Macintosh System 7 ported to x86 with LLM help (hackaday.com)

Kelsi Davis recently reverse-engineered and ported Apple’s classic Macintosh System 7 — originally built for Motorola 68K CPUs — to the x86 architecture using a combination of Ghidra and a large language model. Working from binaries (no source), she used the LLM to speed up disassembly, pattern recognition, and translation tasks, producing a functional System 7 desktop with Finder and GUI elements running under QEMU. The repo and artifacts are on GitHub, and Davis says the LLM helped compress what she expected to be a multi-year project into about three days of work. Beyond the novelty, this demonstrates LLMs as practical force multipliers for complex reverse-engineering and porting work: they can accelerate identification of calling conventions, ABI mismatches, and binary idioms when moving between different ISAs (68K’s big-endian, CISC-like model versus x86’s little-endian semantics). The result highlights opportunities for faster software preservation, compatibility projects, and security research, while also raising questions about correctness, maintainability, and legal/ethical limits when repurposing proprietary binaries. Technical details to explore further include how system traps, hardware emulation, and OS-level abstractions were mapped from 68K to x86 and how much manual patching remained after LLM assistance.