A trend of using generative AI to rewrite the software inside smart TVs and home appliances has taken hold online, and a GitHub tool that promises to reverse-engineer anything has drawn more than 420,000 stars, raising concern about the security of code users do not understand.
Rewriting the living room TV
One user wrote on X that they had Claude Code, Anthropic's AI coding assistant, generate new code for their parents' TV. They said that after being told the model number, Claude Code identified the set over Wi-Fi, removed 25 Xiaomi packages and wrote a new launcher.
“It's just like an Apple TV now (lol).”the user wrote on X, smart TV owner
Another user reported reviving a Lululemon Mirror, a workout mirror made by the Canadian company Lululemon that is no longer sold, and turning it into a smart mirror. They bought one secondhand and modified its software following a method published on GitHub.
The user also built in a Raspberry Pi, a small single-board computer, to control the mirror and added the ability to play their own music and video. They posted photos of the mirror displaying the time and a weather forecast on its surface.
A tool for any codebase
Modifying code like this once demanded skills and a great deal of time and effort. AI has changed that.
REA, short for Reverse Engineer Anything, is one example. Registering it with an AI agent and giving it instructions is enough to analyze how existing software works, according to its project page.
The page says REA analyzes an app even without source code, explains how its functions work, shows the evidence and builds a version for the user's own project. The project can be used with agents including Claude Code, Codex, Cursor, Gemini CLI and Grok Build.
Its targets include websites, JavaScript, Android and Windows apps. The official site cites examples such as adjusting the speed of a game in Chrome and changing what the buttons do in the Windows calculator.
Bugs nobody can fix
Some warn that a flood of low-quality code modifications could lead to serious security problems.
“Software reproduced with AI often contains bugs that the person who used the AI does not understand, or bugs they do not know how to fix.”Jason Koebler, a reporter at the US outlet 404 Media, said
If such software spreads, Koebler said, "vulnerable software gets installed on important hardware, leading to devices and networks being hacked." He also predicted that "hackers will break into this process and try to plant malware in the code."
“Even if you don't use AI, or don't think AI is any good, you won't escape its effects.”Koebler warned