Evidence at a glance #
Based on the linked sources. The author's reported result and independent confirmation are separate checks.
- AI involvement
- Model not reported. Tools not reported. The writeup below describes the work attributed to AI.
- Reported result
- Research. This is the directory's source-based outcome classification.
- Independent confirmation
- Not reported.
- Human work and limitations
- The source video title explicitly mentions the 84-day duration for a whole game. The methodology is described as a loop involving AI generation, compilation, and byte-level verification.
Original sources
What was done #
AI-assisted matching decompilation uses an automated loop where an AI generates code guesses, which are then compiled and compared against original binary bytes to verify function reconstruction.
The project, documented by Chris Lewis, utilizes an AI-assisted matching decompilation workflow to reverse engineer the game Snowboard Kids. The method relies on a "guess, compile, compare" loop. In this process, an AI model generates a candidate source code guess for a specific function. This guess is then processed by the original compiler. The resulting binary output is compared against the original game's binary bytes; if the bytes match, the function is considered successfully solved. This iterative approach allows for the systematic reconstruction of game functions. The author reports that this specific implementation required 84 days to complete the decompilation of the entire game. The process is categorized under game development and decompilation research.
Sources: Source 1