// Contents
Developer Gathin T.’s parents just wanted to watch Netflix, but their Xiaomi TV made them sit through a sponsored clip first, every single time. Their son gave Claude Code the model number. Claude Code found the set on the home Wi-Fi, switched off 25 Xiaomi apps and wrote a new home screen. “Looks like an Apple TV now lol,” Gathin posted on October 17, 2026.
Within a couple of days his instructions had become an open-source project called CleanYourTV. By October 24, according to PPC Land , it had 30,900 stars on GitHub and app lists for 260 TV models, and the count keeps climbing. A GitHub star works like a bookmark: someone flagged the project to come back to it. Tens of thousands of bookmarks in a week means a lot of people are fed up with home-screen ads.
The more interesting part is what those ads are. In 2023 the American TV maker Vizio spent more on its TVs than it took in for them, and made its money on ads and data about what viewers watch. So the ads are neither a glitch nor one brand’s greed. They hold up the economics of the cheap smart TV, and an AI agent has, for the first time, handed ordinary owners an easy way out of that deal.
The TV that sells at a loss
Vizio, one of the big budget TV makers in the US, shows how it works in detail. Until Walmart bought it in 2024, the company was public and had to disclose its numbers. Its annual report for 2023 shows that TVs and soundbars brought in $1.08 billion in revenue and cost $1.09 billion to make. The arithmetic is blunt. Subtract one from the other and Vizio lost about $8.6 million selling hardware.
The money came from a different division, Platform+. It covers on-screen advertising and data about what viewers watch. That division brought in $598 million in revenue and, after its own costs, about $365 million in gross profit (our calculation from the same filing). By the same calculation, Vizio’s gross profit as a whole came to about $356 million. In other words, ads and data earned more than the entire company did, and the hardware ate part of that money.
// Vizio, 2023
The TV is a printer sold cheap to sell the ink, except the ink is bought by advertisers, and what they pay for is your attention.
In 2023 a startup called Telly took that logic to its limit. It promised to give away 500,000 TVs for free . Below the 55-inch screen sat a second, smaller one that ran ads around the clock. You could dim it. You could not switch it off. In return, owners handed over their name, age, gender, address and habits, including shopping habits. An ordinary smart TV differs from Telly in one respect: you still pay for it at the store.
A screen that watches back
The second revenue stream is even less visible. Automatic content recognition, or ACR, is Shazam for your screen. The TV regularly captures what it is showing, turns each frame into a short digital fingerprint and matches it against a database. That way the manufacturer learns what is playing in your living room and can sell the knowledge to advertisers.
Even when you are gaming on a console or playing a film from a laptop over HDMI, the TV keeps matching frames. That is how Samsung and LG sets behaved when researchers from the US, the UK and Spain took them apart in 2024. LG captured the screen up to a hundred times a second and Samsung twice a second, and both sent the fingerprints off in batches, as UC Davis and other outlets described it. Opting out of ACR in the settings did stop the data flow, provided the owner knew where to look.
In December 2025 the Texas attorney general sued Samsung, LG, Sony, Hisense and TCL for collecting viewing data without consent. In February 2026 Samsung reached a deal with the state , promising not to recognize Texans’ screens without their explicit permission. Everywhere else it still comes down to whether the owner found the right checkbox.
A screwdriver that was inside all along
Every Android TV and Google TV set has ADB, the Android Debug Bridge. It is a service channel for developers: a computer sends the TV commands through it, and you switch it on in a hidden menu. Forum enthusiasts have used it for years. But you need to know the names of the system apps, each brand has its own, there are dozens of them, and one mistake breaks the home screen.
The AI agent removes exactly that barrier. It finds the TV on the network, reads the list of installed apps and works out from their names which ones serve ads and which ones the TV needs. CleanYourTV’s instructions tell the agent to save the app list and write a restore script first. Only then does it switch off the manufacturer’s telemetry, ad services and recommendation rows. Nothing is uninstalled, so every change can be rolled back with one command. A similar project, dumbtv, according to TechCrunch , has been tested on Fire TV and sets from half a dozen brands, and it is even easier to use: you just give your AI agent the GitHub link.
The strangest story began with a failure. A Reddit user asked Claude to clean up their TV, and everything pointed to the usual route through the Android bridge. But the set turned out to be a Samsung running the company’s own Tizen system, which has no such bridge, and the agent couldn’t see the screen. So, as XDA recounted , it wrote a small local website, used it to get access to the user’s phone camera and photographed the TV every four seconds to check its own progress. After 11 hours and a hefty chunk of the subscription’s limits, it had wiped Samsung’s advertising ID, hidden the unused apps and built a new home screen.

Where the trick stops working
The method has plenty of limits. CleanYourTV only works on Android TV and Google TV. It does nothing about ads inside YouTube or Netflix, since those come from the apps themselves. And the first firmware update can switch the disabled apps back on and restore the stock home screen.
It can go worse. TechCrunch cites a user who bricked his Android TV, leaving it unable to boot. Cybersecurity expert Ritesh Kotak told CBC that many TVs have microphones, and a tampered system that stays online is more exposed. The updates people find so annoying are usually the ones patching security holes.
There is a sober view of the payoff, too. An XDA writer tried the method on his own TV : the agent asked permission before each change, the TV became more responsive, but the lag at power-on stayed. The real trouble with most smart TVs, he concluded, is weak hardware. No AI agent can fix that.
A refusal nobody explained
In late October a screenshot spread on Reddit: Claude Opus 5.5 declined to help with a TV, citing its safeguards. The task was reverse engineering, working out how a program is built from the inside. According to the poster, the model did it for a video game but refused for a TV. XDA stresses that this is unconfirmed: a filter aimed at entirely different risks could have caught TVs by accident. Anthropic has offered no public explanation. As early as October 19, TechCrunch noted that Claude would not respond to a short “hack my TV” request, and people had to describe the task in more detail.
For a company that sells AI agents, this is an awkward fork. Helping people fix their own stuff sells well. Helping them switch off someone else’s business model is riskier, especially if TV makers ever come knocking.
A TV without ads plays movies just the same; it simply stops earning money for the maker that sold it at a loss. So whose TV is it: the person who paid for it at the store, or the company that planned to keep making money from it for years?







