That dovetails with the standard prescriptive vs. descriptive issue, don't you think? The speaker in your podcast example seems to be (mis- ?) using the word to refer to a stereotype, likening an Internet trope to reality, and (thus) implying that the Internet now has the status of a canon. So, it's an inaccurate use, prescriptively; but it's sufficiently related to the prescriptive use that it gives us some insight into the speaker — probable age, reading habits, opinion about the validity of what the speaker reads online, etc. — provided we have some context. It's a pleasing instance of the game of language being played.

Later edit: I guess one point I'm making is that real dictionaries are still of great use and need in the age of LLMs.

I agree that traditional dictionaries are still very useful. I wouldn’t trust today’s LLMs, for example, for information about etymologies, pronunciations, alternate spellings, conjugations and declensions, etc., unless they had access to human-curated lexicographic databases. But for grasping the meanings of modern words in real-life contexts, LLMs are much better than static dictionaries.

On the prescriptive vs. descriptive issue: As a (former) lexicographer, I think I can say with confidence that it is very difficult to maintain a consistent prescriptive stance when trying to create a general dictionary of a language. You have to have some basis for declaring that a particular usage is wrong. In a few cases, such as hopefully used as a sentential adverb or data as a singular noun, you can find prescriptive grammarians who condemn it or a systematic reason (logic, etymology, etc.) for excluding it. But the vast majority of words in a language acquire and change their meanings through people using them in various ways and situations, without being noticed by prescriptivists and without following clear patterns.

Canonical seems to be such an example. I think I’ve seen that podcast usage before, but I can’t say how old or well established it is. The entry for canonical at the online Oxford English Dictionary does not include it, though that entry has not been updated recently. And the most recent edition of the huge, prescriptivist-friendly Garner’s Modern English Usage says nothing about it. If further investigation revealed that the word has in fact been used in that meaning fairly widely for more than a decade or two, I think the meaning should be included in dictionaries without any marker of incorrectness or inaccuracy.

I don’t know how many dictionaries other than the OED are being regularly updated, though. The market for conventional dictionaries seems to have collapsed.