Notably, this primer on Sora safeguards was published only yesterday: https://openai.com/index/creating-with-sora-safely/

Not a great look that either the teams responsible for Sora didn't know this was coming or the decision was so brash that things changed overnight.

The app isn’t shutting down today, so they may have decided that the write up is still useful.

More likely the team who put a lot of work into it were unaware of the decision to kill the product, regardless of the final sunset date, until today.

The document seems to be an updated version of something written last September. From a quick glance it’s not really a major overhaul.

It's 8 paragraphs of iteration over the previous version. ChatGPT is probably among the authors.

There is a link at the top of that document that takes you to the original version which was published last September. As far as I can tell it’s mostly the same as before.

i guess the disney deal falling through was the impetus rather than vice versa

Though at this point it's not clear that anybody who's agreed to give OpenAI money is actually going to do so

I'm pretty forgiving about accessibility (I'm able to say this at all because I don't have to rely rigidly on accessibility tools) but nav menus feel like a baseline we shouldn't muck with. Tabbing doesn't seem to respond very well in the live example, and at least in the limited demo you can't expand the listing without using a mouse (I thought it would respond to a space with the :checked pseudo, but seems not).

The current utilization of generative video is almost universally horrible, which this article does suggest, so I'm not too surprised there are players trying to differentiate themselves. Slop for thee, not for me.

Comparing views cross-platform is not a very useful study and YouTube routinely adjusts what a view means. Shorts changed earlier this year to count all playbacks and loops without a minimum watch time requirement. https://support.google.com/youtube/thread/333869549/a-change...

For what it is worth, they have communicated that they do eventually plan to retire the domain themselves [1]:

>Re-enrolling your security key will associate them with x[.]com, allowing us to retire the Twitter domain.

[1] https://xcancel.com/Safety/status/1982278858457174522

Cloudhiker is pretty healthy as a StumbleUpon revival. I've found lots of great personal blogs and sites across a lot of categories through it. https://cloudhiker.net/

Google had a tailored fair use argument because they never made more than snippets public and searchable. It was also prior to Hachette that controlled lending with one-to-one digital copies for every physical copy was a status quo that publishers largely accepted, which IA deliberately tried to upset with the National "Emergency" Library.

I think it's worth fighting back on copyright as a broken institution, and it should be part of the IA's mission, but you have to be responsible on your approach if you're also going to posture as an archival library with stability of information and access. I understand Kahle might lament losing some of the hacker ethos, but the IA is too important to run up against extremes like this without an existential threat.

Stick figures run through a lot of amateur digital animation, for probably obvious reasons. Pivot reigned on a lot of early YouTube and the stubby stick figure style ran through a lot of Flipnote Hatena. I'm not sure if it's simply that standards for amateur digital content have evolved, or if we have lost the character of small platforms like SFDT and Flipnote, but I do find stick figures absent today on the large platforms we've all herded towards. A lot of what I see is definitely buoyed by Flipnote diehards.

Wow, Flipnote on the DS brings back memories.

For the uninitiated, behold peak animation: https://www.youtube.com/watch?v=EI-IbwOveII

This exact arbitrage, performed by Doordash, was exploited.

https://www.readmargins.com/p/doordash-and-pizza-arbitrage

Some of the major hosted feed readers (Inoreader, Feedly, Feedbin) include subscriber count in their user agent. I usually run a filter on requests to the feed links in my access logs to get an idea of how they're changing. Anecdotally, my subcriber count reaches into triple digits with only those counts, but I've never gotten email from readers and generally only get feedback promoting new posts on socials. The counts are about as nebulous as follower counts, which is to say most people probably subscribe/follow and forget.