Usually, when someone talks about "starting from scratch" they mean "starting over from scratch," meaning throwing out something good enough in the hope of building something perfect.
This article is more about "first mover disadvantage" or "handicap of a head start" where an early implementation is leapfrogged by competitors that don't have legacy baggage.
It's 100% the endorsement I think it is, someone searching for "is there a fire near me" wants an answer. The main issue with websites (and why Twitter grew) is it became easier to colocate information and get real(ish) time updates. For sites that don't care about actual traffic being able to surface that data to AI bots is perfect. If it reinforces a loop of "oh [site] is the actual source of information" even better.
Why not? I'd much prefer if I can just send my agent to gather info among these sites than have to sign up for some Twitter clone, enable Javascript, then click around through a sea of algorithm bait.
This article is more about "first mover disadvantage" or "handicap of a head start" where an early implementation is leapfrogged by competitors that don't have legacy baggage.