on macOS.
• Controls and UI: Fully remappable keyboard and controller support, dynamic button prompts (Xbox, PlayStation, or Nintendo), and an in-game menu (F1 or controller Home button).
• Additional options: Multi-language support, mirror mode, and other quality-of-life features added across beta releases.

Current Status and Known Issues

The project is explicitly labeled a beta. Official notes warn that bugs, crashes, and unsolvable randomizer seeds (especially in certain dungeon room shuffle modes) are possible.

Documented issues include:

• Graphics glitches on wide and ultrawide aspect ratios (black boxes, lighting lines, shadow problems in large rooms).
• Specific crashes, such as checking the dresser in BowWow’s mom’s room or clicking “Title Screen” in the custom settings menu.
• Occasional controller issues on the start screen.

Community feedback indicates that basic playthroughs are viable for many users, but edge cases and heavily randomized seeds still require caution. Vulkan performance on Windows and Linux has received optimizations; frame-rate limitations no longer force slow-motion on those platforms (macOS improvements followed in later patches).

Legal and Usage Requirements

Users must supply their own legally obtained copy of the v1.0.0 Switch version. The repository strictly prohibits committing any Nintendo-owned files, extracted data, generated code, or fonts. Early discussion around encryption-related tools led to their removal to reduce legal risk.

The project provides no instructions for obtaining game files or keys. Possession of a legitimate copy is required.

Community and Resources

• GitHub: peiikai/LASRecomp
• Discord server for bug reports and discussion
• Releases include macOS DMG, Windows zip, and Linux AppImage builds (beta updates continue)

Conclusion

LASRecomp demonstrates the current practical reach of AI-assisted software engineering. With clear human direction and rigorous testing, models such as Claude can handle substantial reverse-engineering, code generation, and documentation work, producing a functional cross-platform native application.

It is not an official port and remains a beta with known crashes and compatibility gaps. For players who already own a legal copy of the Switch Link’s Awakening remake and want native desktop performance plus extras such as ultrawide support and a built-in randomizer, it offers an interesting experimental option.

Check the latest beta on the official repository, review the known-issues list, and test in a controlled environment before relying on it. Continued progress on crash reduction and platform stability will help determine the longer-term viability of this AI-assisted recompilation approach.