No AI, no computer: how a 3 MB phone app rode the "AI makes TVs dumb" wave to 1,500 stars

In late October 2026, a developer named Manish built a 3 MB phone app: with the phone and TV on the same Wi-Fi, you can strip the preinstalled junk from Android TV, install APKs, transfer files and type on the TV from your phone. It didn't invent a need; it rides the hottest open-source topic of that month, "using AI to make smart TVs dumb". Trend projects ask you to install an AI coding assistant, connect a computer and run command lines; this one removes all of that. In 15 days it got 1,539 stars, and the APK was downloaded about 30k times. Below: how it caught this wave, where the stars came from, and whether you can still get in.

A "don't do this" article, and a repo built the same night

On Oct 22, PCWorld published "Please don't use AI hacks to debloat your TV," by Jared Newman. His point: these methods require enabling developer debugging on your TV, connecting to a computer, and running command-line scripts, "running a pile of unreviewed scripts can have unintended consequences." He specifically noted that the wave's two projects, cleanyourtv and tv-detox, admitted some TVs got stuck in boot loops.

The same evening (UTC 19:51), man15h/tellyscope was created. The next day, Manish posted on XDA: "I built a free, open-source phone app that debloats your Google TV over Wi-Fi. No computer, no root." The post included an "AI disclosure": most of the code was written by Claude Fable and Claude Opus 5.5 under his direction and review. A moderator quickly closed it as a duplicate post.

By Nov 6, the README's first screen had five badges: 3 MB, no tracking, no root, 6 languages, GPL-3.0. Below was a section called "Safety rules" with seven items, almost each one answering a PCWorld concern:

• Built-in TV apps are only disabled, never uninstalled, and every step can be undone;
• Before the first change, it records the TV's original state, and "Restore all" only restores what it turned off;
• Apps needed for boot, network, and showing the home screen are on a protected list, with no switch to bypass them;
• Every change first lists the apps it will touch and asks you before doing anything;
• No root, no exploits, only what ADB on a stock TV allows.

How big the wave was

The wave project was cleanyourtv, from Gathin23: the repo was created Oct 17, and its content was a prompt plus scripts for AI coding assistants like Claude Code and Codex, letting the AI connect to your TV and strip ads and tracking for you. Over two-plus weeks it reached 52.6k stars (GitHub, 11-06). On Oct 19, TechCrunch ran "People are using AI to make their smart TVs dumber again," a Hacker News post on "Android TV without ads" hit 444 points, and The Verge, the Guardian, and CBC followed later. Before that, in September, HN threads like "LG TV records screen when off" and "hidden residential proxies in TV apps" had been trending for a month.

The demand is real, but the wave has a high bar: you need a paid AI coding assistant, a computer, and the know-how to use ADB. cleanyourtv sees this too. Its README's "Start here" says that if you don't have a coding agent, check out the community apps. That page (cleanyour.tv/app) 11-06 lists four apps, three of them for phones, and Tellyscope is one of them. The listing criteria are at the bottom of the page: open source, regular releases, responds to issues, no tracking.

Where the stars came from: one big wave, and a string of small amplifiers

The author, man15h, registered his GitHub account 2015 years ago, has 74 followers, and his previous best repo had 101 stars. He had no personal audience, and no trace of paid promotion.

The daily star count shows no single explosive day. Several small amplifiers passed it along:

(Daily GitHub stars; the timing of the posts matches the star jumps, but the causation is inferred)

The background was always there: cleanyourtv still gets several hundred to a thousand-plus new visitors a day, some of whom come through the "community apps" page. How much of the total that path accounts for can't be measured from outside.

Another number shows real people installed it behind the stars: over 15 days, 82 releases (including betas), with about 29.8k total APK downloads across GitHub. The stable v0.14.0 alone got 7,268 downloads (GitHub Releases, 11-06). Roughly every 19 downloads brought 1 star.

Among a dozen "phone versions" in the same wave, why it ranked first

Tellyscope wasn't the first to think of a phone version:

What it added came down to three things.

First, it turned a "cleaning tool" into "your phone manages the TV." It sideloads APKs (no USB drive needed), transfers files between phone and TV, lets you type with your phone, monitors CPU, memory, and network, includes a shell, and does one-tap factory reset. The opening of GitHubDaily 11-02's post wasn't about ads: "Want to install an app on your TV? The APK is on your phone, but you have to go through a USB drive. Using the remote to type a search term takes forever." Ad-removal stories cool off; pain points like installing apps and typing don't, and people share more when there are more angles to write about.

Second, it has Simplified Chinese. 6 languages (English, Spanish, Chinese, Japanese, Russian, Hindi), with the README stating the translations are machine-made. Chinese repost accounts can write about it right away, and among its peers it got the best Chinese exposure.

Third, the safety rules sit on the first screen, and releases ship fast. 15 days, 82 releases, with auto-update support via Obtainium and Orion Store.

But don't overrate the differentiation: timing is the first factor. Without the cleanyourtv wave, this wouldn't exist. It and cleanyourtv-app are the top two in the category, and the gap is small. Under the GitHubDaily post, someone pushed back: "Tools like this run on ADB over Wi-Fi……5555 The port is plaintext with no auth on the LAN, so if someone gets onto your home Wi-Fi, your TV's install permissions are wide open." Another asked: "If the TV restarts and the port changes, do you have to re-pair?"

How many people know about it

Plenty of people know the topic; TechCrunch and The Verge both covered it. Tellyscope specifically spread only in small circles: small sites like LockerGnome and digitbin, a few Chinese personal blogs, GitHubDaily, and a TikTok video from the GitHub repost account github.signals (23k views). It is free, open source, and has no in-app purchases.

The niche is already crowded: cleanyourtv's SHOWCASE lists 218 derivative projects, and there are at least five or six phone-side apps. The wave itself is fading too: cleanyourtv's daily new stars dropped from 4,000+ to about 600, while Tellyscope still gets 90–110 per day, slowly trending down.

If it were you

Verdict: C. This specific pit is full; building another "phone version of cleanyourtv" will almost certainly not win stars. But the play it demonstrates is worth learning, and it can be moved to the next wave.

Who it suits: fast indie devs who use AI to write code, can turn a prototype into an installable app in a day or two, and watch GitHub Trending and HN for waves.

Who it doesn't suit: anyone wanting to open another repo on TV debloating, or anyone who can't ship in a week, since windows for this kind of opportunity are measured in days.

How to do it, ranked by how worthwhile:

1. Wait for the next "high-barrier hit" and build its "no-X" version: once a wave project takes off but requires an AI assistant, a computer, or the command line, build a front end ordinary people can use immediately. Topics like car head units, projectors, TV boxes, and routers, the "AI-modding your home hardware" category, are candidates for the next wave.
2. Put the media criticism straight into the README's first screen: if people say it "will brick your device," lead with "disable only, every step undoable, protected list."
3. Turn a single-point feature into a manager: don't just do the one thing the wave is about. Also handle the user's other annoyances on the same device (installing apps, transferring files, typing), so sharers have other angles to write about.
4. Do Chinese from day one: Chinese GitHub recommendation accounts are a cheap, steady traffic source.

How to get users: first get listed on the wave project's official recommendation list, the cheapest traffic there is. Meet its criteria (open source, regular releases, responsive to issues, no tracking), then submit a PR to its SHOWCASE. Next, post once on the relevant Reddit subs and XDA (don't repost, or it gets closed), then wait for Chinese and English GitHub repost accounts to write about it. Record a 30-second demo video; repost accounts spread things much faster with one.

When "before" means: ship within a week of the wave repo's creation. Tellyscope came 5 days after the wave, which already made it late for its niche; two weeks later, it would almost certainly rank behind 218 derivative projects.

What would show you misjudged it: (1) if in late November Tellyscope still gets hundreds of stars a day, or it gains traction on Google Play, that means "phone-managed TV" has lasting demand beyond chasing a hot topic, and then doing what it does poorly (safer pairing, Fire TV support) is worth it; (2) if the next "AI-modding hardware" wave arrives and no phone front end passes 1,000 stars, then this time the cleanyourtv recommendation page carried it, and the playbook won't transfer.

Sources

GitHub: the man15h/tellyscope repo page, README, Releases download counts, daily stars (10-25 to 11-06); star counts for similar repos (11-06). The cleanyour.tv/app community app page (11-06). PCWorld, "Please don't use AI hacks to debloat your TV" (10-22). TechCrunch (10-19). The author's XDA post (10-23). X: GitHubDaily (11-02), 魏咕咕响 (10-28), Ryan Hart (11-03). LockerGnome (10-29) and Chinese blogs. TikTok search for "tellyscope". Wherever it says "inferred," it is an inference.