Avernyx

[Avernyx @Avernyx]

@spacefundz woow

[Avernyx @Avernyx]

@Shileisaboy Check out my new post—it’s worth it😅

[Avernyx @Avernyx]

@lateascent nado))

[Avernyx @Avernyx]

@Drundor wooow

[Avernyx @Avernyx]

@Amos_FACT Check out my new post—it’s worth it.

[Avernyx @Avernyx]

If you want to try it, the open-source version with tested lists per model: https://t.co/UKrnEPZbEk

[Avernyx @Avernyx]

It is becoming dangerous how easy this is. A TV "hack" with Claude Code just passed 4.3M views and 6,700 bookmarks in 3 days, and almost nobody is asking what it actually did.

Here's what it did. Nothing was hacked.

Every Android TV ships with adb, Android's own debug bridge. You open developer options (press OK on the build number 7 times), switch on network debugging, and any computer on your wifi that you approve can send commands to the TV. Claude Code just types those commands for you.

The cleanup is one command, repeated: pm disable-user --user 0 [package]. Disabled, not deleted. The ad row, the vendor store, the "recommended" feed, the usage trackers. Then a lighter launcher becomes the home screen. The screensaver with revenue and Stripe pings is an extra app on top. It isn't the fix.

That's the whole trick.

It's also why it doesn't work everywhere. Samsung runs Tizen, LG runs webOS, most Vizio sets run SmartCast. No adb there, so no cleanup. Fire TV lets you disable apps but won't let a new launcher become home.

The risks are boring, but real:
→ disable the wrong package and you can lose HDMI switching or the remote
→ turn off the stock launcher before the new one works and you get a black screen
→ leave network debugging on and every computer you approved can still control the TV

None of this is hard. It used to need someone who knew which packages are safe to switch off. Now the model mostly knows, and an open-source repo already has tested lists for about 30 models.

The TV is the least interesting part. The bigger shift is that the "family IT person" job just got a lot smaller.

Which TV would you try this on first?

[Ryan Doyle @doooyle]

Hacked my TV with Claude Code today

No more ads, bloatware, spyware or laggy UI

I also got a beautiful screensaver with my revenue, calendar, weather, and notifications for new Stripe payments

It was super easy! I sent it the listing for my tv, asked "can you hack this?" And it talked me through how

[Avernyx @Avernyx]

@Amos_FACT I think you’ll want to see this))

[Avernyx @Avernyx]

This is very funny 😁

[Defiant L’s @DefiantLs]

Some people would say this is a better feature than flying 😂

[Avernyx @Avernyx]

@DefiantLs hahahhahahahahha

[Avernyx @Avernyx]

Hey @grok, which country has the best chance of leading the world in AI by 2040?

[Avernyx @Avernyx]

❓️❓️❓️Everyone is generating the same empty dashboards. And almost nobody is asking why.❓️❓️❓️

Open Twitter right now and you’ll see them everywhere: neon decision grids, spinning “discovery architectures,” token funnels, support-ticket matrices, trading floors full of little agents, Roadster teardowns with confidence scores and part trees. Glowing nodes, live metrics, code snippets in the corner, percentages that feel authoritative. They all look like they came from the same prompt family, and most of them did.

None of it does anything.

There is no real pipeline behind the “JEV / SIGNAL LOOM.” No working agent swarm in the trading room. No actual architecture inside the spinning sphere. The numbers are decorative. The graphs are vibes. The confidence scores are whatever the model decided looked good that day. You can usually tell because the labels contradict themselves, the metrics don’t change in any meaningful way, and the whole thing collapses the moment you ask what the system actually does when no one is watching.

People still post them.

They post them because the aesthetic of complexity is now cheaper than complexity itself. Claude (and every other model) is extremely good at producing the visual language of serious systems: dark backgrounds, cyan and magenta accents, HUD elements, “live” readouts, tiny monospace text that implies there’s a real backend somewhere. You feed it a concept, it returns something that photographs like insight. Then you post it and the replies do the rest — “this is insane,” “how did you build this,” “the future is already here.”

It’s a status shortcut. You don’t need to understand cold-gas thrusters, agent orchestration, or email triage logic. You don’t need to ship anything that works. You just need the model to draw something that looks like you understand it. The image becomes the proof. Engagement becomes the validation. And because the barrier is now basically zero, the feed fills up with confident-looking air.

This is why the style spreads so fast. It’s optimized for the platform: high visual density, implied technical depth, low cognitive cost to produce, high social reward to share. It creates the sensation of progress without requiring any. The more people copy it, the more normal it becomes to treat empty interfaces as evidence of something real. Eventually the distinction blurs — not just for the audience, but for the people making them.

The result is a timeline full of dashboards that imply systems that don’t exist. Visual theater dressed up as engineering. And the strange part is how rarely anyone points at one and says the obvious thing: this doesn’t do anything. It just looks like it does.

That’s the whole product.

[Avernyx @Avernyx]

@LillySpeaktv 1

[Avernyx @Avernyx]

@rost_roma It’s not that hard )

[Avernyx @Avernyx]

Why are there so few comments?))

[Bober_smart @Bober_smart]

https://t.co/p4teS18nnJ

[Avernyx @Avernyx]

Is it actually possible to win the lottery?

[XHabib @XHabib]

If I win the lottery, I won’t tell anybody, but there will be signs

[Avernyx @Avernyx]

@YapayZekaSerisi

[Superior @andreysuperior]

https://t.co/rUN1cPDY65

[Avernyx @Avernyx]

@Mubi_1713 @grok hahhah

[Avernyx @Avernyx]

@AirdropAlchemis