A developer says he used Claude Code to clean up an Android smart TV and make it feel faster, according to Tom's Hardware. The report centers on Mert Cobanov, identified by the outlet as a Senior AI Engineer at Refik Anadol Studio, who posted on X about an “unbelievable” improvement in how his TV ran after the cleanup. According to Tom's Hardware, Cobanov said he opened developer options on the Android TV and instructed Claude to connect to the device via ADB. He said the agent then controlled the TV, deactivated apps, installed FLauncher, and shortened animation durations. Cobanov also said the process did not require root access. The strongest performance claim is Cobanov’s own: he asserted that the TV now runs smoother than it did when it was new four years ago. Tom's Hardware did not report independent benchmarks or a second test, so the result should be read as an individual account rather than proof that the same process will deliver similar gains across Android TV hardware. The report is still notable because it shows a practical use case for coding agents outside a conventional software repository. Instead of generating application code, the agent was used to operate through developer-facing controls and apply a sequence of system changes on a consumer device. Cobanov has also published a “Clean up your Android TV” site, Tom's Hardware reports. The outlet says the guide has four steps and includes the prompt Cobanov used, with the instructions framed for Claude Code or Codex. There are warnings attached to the process. According to Tom's Hardware, Cobanov says the guide does not uninstall apps; it disables them, which he says makes the steps reversible. He also says the agent keeps a list of what it disabled, allowing a user to ask it to re-enable everything it changed, and the guide includes instructions to restore individual apps. The broader context, as Tom's Hardware notes, is that many smart TVs ship with limited processors, memory, and storage, and can feel slower after app or operating-system updates. That has helped create demand for separate streaming boxes and HDMI sticks, but Cobanov’s example points to a different path: using an AI coding agent to remove or reduce software overhead on the TV itself. Who benefits: Technically comfortable Android TV owners may benefit if they can follow the guide and safely reverse unwanted changes. Developers of coding agents also get another real-world example of their tools being used beyond source-code generation. Who's exposed: Users who attempt similar changes without understanding the warnings could disable apps they still need, even if the guide says the changes are reversible. TV makers with sluggish Android TV implementations may face more public pressure from users who see debloating as necessary maintenance.