In recent weeks, native PC ports of games originally made for consoles have become increasingly common. The reason is not a single breakthrough but a combination of mature tools and AI agents able to follow complex instructions for hours without stopping.
At the centre of this acceleration is Reverse Engineer Anything (REA), a project that does not do the work by itself: it is a mix of toolkit and agent harness, linking under one roof tools that were previously separate, such as binary analyzers, decompilers, database readers and firmware image tools. These tools are made available to the AI agent of the user's choice, along with skill files containing additional instructions. The result is glue that significantly cuts the time developers spend on repetitive chores while reverse-engineering code.
The groundwork was laid years ago. Console emulation on PC emerged in the 1990s with 2D games and moved on to 3D consoles such as the PlayStation 1 and Nintendo 64. Through the early 2020s, emulators and toolkits grew rapidly, covering more systems than ever, and hardware became easier to take apart and inspect, helped by cheap and powerful devices such as the Raspberry Pi. The ultimate goal, fully decompiling and recompiling a game so it runs as native PC code rather than through emulation, was within reach. Only one barrier remained: dedicated development time.
Reverse engineering is tedious and cannot be done half-heartedly: one errant byte is a bug. Modern games can contain millions of lines of code, and every step requires checking which files are read, which calls are made to the operating system and what memory is allocated. This kind of meticulous, repetitive work is exactly what computers excel at, even non-deterministic ones such as LLMs. This generation of AI models, including the GPT-6 and Claude 5.5 families, follows complicated instructions more closely and keeps better track of its tasks, granting a higher degree of autonomy. Their high token efficiency also makes them affordable enough, if not cheap, for a project as large as taking apart an entire game.
Given that AI models can hallucinate and produce invented code, how does such a project get done? The mechanism is simple: the agent hands over decompiled source code and recompiles it to check whether it matches the original byte for byte. Even when it errs, the agent can retry until it gets the result right, and developers can step in when needed.
Compared with before, when developers had to juggle several tools across separate terminal windows, the agent can now manipulate bits and have programs talk to each other on their behalf, as long as it receives the right orders, without needing breaks.
Cost remains a real factor. With relatively simple prompts it is possible to explore and break down a game or application, provided you have one of the pricier agent subscriptions or can absorb high token bills. The reconstruction of Modern Warfare 2 is an example: the project took up to 700 billion tokens. According to the article, its developer could have saved time and tokens by using REA.
No comments yet. Be the first!