stunt work is going to be a dead job by 2027, and this clip is exactly why.
Sylvester Stallone rides a Harley straight off a rooftop and drops all the way down to a giant trampoline in the street, with a huge crowd packed around it.
none of it happened. a creator made the whole thing on a laptop for about $7. the kind of stunt that used to need a closed street, a crane, and a very nervous insurance guy now takes one evening.
it's already all over the feed, with half the comments swearing an old action star really pulled it off and the other half staring at it trying to spot the fake.
what sells it:
> the bike sits against the open city skyline so the height reads as real
> the trampoline below is scaled right against the crowd around it
> the motion blur on the bike mid-air gives it real speed instead of a frozen pose
the tools behind it:
> Kimi K3 wrote the launch, the drop, and the crowd timing
> Kling 3.0 built the bike, the rider, and the fall
> Seedance 2.5 handled the trampoline bounce and the light on the metal
> ElevenLabs made the engine and the crowd noise
> CapCut did the slow motion and the color
a real version of this needs a licensed rider, a rooftop permit, an airbag rig, medics on standby, and weeks of setup. everyone signs a waiver.
this one needed nobody to sign anything.
i spent a week putting the full money guide together. it's in the article below
THE NEWS OVERLAY SPELLS PURSUIT WRONG
151 likes. 8 comments
A humanoid running down an interstate, cruisers behind it, with a lower third that reads ROBOT PURSUITE ON INTERSTATE
that typo is the whole post, because it is the most reliable detection signal in the clip:
โ broadcast graphics go through a template, an operator and a producer. a misspelling in a lower third survives none of those
โ the same phrase appears twice in the frame, in two different boxes, with the same error in both. a render copying itself, not a control room
โ everything else is convincing. the LIVE badge, the traffic, the camera position, the motion blur
โ the model reproduced the look of the text layer without reproducing the process that produces text layers
โ and 151 likes means almost nobody stopped long enough to read it anyway
the operational point is that detection has moved to process artefacts rather than to pixels. faces got solved, physics got mostly solved, and what remains unsolved is everything produced by an institution with review steps in it
broadcast graphics, official document layouts, regulatory disclosures, standardised forms. these are outputs of workflows, and a model reproduces the appearance of the output while skipping every check the workflow contains
so the practical heuristic for anyone assessing synthetic media at volume: stop examining the subject and start examining the institutional furniture around it. spelling, font consistency across two overlays, whether a timestamp advances at the right rate, whether a station ident matches the market
that is also where automated detection should be aimed, and mostly is not. the detectors look at the image, and the tell is in the paperwork
if you want to test where your own eye breaks, making one yourself is the fastest calibration. image-to-video off a still, one line about the motion, @Picsart runs it from a phone
the robot is flawless. the chyron cannot spell
THE COMPUTER IS TINY.
THE COOLING SYSTEM IS MASSIVE.
At first, this looks like some kind of industrial machine.
Pipes.
Vacuum lines.
Insulation.
Hundreds of connections running everywhere.
But this is part of the infrastructure required to operate a quantum computer.
And this is what caught my attention.
The processor itself can be incredibly small.
But everything around it can become enormous.
Multiple temperature stages.
Specialized cooling systems.
Vacuum hardware.
Precision plumbing.
Electrical and microwave connections.
The processor is only one piece of the puzzle.
The real engineering challenge is creating an environment where it can actually operate.
We usually talk about the future of computing as if it's all about better chips.
But videos like this show another side of the story.
The chip may be tiny.
The machine required to make it work can be massive.
Sometimes, the most impressive part of a computer isn't the processor.
It's everything built around it.