I used AI to disable unused apps and shorten the animations on my TCL 98C7K. I tested four tune-up configurations and tried Projectivy separately, with 144 app launches in total. I wanted to see how much faster these changes could make a modern television.
Watch the Full Investigation #
I Let AI “Hack” My TV. Then I Ran 144 Launch Tests.
AI Debloating on My TV #
AI agents that clean up a sluggish smart TV have been all over the tech news lately. In September, Tom’s Hardware covered Mert Cobanov’s experience using Claude Code on an older Android TV. His changes included disabling apps, installing FLauncher, and shortening animations. On 19 October, It’s FOSS reported a broader wave of similar experiments, including the community project now called CleanYourTV.
Those reports got me curious enough to try it myself. Shorter transitions can make a TV feel quicker even when an application takes just as long to start. A simpler launcher can also clear the screen of clutter while YouTube and Plex perform as before. So I measured each change on its own.
I used my 98-inch TCL 98C7K and tested the reversible tune-up already applied to it. I had disabled six unused applications, reduced three animation scales to 0.5x and restarted the TV. I used Codex over Android Debug Bridge (ADB) for the changes and measurements. The Tom’s Hardware story used Claude Code. The Android commands work with either assistant, and I could also type them myself at a terminal.
The main test covered 96 launches across four configurations. I then ran a separate, balanced 48-launch comparison of Google TV Home and Projectivy, with Projectivy running as an ordinary app and the tune-up unchanged. Those results are on page 5.
The six-app tune-up changed median launch times by only 0.6% for YouTube, 0.5% for Plex and 1.5% for VLC. Available memory barely changed, and my Home-screen trace showed no improvement in smoothness. Idle CPU load was lower during the sampled periods. I would keep my expectations modest.
Test TV Specifications #
Android identifies my TV as Smart TV Pro. The retail model is 98C7K, as the Product Information menu confirms. Before using a debloat list from another TCL, I would check each package against this model and its firmware. Sharing a manufacturer does not guarantee that the packages do the same job.
| Verified Item | Result |
| Retail model | TCL 98C7K; generic Android label: Smart TV Pro |
| Operating system | Android TV 14 / API 34 |
| SoC property | MediaTek MT9689 (device-reported) |
| TCL firmware | V8-0012T01-LF1V643.003175 |
| TV+ OS | TVM7.1.0_71644eaa1_260320 |
| Security patch | 5 December 2025 |
| Kernel | 5.15.180-android14-11 |
| Android build | tcl9618-user 14 UTT2.250416.001 AU02 release-keys |
| Product-menu capacities | 3.0 GB RAM; 64.0 GB storage |
| Android-visible memory | 2393.8 MiB total in the initial audit |
| Physical / UI size | 3840×2160 physical; 1920×1080 UI override; density 320 |
| Initial free app storage | 45.23 GiB; 92.69% of the app-data partition |
| Package inventory | 168 packages, including Android components; six optional packages disabled |
The feature menu also lists 144 Hz, DLG 288 Hz, 120 Hz MEMC, a 48–288 Hz VRR range, Dolby Atmos, DTS, Dolby Vision IQ, Dolby Vision Game and IMAX Enhanced. I did not verify these capabilities in this software test. The highest refresh-rate mode depends on resolution: TCL’s 98C7K specification sheet qualifies 288 Hz gaming at reduced resolution. I did not test panel refresh, HDR accuracy, audio performance or HDMI input lag.
My initial audit found approximately 45.23 GiB free on the app-data partition, so storage was hardly a problem. The six optional apps were already stopped and absent from the resident-process snapshot. I would not expect much RAM back from disabling them. The TV had been up for about 51 days, including standby, so I restarted it before the controlled comparison. Otherwise, a fresh boot could have received credit for the tune-up.
I tested a software change on the 98C7K, with the original screen, firmware, launcher and display settings retained. A complete picture-quality review would require separate testing.