I’m describing the vast majority of fights that happen in the public.
But the comment you replied to already addressed those fights, and bike thefts, and the vast majority of cases that you’re talking about, by saying
If there is a long-lasting visual cue that the event has or has not happened yet (e.g. a window is either broken or not), then a binary search is very useful.
No one is moving goalposts. The parent comment said that binary search is useful in situations like bike thefts where visual cues are present, and not useful in situations where visual cues are not present.
In your hypothetical situation involving AI, the AI would use visual cues that are present, and so the situation is covered by the parent comment’s second paragraph. In a situation where there are no visual cues for the AI to use, it would be covered by the third paragraph. They still aren’t wrong about anything.