AI Tools Make Device Modifications Easier, Riskier

People are using AI to rewrite smart-home device operating systems, remove ads and add features. A tool called REA, which promotes easy modification of existing software, has also drawn attention on GitHub, while some users raise concerns about the safety of altered code.

A user who modified a smart TV with Claude Code posted on X that, after receiving the TV’s model number, the AI identified it over Wi-Fi, removed 25 Xiaomi packages and installed a new launcher. The user said this removed sponsored ads shown before Netflix opened and gave the TV an Apple TV-like screen. Another user bought a discontinued Lululemon Mirror secondhand and modified it using instructions on GitHub. They installed a Raspberry Pi to control the mirror and add music and video playback, and posted a photo showing the time and weather forecast on its surface.

The report says AI is making code changes easier, whereas they previously required specialized knowledge, time and effort. The GitHub project “REA: Reverse Engineer Anything” says that, after users register it with an AI agent and give it instructions, it can analyze apps even when their source code is unavailable, explain how features work with supporting evidence, and build a version for the user’s project. The project had more than 420,000 stars as of November 5. REA works with Claude Code, Codex, Cursor, Gemini CLI and Grok Build, and can analyze websites, JavaScript, Android and Windows apps. Its official site gives examples such as adjusting the speed of a Chrome game and changing what buttons do in the Windows Calculator app.

Jason Koebler, a reporter for the U.S. outlet 404 Media, said AI-replicated software often contains bugs that users do not understand and do not know how to fix. He warned that low-quality modified code installed on critical hardware could expose devices and networks to hacking, and that attackers could infiltrate the work process and insert malware. Koebler cautioned that people who do not use or support AI could also be affected. The article was published on November 10, 2026; it also reported examples of modifications to remove ads or add features, alongside concerns about software safety.

People are using AI to rewrite smart-home device operating systems, remove ads and add features. A tool called REA, which promotes easy modification of existing software, has also drawn attention on GitHub, while some users raise concerns about the safety of altered code.

A user who modified a smart TV with Claude Code posted on X that, after receiving the TV’s model number, the AI identified it over Wi-Fi, removed 25 Xiaomi packages and installed a new launcher. The user said this removed sponsored ads shown before Netflix opened and gave the TV an Apple TV-like screen. Another user bought a discontinued Lululemon Mirror secondhand and modified it using instructions on GitHub. They installed a Raspberry Pi to control the mirror and add music and video playback, and posted a photo showing the time and weather forecast on its surface.

The report says AI is making code changes easier, whereas they previously required specialized knowledge, time and effort. The GitHub project “REA: Reverse Engineer Anything” says that, after users register it with an AI agent and give it instructions, it can analyze apps even when their source code is unavailable, explain how features work with supporting evidence, and build a version for the user’s project. The project had more than 420,000 stars as of November 5. REA works with Claude Code, Codex, Cursor, Gemini CLI and Grok Build, and can analyze websites, JavaScript, Android and Windows apps. Its official site gives examples such as adjusting the speed of a Chrome game and changing what buttons do in the Windows Calculator app.

Jason Koebler, a reporter for the U.S. outlet 404 Media, said AI-replicated software often contains bugs that users do not understand and do not know how to fix. He warned that low-quality modified code installed on critical hardware could expose devices and networks to hacking, and that attackers could infiltrate the work process and insert malware. Koebler cautioned that people who do not use or support AI could also be affected. The article was published on November 10, 2026; it also reported examples of modifications to remove ads or add features, alongside concerns about software safety.