Successfully Ported 1993 Amiga Assembly Game to Godot and Verified Original Binary with Claude Code
Key point
Claude Code analyzed 70,000 lines of 68000 assembly code, ported it to the Godot engine, and verified byte-for-byte consistency with the original binary.
Details
A case study was released showing the successful porting of 72,758 lines of 68000 assembly code from the 1993 Amiga game 'Babylonian Twins', developed in Iraq, to the Godot engine using Claude Code.
Assembly Code Analysis and Binary Verification
The AI reverse-engineered undocumented legacy code to interpret complex hardware control logic, including level map encoding, object tables, and Copper lists. In particular, it automated the process of reassembling the 1993 source code using the vasm assembler to verify that it was byte-identical to the released binary. This resolved discrepancies between 30-year-old memory snapshot data and uninitialized variables, and matched all five levels pixel-perfectly, including Copper effects such as sky gradients and water color cycles.
AI-Based Porting Workflow and Results
The porting process proceeded as follows:
- Step 1 (C++ Port): Migrated 34,000 lines of C++ code written in 2010 to Godot 4 overnight, generating a playable build.
- Step 2 (Assembly Port): Rewrote the original assembly behavior in GDScript to maintain 50Hz timing and integrated it into the Steam version.
- Verification Automation: Added command-line flags such as
--probeand--screenshotto convert game physics and rendering results into machine-readable data, and confirmed visual accuracy through screenshot comparison.
This project demonstrated that AI coding tools are effective not only for simple code generation but also for complex reverse engineering tasks, such as understanding the structure of low-level assembly code with minimal comments and migrating it to modern engines. The final product is live on iOS/Android and available as a free demo on Steam.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.