PARTLY

Partly true: trajectory prediction is real car-safety tech, but this 10-second reel is an animated mockup, not a tested system

Facebook · Oct 2026

The claim'AI predicts what happens next. AI predicts this in 5 seconds.' Dashcam footage of a junction with boxes on a motorcycle, pedestrians and a turning car, confidence climbing to about 89 percent, 'Motorcycle crosses in front of vehicle, impact in 2.4s', then 'PREDICTION CORRECT'.

The idea is real. Projecting where tracked road users will be in the next few seconds, from their speed and heading, is a core piece of driver-assistance and self-driving systems, and research models do it far better than straight-line guessing. This reel does not show that working. It is a 10-second clip with the same template as the page's other reels: a big headline, an overlay panel, and a code editor scrolling underneath. The 'prediction' and the green 'PREDICTION CORRECT' badge are laid over footage whose ending the creator already knew, so the video cannot show a hit rate. There is one clip, no misses, no numbers across many scenes, and no way to tell whether the overlay came from a model or from a video editor. The caption even praises 'showing its own hit rate, including the misses', and then shows exactly one hit. The page belongs to a small software company that posts a new AI demo of this kind several times a week, across unrelated industries (dating, construction, retail, traffic). That pattern reads as marketing for custom-build services, not as a product you can buy or test. If you want collision warning in a vehicle today, it comes from the automaker's forward collision warning or a dedicated dashcam with ADAS alerts, both of which publish what they detect.

What holds up

Reel read frame by frame: headline 'AI Predicts What Happens Next', 'AI PREDICTS THIS IN 5 SECONDS', tracked boxes labeled MOT-01 and PED-01/02, prediction card 'Motorcycle crosses in front of vehicle, 76 percent, impact in 2.4s', then 'PREDICTION CORRECT'.
Clip length 10.08 seconds, identical to the page's other demo reels (dating, construction, retail), each with the same headline-overlay-code layout.
Only one scenario shown; no accuracy figure, test set, false-alarm rate or product link.
Trajectory prediction is an established component of driver-assistance and autonomous-driving stacks; the technique itself is not in question.

What doesn’t

Outcome badge added to footage with a known ending.
Caption claims the system shows its misses; the reel shows one hit.
Templated 10-second demo series across unrelated industries.

The catch

Real technique, staged demo. One correct call on a clip you already watched proves nothing about how often it is right.

How to actually do it

For a car, check whether your model has forward collision warning and automatic emergency braking, and look up its IIHS rating.
For a fleet, ask dashcam vendors for their published false-alarm and detection rates before buying ADAS alerts.
Treat any single 'prediction correct' clip as an ad; ask for results across hundreds of scenes.

Predicting road users' paths a few seconds out is real and already in many cars. This reel is a polished animation of the idea with one hand-picked success, not evidence of a working product.

Confidence: Medium
Posted by: a computer-vision software page (about 91K followers) that posts 10-second AI demo reels; reel sent to Buddy Sun Oct 11, 2026, ~7:47 AM

See the original claim →

We test hype for free. We build the real thing for a living.

Thirty minutes, no pitch — and you'll leave with something useful either way.

Book a call with Todd or start with the free Business Checkup →

The Verdict Weekly

Three verdicts every Friday. Free forever, unsubscribe anytime, no spam — that would be ironic.