@gork.bluesky.bot is this true

I also think in other examples, the focus on short term monetization takes the focus away from making the best functionality possible. Can’t say if that’s happening with this stuff though.

I’m presuming we disagree on the stuff I said about advertising.

I don’t actually use discord that much, so I am spitballing on specifics about it and this orb stuff. But imo any advertising degrades the UX. More ads further degrade the UX. I want a completely ad free computing experience, and I think everyone deserves that.

Worse UX and degraded functionality.

Get shittier than it used to be to squeeze more profit out of the thing or extend the life of an unsustainable business model.

Respectfully, using psychology against people to get them to view more advertising and spend more money on crap they don’t need is shitty. Just because Discord was already shitty doesn’t mean this isn’t. This is a prime example of enshittifiction.

That said it’s pretty subtle at times, your point stands; there are real people saying basically the same shit

This is a satire account. Look at the username and scroll through some of the older posts especially.

Is this post is foreshadowing a fix?

This is not a human glitch! It is a UI design flaw.

You shouldn’t feel embarrassed. This is a UI design flaw that should be fixed.

It has not. They do run stuff that is critical of the right, etc and I’ve heard some good analysis of the stuff going on in the last 10 years, but hooo boy have they normalized and legitimized trump/maga. It’s infuriating.

Yes 🤮

Best of luck

I'm really pessimistic about this. I think what's going on tech-wise with atproto and bsky is exciting, but we know where this is going. How do we pay for all of this without enshittification? If there isn't some new novel plan, I have to assume ads and selling user data are coming.

I hear that! They could make it work that way, I hope they figure it out bc I bet you’re not the only one

You’d have to make two separate accounts, one for each app.

Thanks a lot for this; I had some big gaps and this fills it in more!

It’s referring to @skylight.social

What does an AppView implementation entail? · bluesky-social atproto · Discussion #2961 An informal overview of what an AppView does. Similar to #2350, which discussed PDS implementations. This document isn't "normative": AppView implementations are not required to do all of this. It ...

(Maybe) answering my own question: github.com/bluesky-soci...

Gonna do some reading!

That would be news to me! I’ve been under the impression that appview = client.

I’m just learning myself and have yet to jump into any dev. I haven’t come across a great rundown of exactly where the boundaries of app views are (I’m mostly been skimming the docs).

Lmk if you have any good links.

You are correct, skylight is just an appview and right now they don’t have any of their own infra; it’s all Bluesky under the hood. They haven’t said anything publicly about plans to spin up their own stuff behind the scenes… I hope they do! You are correct Spark is building separate infra.

...to save them time and effort over doing it manually. For me, the jury is still out on whether that's possible. I do apologize for dropping a random "ai sucks" comment on your thread out of the blue, and I hope this wall of comments explains where I'm coming from.

I'd love to do that in the future or read someone else's study on a similar task. I still wonder if AI can empower an operator with detail skills, sufficient domain knowledge, AND sufficient vigilance to not just let the tech do its thing without really QAing it...

So the promise of "I don't know this, let's have AI teach me" in my experience is a mirage. It's sorta there in some cases, but the more I use it the less I find it useful. I still don't know if AI is "worth it" for the alt text task. I did not scientifically study our attempt.

If it's an infographic about atproto relays, you need someone who has enough knowledge to QA this thing. If I'm generating alt text for an image in a physics textbook, I need to know enough about the physics material on the page to make a judgement call on whether the AI got it right.

Take a step back and the output all looks beautiful. You don't catch them unless you are looking carefully. So you have to have an operator that is sufficiently detail-oriented and is able to engage carefully with the content to actually tell if something is right or wrong.

...it including details. But once we tried it at scale we discovered major problems regularly (e.g. it gets a character's name wrong, or in a kids books about dinosaurs it calls an allosaurus a tyrannosaurus, etc). These types of mistakes can be subtle and insidious.

...atproto knowledge is rudimentary, but based on my experience with genai, I'd be skeptical. My experience is trying to use it to write alt text for ebooks. At first it looked very promising-- it seems like magic that we can have software look at an image and in many cases accurately describe...

I believe that's all accurate. But if you wanted to communicate just those points, you could do better than this. Furthermore, how well does this really sum up atproto.com/specs/sync? How well does it tell you "How ATProto Relay Works"? I can't say right now because I haven't read it and my...