I tend to refer to OpenID for the branding concept but I don't know how successful OpenID's branding was.

(So you can distinguish apps that utilize atproto more broadly from the use of atproto purely as an IDP, mostly)

I am once again asking us to design consistent branding for atproto login that also indicates type of use.

My kingdom for a pair of thick, strong, calloused hands with faint blood stains in them slowly and gently caressing someone.

Early information was wrong in that it didn't turn out to be a provider issue. Instead a small mistake from a long time ago came back and bit bsky in the ass.

This is pretty normal in software.

Weird angle to take when game streaming continues to be an incredibly niche concern due to whatever the inverse of a chicken-and-egg problem is.

Well that's upsetting.

it's all happening!

plan9 really makes so much more sense in a world where tailscale exists.

aaaaa that's so sick

Don't even document it, just silently attempt a connection to the websocket port on localhost if it's available and send a jolt though an available device if the video has been paused for too long to remind the user that they're in the middle of watching a talk.

Is. Is that why there are all these weird TERF doctors like Hillary Cass running around?

Most of this information is half-remembered from Strange Aeons's interview with former Tumblr engineer cyle on YouTube, and from LinkedIn. Any errors are probably mine.

I should clarify: the person I mention in this thread DID NOT run Denny's social media WHILE AT TUMBLR (and they also ran the Twitter and Facebook accounts for Denny's). It's just that Tumblr saw their work and hired them because those skills are what they wanted.

What the fuuuuck is this bullshit.

Memex 2 is happening.

enclose.horse Day 101
💎 PERFECT! 💎 100%

Now that's an enclosed horse.

This is actually effective AI slop, which I don't like.

Oh is it §1466A education hours again?

Curious how much of this is Google's LLM specifically. Like, this problem isn't solvable, and will always limit LLMs in search problems, but also, the fact that the AI Google uses for search is actually WORSE about this than any other AI (worse than Real Gemini!) cannot be helping.

(since every modern operating system now providers some kind of isolation primitives, this feels possible? But I'm uncertain due to the differing capabilities of different systems.)

One day there will be a tool that delivers Nix's fundamental model in a way that is rigorous and cross-platform, using containers as the isolation primitive. I am very eager for that day to arrive. It has not arrived yet, so for now everything sucks, just in different ways.

Nazism isn't bad because republicans do it. Republicans are bad because they want to do nazism.

The reason I'm asking you is that I want to understand your opinions about AI. I don't want to spend a bunch of time talking about issues you don't care about or areas where we fundamentally agree.

Anthropic isn't saying "pay us to protect you from our model," they're saying "pay us to protect you from the model a chinese lab will build a year from now."

This is the nature of the vulnerability arms race, and was long before AI entered the picture. State actors spend vast sums of money to find vulnerabilities that are not publically disclosed. Defensive groups spend huge sums to find those issues and get them fixed.

The thing you seem to be missing is that if Anthropic can do this, so can anybody else.

like, the fate corrupted joan of arc? or a different one?

👀

Well, what does a positive case even look like to you?