The "Drizzle Queries" section of the docs describes additional APIs for relations (referred to in the docs as db.query). There is also an API that looks/works much more like SQL (see db.select(), db.insert(), db.update()) with good types.
I think I read somewhere (can't find the source at the moment) that he mainly used his existing access (to slack?) from his phone and didn't actually do much hacking at that point.
It probably depends on the reference point. If I now use 4 words (2*2 rhyming words) instead of passphrases with 2 words, then the password is more secure than before (even if the attacker knows for sure that I use a rhyming password).
Besides, there is the rhyme scheme, which is not necessarily known to the attacker.
The idea is fun. I'm just not sure it helps much in practice. If I can't remember the exact words, the rhyme makes me only marginally less lost than without it.
Twitter could, of course, have just given a warning. But I also don't see any problem treating the tweet as if it were current. The age of the tweet does not change whether it is acceptable or not, does it?
I think that's part of the problem. Does a new video, which may have nothing to do with the originial video, really deserve the views and popularity of the old video?