If that's the limiting factor, doesn't that lend itself to some sort of shared solution where multiple apps share one Tap server? Especially if there are shared lexicons, but perhaps even if there aren't
That is a great way of putting it.
Would be so useful for revealing which accounts are actually trying to be truthful vs those intentionally spreading disinformation.
And for people trying to stay informed via social media, allow you to quickly get updated after reading wrong info
Only tangentially related, but this reminds me that one feature I have always wanted for social media was the ability to issue corrections for previous factually incorrect posts so that the corrections could be shown to everyone who saw the original
and if all the engagement is wiped out, then editing is just delete and posting again, but with the time backdated to the original post time so it won't show up at the top of feeds and get the engagement again which in some ways is worse
Which then also creates the opposite problem where someone unknowingly / accidentally wipes out a bunch of engagement by fixing a one character typo that doesn't in any way change the meaning
Right now, model updates are months and weeks behind the news, but perhaps there will come a time when updates are quick enough that days or hours matter and there is a tech business model that can be built around having access to news first
At some point it feels like the big tech giants will need to just subsidize the production of high quality, original, truthful text about current events to feed their LLMs. Just hope they realize that ahead of time and won't just wait until everything breaks
I want to go even bigger and have sports with leagues with teams and then teams have games with scores (and maybe more stats?)
But you are making me question whether it's actually a separate thing or better done as a specific implementation of communities/topics/events
What you are describing sounds a lot like what I am working on for sports.
You contrast it to live events, but what is a live event discussion but a community of of users posting about a topic or collection of posts about the event? bsky.app/profile/dana...
And speaking of the EU, I think one of the more useful approaches to this is to think about how to create a fully legal LLM training data set would work that would comply with even the most strict laws. Can't stop bad actors, so better to try to work to enable the good ones
EU almost certainly the place that anything is most likely to have some legal teeth.
Decentralized nature of atproto also does lead to some really unique situations where copies of parts of user data exist in different spots. Not sure how an obligation to seek out user permissions works legally
Like the proposal says:
"These schemas allow users to declare how their public data may be used by external consumers"
and I'm not really sure that's true in any sort of legally binding sense. Seems like at minimum it depends on your PDS terms of service
To me, this is less about Bluesky devs and far more about them paying some lawyers to figure out what, if anything, this sort of thing does or could do in a legal sense. Or is everyone just going in assuming this is purely just a user preference thing with no legal implications?
But what he is showing with that graph is the growing number of accounts that keep their data on servers not owned by Bluesky.
Blacksky is one option, but so is @eurosky.social and more people are starting to hosting their own just for themselves.
Not all atmospheric apps have a separate PDS
Any app built using the AT protocol that Bluesky created for their app.
The person you responded to runs @blackskyweb.xyz which doesn’t use any Bluesky infrastructure, so actually technically you just did with this reply.
Other examples: @anisota.net @leaflet.pub @tangled.org @cartridge.dev
Graph of Bluesky daily likes since the beginning of the site. Source: https://bsky.jazco.dev/stats
Bluesky user data is a start of the comparison cherry picking choose your own adventure. Simultaneously exploding, dying and plateauing depending on which narrative you would like to support bsky.jazco.dev/stats
Working on something myself where I think I am going to try to split the difference and launch an "unbranded" generic lexicon project at the same time.
But it creates extra work and a bunch of extra decisions about what portions are generic vs app specific without the benefit of a 2nd app existing
Had a big discussion that starting from branding and ended up on a bunch of similar topis with @bmann.ca and @aka.dad about 2 weeks ago starting here: bsky.app/profile/aka....
I am very much with you, but the issue is that waiting for collaboration and requiring in essence a second project to be completed and launched in order to start working on the first really slows things down.
Unfortunately, I don't know that there is a good solution
and make it elected during the midterms so we can always make sure that the AG is of the opposite party as the president for at least the 2nd half of the term
It's no wonder a plagiarism machine works better when your project will never be less than 99.99% stuff other people wrote compared to something like a journalist doing original reporting about events that just happened
One underrated aspect of why LLMs are so useful for coding is that at this point you have to go install and run so many lines of someone else's code just to get set up to start writing your own. When you consider OS, language, frameworks, etc the original code probably rarely reaches 1% of 1%
Perhaps crypto makes less sense as a comparison and more as a contrast of something that took off not because it was a good idea, but because it was attractive to investors. Could be possible that atproto is a great idea, but will end up held back by being unattractive to investors
True, but at risk of sounding like all the bitcoin people for so many years, it is actually still early. And that's why I thought it was worthwhile to try to make the positive case for why atproto stuff is more investable than it seems
Wrote up some thoughts inspired by this post and now shameless piggy backing off of it to share pmcewen.leaflet.pub/3mihbkh7fj22c
Especially about 2 sentence thoughts you see someone posting where in a best case scenario you already know you are missing some context, tone and body language
Working on a project and to the point where I have more strategy questions than anything and would really appreciate some advice and guidance. Can I take you up on this offer?
Finally done. I was able to talk with @pfrazee.com, as well as a handful of other folks with insight on the situation.
#atmosphereconf
trezy.com/blog/the-mar...
