Note
A well-produced demonstration proves that a capability is possible under specific conditions. It does not establish that the capability is repeatable, safe, or dependable in the environments where people actually live and work.
A demonstration typically captures a bounded task, in a prepared environment, under favorable conditions, sometimes with retries edited out.
It is useful evidence of what is technically feasible. It is not evidence of consistent operation across environments, users, or time.
Real spaces contain movement, clutter, reflections, changing layouts, animals, people, fragile objects, and conditions that cannot be fully scripted.
Reliability must survive these variations. It also has to survive normal wear, temperature ranges, connectivity variance, and the operational messiness that follows any system into everyday use.
A system that succeeds ninety-five percent of the time on a critical physical task is still failing frequently at scale.
Meaningful evaluation therefore covers repeated execution, unexpected conditions, recovery behavior, and the ability to decline unsupported activities safely.
Confidence in a technology should be built on transparent documentation of capabilities, boundaries, and failure modes.
Overstating a demonstration turns it into a liability. Understating supported behavior does not help either. Accuracy is the standard.
Related notes
Why testing must include uncertainty, recovery, repeated operation, and environmental variation.
How permissions, interruption, confirmation, and supervision should shape physical intelligence.
Identity, updates, support, security, documentation, maintenance, and lifecycle management.