Demonstration
Imitation preserves the human reference.
Useful for establishing a behavior, but the copied motion does not by itself encode whether a later attempt was faster, more reliable, or expensive when it failed.
An educational outcome model
A robot that only copies can only inherit the limits.
Explore the shift described in the post: from replaying a demonstration to weighing speed, reliability, and the cost of failure.
The learning boundary
Copy the move. Or learn from what happened next.
Speed
Completion time becomes a visible outcome, not an invisible side effect.
Reliability
Repeated success can be compared across trials.
Failure cost
A failed attempt can carry a penalty instead of disappearing from the lesson.
Outcome
Transparent scoring makes the tradeoff inspectable.
Source boundary
This lab illustrates the post's idea. It does not claim to reproduce Kinetiq Ascend or disclose how it works.
Two ways to read the same trial
A demonstration tells the robot what a person did. An outcome tells it what the attempt cost.
Interactive trial
Give the attempt consequences.
Use the controls to model one trial. The formula is deliberately simple and visible: an educational lens, not a robotics benchmark.
"A robot that only imitates can never move faster or more reliably than the person it copied, and it never learns the cost of failure."
Grounded in the supplied post naming @thehumanoid and Kinetiq Ascend. The post ends mid-sentence in the provided source, so this artifact makes no claim beyond the visible text.