AI tool REA cracks paid software: Is the software industry's business model coming to an end?
It was predicted, just in a slightly different way: rather than having to buy software in the long run, you'd simply have your AI agents program the features you need. Now there's another development: take an expensive paid app on your PC, run REA on it, and your AI can learn the professional tool's entire codebase and functionality before rebuilding it as open-source software.
A few days ago, we reported on programmer and entrepreneur Brandon Thomas, who, frustrated with Adobe, decided to recreate almost the entire Adobe suite as open-source software under the project name ArtCraft, making it freely available to everyone. This includes Premiere, After Effects, and more. Thomas has years of programming experience, but what if anyone could do this soon?
That's where the new open-source tool REA comes in. The acronym stands for Reverse Engineer Anything. REA connects AI coding agents such as Claude Code and Codex to reverse-engineering tools like Ghidra, Hopper, and IDA through the Model Context Protocol (MCP).
What does REA do, and how does it work?
REA analyzes software based on instructions given in natural language to understand how it works, even when the original source code isn't available. While REA can't always recover the original source code exactly, it can reconstruct readable code based on the software's functionality. This reconstruction gives an AI agent enough information to replicate parts of the software, ideally achieving comparable results with newly written code. The result is a custom implementation that delivers the same functionality.
Using a new codebase for the implementation could also make copyright enforcement more difficult in practice, since the original source code hasn't actually been stolen. And that's precisely where the potential for market disruption lies. Functionally recreating software is legally different from copying its protected source code.
Why is this relevant?
REA significantly lowers the barrier to entry, and this app is almost certainly just the beginning. Instead of having to understand everything yourself, you can ask the AI questions and give it tasks in natural language. However, a fair amount of technical knowledge is still required for now, including installation, working with the tools, and verifying the results. That said, YouTube and other platforms promise easy-to-follow tutorials that can help users get around some of these hurdles.
More accessible reverse engineering could mark a major technological leap. Recreating software is becoming cheaper. But that doesn't automatically mean smaller players will necessarily benefit. At the same time, competition could intensify, shifting the focus away from features and exclusive code toward usability, reliability, data, ecosystems, customer support, trust, and ongoing development. This is where large corporations have an advantage, potentially putting even more pressure on smaller developers.
Does it work, and if so, how easy is it?
I didn't try REA with the powerful Claude Code because it's too expensive for me. Instead, I'm using the ad-supported, free app Freebuff (don't provide sensible data!).
I instructed the AI to install REA and all its dependencies, along with MCP and Ghidra. From there, the entire process ran automatically, and I didn't have to do anything else.
Next, I wanted to recreate an app. It shouldn't be as complicated as a game, but not as simple as a calculator either. The VLC Media Player came to mind, so I entered the following prompt:
The agent got to work, asked me three more questions, and finished after about 10 to 15 minutes. I had my own VLC clone. I was also informed that it used Chromium rather than libVLC as its engine, meaning some video formats wouldn't work. That could be changed, though.
The “Own Player” worked right from the start. I simply dragged an MP3 file into the window, and playback began immediately. I could also open folders, although subfolders weren't included. Keyboard shortcuts for seeking, muting, and other functions worked perfectly. The only issue was that fullscreen mode displayed a black screen.
Overall, though, the result was impressive for just a few minutes of work. Once everything was installed, virtually all it took to create an app was a single prompt and answers to three multiple-choice questions.
Conclusion
Reverse engineering is nothing new. But AI is lowering the barriers to entry and making the process increasingly straightforward, even for non-coders. And that's precisely what could put software copyright law to the test: if anyone can easily recreate their own apps from paid software, this could threaten software vendors' business models.
Not every copy is legal, of course, but under certain conditions, functionality can be recreated without using the original code. As a result, the code that software vendors spent so much money developing could lose some of its value as a competitive advantage, even if their copyrights remain untouched. This could reshape the software market in the long term, potentially in disruptive ways.
In the past, you bought software and owned a copy. Then came the subscription era, when software was licensed for use rather than sold outright. If everyone can build their own apps—or recreate existing ones—the software licensing model will likely be affected as well.
Add as a preferred source on Google
