good shout, I think some recent typography changes made the bolding more subtle on mobile. I've pushed an update underlining them and it should be live momentarily!
Blog post author here! Always happy to see people looking at and reading about these paintings. Happy to answer any questions or job offers people here may have!
I'm speculating a bit here, but I suspect that answer reflects a practically risk-averse read of the same case, for two reasons. One is that, as a district court opinion, Bridgeman isn't _necessarily_ binding federal precedent. (I don't want to overstate that point, because Bridgeman is well established and has been found persuasive by other courts. Still, given that this site appears to distribute the resulting ebooks itself, I assume their appetite for risk here is pretty small.)
Even following Bridgeman though, there is a teeny bit of analysis that you have to do to determine that the reproduction is a faithful photographic copy of the original. In most cases that's straightforward, but again, if you really want to limit risk and as in this case your application doesn't require a particular image, it might be worth cutting out that analysis altogether and going with media that is simply outside of the range of copyright terms.
Really lucked out here that .gov is explicitly enumerated in a published list. For basically any other TLD your first (big!) issue will be finding such a list.
Thank you for sharing! If folks are interested in the code that powers the bot (it uses OpenCV to find the label, then draws frames and animates with ffmpeg) it's available here: https://github.com/thisisparker/78_sampler
Oh man! I didn't know there was prior art. I think yours looks great, Rob, and I'm very eager to get it installed and compare notes. I'm impressed that you got all the clues displayed at once.
Raph! I loved that puzzle (and I've solved it in cursewords)! I may try to get together a collection of freely licensed demo puzzles, and I'd be honored to include this.
There isn't such a way yet, but I want to do that in the future. (This 1.0 version doesn't have any kind of persistent settings, so there's not yet a place to store custom key-mappings.)
In case you want to try out my defaults: `tab` and `shift+tab` advance/retreat words to the next blank, and `page up` and `page down` (on a Mac, `fn` + `up` or `down`) do blank-agnostic word advance/retreat.
(I also haven't documented this yet, but `[`/`]` and `{`/`}` do cursor-perpendicular box-wise movement in blank-agnostic and blank-aware steps, respectively. One of the hardest parts of this is just describing these kinds of movement.)
Huge fan of Phil! Keiran and I are both alumni of the Recurse Center (https://www.recurse.com/) from around the same time. It's become a real crossword programmer powerhouse.
ArchiveTeam and URLTeam do great work, but those projects and the OP have different goals. (I'm the OP.) This bot will NOT produce any kind of corpus of tweeted links or links shortened by Twitter's t.co shortener; in fact, it bypasses t.co entirely and I don't have any kind of record of those shortened links.
Instead, it backs up the contents of the pages that are linked to at the time of the tweets. Frankly my tool doesn't do anything interesting at all with the URLs — it just submits the "expanded URL" that was tweeted to the Wayback Machine and lets it sort out any and all 301s.
If there are words in the original post that give the impression that we're providing an exhaustive summary, let me know and I'll see how we can clarify.
One of the arguments ESA made seems to resonate with people. I don't think it's at all dishonest to point that out.
We will address the comments more completely with the response filing. In the meantime, I'm glad your 9 minute list exists in case people prefer it to the original.