Skyline view of Vancouver from a beach with very fluffy clouds illuminated above, mountains in background
A sunset over an arcing beach, with a ship and a mountain in the background
A bay with several large barges, mountains in the background
A gaggle of Goosetophers at dusk, with the Vancouver skyline in the background
Some pre #AtmosphereConf photos from around Vancouver
Hope you feel better soon!
You as well!
Same here, was knocked out all of yesterday
I’ve been using Habitat Docs all day to take notes on all the great #ATscience talks. Hope everyone at the conf gets a chance to check it out! 📝 🔬
We’re back in business 😎
@atprotocol.dev looks like the atmosphere conf website is down
Should be all set!
I am really looking forward to Arushi's @offline.arushibandi.com talk.
As we were wrapping #AtmosphereConf 2025, she said thank you & a few other kind words on her way out the door.
So I'm very pleased to see this talk, "in conversation with several of the talks from last year"
Are you able to see how old those are? Probably from a very early iteration of this: habitat.leaflet.pub/3mhglyteqe22i
The other reason is that at some point we’d like to build a management pane to see the global view of all permissions you have granted. This isn’t easy to do if the UCAN is being delegated / sub-delegated
The fun part is when the problems you actually hit turn out not to be the problems you thought you were going to hit
There is a bunch of speculative protocol creation in this post, so would love to hear what everybody thinks! Are the benefits worth the drawbacks? Is there another way we can make AppViews work for small world cases? Should something other than the PDS handle the new protocol elements we described?
In part 2 of our calendar series, I explore what it would actually take to build a permissioned data app without AppViews. There are a couple twists and turns along the way, but in the end, we got a fully functional app 📆
spaces, meet cliques!
how we built a minimal calendar on pear (our permission-enforcing ATProto repo) and what we had to build to get there.
For the same reasons, the likelihood AppViews take advantage of grey areas is a lot higher. They might be using your data for training models, or to help target ads. If an AppView does these, do you sign a terms of service agreement with them?
OTOH, a client handling data properly will not keep your data around for that long (tho it may be cached, especially if local first). But, a client is much less likely to be hacked because 1) its unlikely the client is on a device exposed to the public internet and 2) its a much less juicy target
If a AppView handles data properly, it is still creating a long-lived replica of it that is out in the wild. If your friend's choice of AppView gets hacked, your data will be taken, even though you didn't make the choice to use the AppView (though you did auth it).
In the case of malicious actors, granting access for someone to read from your PDS is not from different authorizing an AppView. The bad actor can exfiltrate your data either way.
But a big difference emerges when we consider less malevolent cases...
By comparison, the number of 3rd party PDS providers out there in the world will likely be much smaller than the number of AppViews. So, it would be a lot easier for the community to work out which ones are malicious and get people to stop using them.
Its partly a matter of blast radius. In the future, there could be 100s of categories of apps, each with dozens of potential providers. That is a really large number of 3rd parties to keep track of. Any system for detecting malicious actors would be strained.
There's a couple different reasons, will try to go over them quickly 🧵
Not quite ready yet, but we got something cooking on that front too
If I wanted to send you an invite, but you used a different calendar app than me with it’s own AppView, I’d need to authorize your calendar’s AppView to see my event, so it can index and serve it to your client. Current ATproto apps don’t have this problem because all PDS data is public right now
Yup, exactly. A 3rd party client might also be malicious of course, but since the software runs on your machine, what it does with your data will be a lot more transparent
Check out my first blog! As the ATmosphere moves towards permissioned data, how do we make sure privacy is really preserved? Can we build a permissioned data app today? All that and more in the post 👀📅
over the next couple weeks, we're going to be sharing more about what we're building at @habitat.network :)
here's the first ... excited to see things break for the first time I guess!
habitat.leaflet.pub/3mggkuhjzjs2p
@habitat.network, we are so excited about ATProtocol not so we can get new social networks, but because of the primitives it provides for a world with user data agency. let's build that world.
beyond excited to be giving a talk at #ATmopshereConf and officially unveiling @habitat.network!
i'll be talking about how we're trying to get away from skeumorphism (callback to @rude1.blacksky.team's talk from last year) and building for privacy-first usecases on the protocol. can't wait 🌱
