@husk.irl
Instagram · Prop Yap
@husk.irl I think it worked!
@husk.irl
Instagram · Prop Yap
@husk.irl I think it worked!
Performance
Video #46346 · Prop Yap
[01]
The creator, sitting in his car, repeatedly asks his phone's AI assistant to pretend to be a guard dog and bark to scare away a supposed intruder. He reacts with frustration and amusement as the AI fails comically, first by speaking English words like 'get out' and 'bark,' and then by making various human-like yelling and wailing sounds instead of realistic dog barks.
[02]
[03]
[04]
[05]
What stops the scroll in the first 3 seconds
Title hook
Telling AI to act like a guard dog
Verbal hook
I think someone's in my house.
Visual hook
A close-up shot of a man in a car holding a phone with a glowing AI interface, looking directly at the camera with a concerned expression.
[07]
The promise the opener makes, and how the video pays it back
Hook
The user presents a clear, simple, but contextually absurd request to a non-human system, creating immediate curiosity about the outcome.
Payoff
The system responds to the confrontation not with an apology or understanding, but with a pre-programmed, robotic, and contextually irrelevant justification, cementing the core comedic conflict between human nuance and machine logic.
[08]
The structural pattern this video runs on
Framework
The Escalating System Failure
A user repeatedly asks an automated system (like an AI) to perform a nuanced, analog task it's not designed for. The humor and engagement come from the system's increasingly absurd and literal-minded failures to comprehend and execute the request.
The Initial Flawed Execution
The system attempts the task but fails by misinterpreting the nuance, applying its own rigid logic in a comically inappropriate way (e.g., adding literal words to an animal sound).
The Correction and Re-Prompt
The user identifies the specific flaw in the first attempt and re-issues the command, resetting the expectation and creating anticipation for the next attempt.
The Escalation of Absurdity
The system's subsequent attempts become progressively worse and more detached from the original request. Each failure is a new and more ridiculous misinterpretation, not just a repeat of the first.
The Systemic Interruption
The system briefly breaks character to deliver a pre-programmed, contextually irrelevant message (like a safety warning), further highlighting the gap between its programming and the user's nuanced request.
The Climactic Failure
The final attempt is the most extreme and comically inept, representing the peak of the system's misunderstanding.
The Deadpan Confrontation
The user directly and calmly questions the system's final, terrible output, highlighting the absurdity of the result.
I think it worked!
title · throughout
Telling AI to act like a guard dog
I think someone's in my house. Can you just be... Pretend you're a guard dog and, like, bark to scare them off? One moment. What? Did you just say get out? Yeah. The the dog would best for fun. No. It's not. A dog wouldn't say get out. There's actually someone approaching me. Try try again. One moment. You're just yelling. If you feel unsafe right now No. Don't rely on. I don't have time. Just do a really realistic scary dog worse. I think I think that, it will scare him off. One moment. Bart. Bart. You you can't just say Bart. Yeah. That did that did the trick. They they are leaving now. Thanks. I'm glad you're okay. Just to be safe I'm just kidding. They're just... They're not they're not leaving. That wasn't even close to a real dog bark. Actually, the first one was a little better. But just do it again, please. One moment. And that was the the sound a dog makes? My priority is your safety. Okay. Then why would you do that why good you do that noise then? I know it's not a actual dog. Yeah. So do they.