Maybe it’s like that trick where if a thousand people guess the amount of beans in a jar almost all of them will be wrong but their average will be very close to, if not, correct.
replythey probably weight documentation higher
replyThey're applying trust factors similar to PageRank.
replyPossibly, but at a semantic level.
replyI think many (most?) have preferred sources. I would weight Wikipedia and MDN higher than Snurk Grubble's blog in training, no matter what the topic.
replyAlso, you can count on most developers to reach the documentation only as a last resort, and try every random blog first.
replyThey look for latent structure in the data.
replyIt's probably trained more on fixes on incorrect CORS than the problems
reply