11th October 2026
AI is getting damn good at decompiling software.
We are getting to the point where an agent costing a few dollars a month can help anyone turn an application or game into readable/editable source code.
In 1999, Napster changed the music industry. Music piracy wasn't new, but Napster removed the expertise and made it way easier. Suddenly, anyone could access practically any song ever recorded.
The music industry responded with lawsuits, copyright enforcement and technical restrictions, but none could reverse the underlying change: the cost of copying and distributing music had collapsed. The industry had to reinvent its business model.
I think we’re approaching a comparable moment for games, and potentially all software. Except this time the disruption could be more profound, because AI isn’t simply making software easier to copy. It’s making it easier to modify, remix and personalise.
For decades, commercial software has relied on a simple form of protection: complexity. Developers write source code, compile it and give customers the finished product.
Reverse engineering has always been possible, but reconstructing a sophisticated application would require years of work by an expert. Modern games contain millions of lines of code, proprietary file formats and custom graphics systems and it all needs to be decoded.
AI makes this task way easier, Claude/Codex can use the reverse-engineering tools and can help developers get back to something resembling the source code quickly.
That doesn’t mean anyone can remake a Photoshop or a PS5 game over a weekend. Encryption, proprietary assets, copy protection are still major obstacles. But the expertise required is now widely available.
In 2026, Open Community Edition showed Halo: Combat Evolved running inside a browser. The project also ports to Windows, Linux, Android and browsers through WebAssembly, with users providing their own game data rather than anyone re-distributing Microsoft’s original assets.
Something slightly different is happening with commercial applications. Brandon Thomas has created ArtCraft, seven open-source applications intended to reproduce functionality associated with Photoshop, Lightroom, Premiere Pro, Illustrator, After Effects, InDesign and Acrobat. The applications are pretty basic, compared to Adobe’s but the extraordinary part is that a single developer using AI can attempt to reproduce a family of professional applications.
These are actually two different threats. One is reverse engineering existing software; the other is making it dramatically easier to build something that does roughly the same thing. Both weaken an advantage established software companies have enjoyed for decades: their products are extremely expensive and difficult to reproduce.
Traditionally, software piracy has depended upon a relatively small number of technically skilled people cracking applications and distributing them. Publishers can then pursue those distributors, shut down websites. But what happens when the tools required to do this are available to everyone?
Instead of searching for somebody else’s modified version of a game, you could increasingly use an AI coding agent to create your own. You might analyse software you legitimately possess, recreate particular functionality or build a compatible implementation yourself.
There are legal limits. Owning software doesn’t automatically give you permission to circumvent its protections or reproduce copyrighted material. But if you never distribute, who would know?
AnyPS5 is an open-source project attempting to make PS5 games run on PCs. Rather than emulating every component of Sony’s console, it converts PlayStation executables and recreates the system libraries upon which they depend.
As of October 2026, the project reckons to be at 85% implementation of the system functions. The progress is remarkable.
A small community is systematically reconstructing enough of a modern console to run its software anywhere. Projects like this usually require years of highly specialised work, but increasingly capable AI agents can help identify undocumented behaviour and understand the low-level code.
It's not about if PS5 emulation suddenly works tomorrow. It’s what happens when tasks that once required decades, take weeks.
The timing is fascinating because GTA6 launches on 19 November 2026, initially for PlayStation 5 and Xbox. Few products better demonstrate the economics of modern blockbuster software: years of development, huge teams and huge amounts of money have gone into this release.
A fully functional unofficial PC version of GTA6 appearing within weeks would be an extraordinary achievement, given the encryption, proprietary technology and complexity involved. But serious attempts to dismantle and understand it will kick off immediately.
What happens when the number of people participating in this kind of thing increases to hundreds of thousands of developers armed with AI?
That changes the economics of console exclusivity, software licensing and technical protections.
This is where I think the argument becomes more interesting than piracy. Imagine you love a football game from the late 1990s. You like its simplicity and interface but want modern transfers, competitions and player psychology. Today, recreating that experience would require development expertise.
In the near future, you might simply describe what you want to an AI agent. It could analyse how the original game behaves and reproduce the elements you like and build new systems around them. You wouldn’t necessarily download somebody else’s remake. You’d create your own.
The same principle applies everywhere. You could recreate an abandoned Mac application so it works on a modern operating system, or build a simpler version of an bloated productivity tool.
This changes our relationship with software. For decades, users have accepted the decisions of developers. AI introduces a new option: make the software you want instead.
Everyone being able to make their own version is way more disruptive than piracy.