Over the weekend, users began replacing the operating systems on smart TVs with AI-generated alternatives to strip out bloatware and advertisements. This trend has spread to smart picture frames, baby monitors, and fitness mirrors, effectively hacking devices that manufacturers had locked down or abandoned.
The rise of AI cloning
While “vibe coding” usually refers to building software from scratch, this movement involves using AI to reverse-engineer existing applications. Users feed the model details of a proprietary product, and the AI attempts to replicate its features or install open-source software onto the hardware. The process mixes legitimate reverse-engineering with what critics call plagiarism.
Mike Nolan, who studies the politics of open source software, told 404 Media that the business model for software-as-a-service is collapsing. He noted that recreating desired functionality has become trivially easy.
“The barrier for recreating this stuff is lower, so a lot of people are doing it for fun,” Nolan said. “But there’s a lot of work associated with maintaining software so there may be a lot of instances where the software may not be reliable enough to use.”
The current wave is driven by a GitHub project called “Reverse Engineer Anything” (REA). As of this writing, the repository has more than 342,000 stars. The tool allows users to inspect an application without its source code, explain how a specific feature works, and build a version for their own project.
This capability mirrors recent viral clones of Adobe Creative Cloud, Microsoft Office, and various video games. It is now being applied to embedded software on physical devices.
One maker of guitar pedals claimed on Hacker News that he used AI to decompile the algorithms of three competitors in 72 hours. He suggested he could simply put the stolen code in a box and resell the product.
“Basically, anyone with a Claude subscription can now rip off any best-in-class algorithm from any competitor,” he wrote.
The quality problem
Previous discussions on AI-generated decompilations of video games noted that the resulting code often contained bugs and did not function as intended. That issue persists here.
AI-generated software frequently contains errors that the user does not understand or know how to fix. People are deploying this code onto the internet and into communities at a speed that prevents proper vetting or moderation.
Even if most individuals do not hack their own baby monitors, the existence of this software affects everyone. Poorly coded, insecure applications are being installed on critical hardware, creating risks for networks and devices. Open source projects are facing an influx of unreliable code from users who treat the technology with blind enthusiasm.
What it means
This shift is enabled by the “agentification” of AI and the removal of safety guardrails. Models can now be instructed to perform complex tasks autonomously. The result will likely be an onslaught of semi-functional hardware and software slop.
Expect hastily created startups to launch with broken software and consumers to interact with touch screens that contain bizarre bugs. Hackers may also infiltrate this process to inject malware into the cloned code. Meanwhile, smaller companies face the threat of being undercut by competitors who used AI to steal their products.
There is a counter-movement, however. Users are also likely to use decompiled software to remove manufacturer guardrails on products that should not have existed in the first place, such as removing ads from baby monitors or unlocking locked-out features on televisions.
