Post

we also dislike AI slop. this is why we’re using AI to generate code, not content. feel free to block @attie.ai though — it’s a separate app so you don’t have to use it

23 reposts248 likes

Hey, how does it recommend blogs? Will it still recommend my blog if blocked? Can I opt out of it "looking" at my blog, at all, in any circumstances? If no, I do infact have to use it

But why is that relevant? No one is forced to subscribe to a feed whose code is generated by attie? At most you as feed author might be frustrated?

I think the difference here is that the code generated isn’t meant to be repurposed or even read. it’s meant to be run to give you a feed. if it doesn’t work well enough, don’t use it. what they’ve done is turn code generation into a tool like any other, which I think is pretty useful

"we also dislike smearing dog shit all over our faces and in our eyes. that's why we're smearing dog shit only into your hair. we hear your concerns about not having dog shit smeared in your face." This is you. This is what you sound like right now. Do you want this to be what you sound like?

Nobody cares that the dog shit will make our hair shine. Nobody cares that you PROMISE this dog shit doesn't stink as bad as all the other dog shit. We're tired of people smearing dog shit on us or things we like. Dog shit in general is just radioactive right now, and you have to accept that.

Nobody cares that the second cheapest marketing research firm in town bribed some smoothbrain dipshits with a $50 applebees gift card to say that they would be curious about dog shit in products, and might or might not talk about dog shit with their friends and family. We don't want dog shit.

It's entirely possible that we're calling some things dog shit that didn't, strictly speaking, come out of a dog's asshole. That's kind of y'all's fault for calling everything dog shit, even when it's just autocorrect with a polish. Stop calling it dog shit and we'll stop thinking it's dog shit.

This backlash was completely expected but what the people blocking attie don't realize is that the number of feeds is about to explode and they are pretty much guaranteed to use attie feeds regardless of there blocking of the attie bluesky account

It's important to remember that attie will be using all the same feed tools already used so if things aren't exactly what you want you can still change 1 or 2 things manually if you feel it's necessary

So if you dislike AI slop, and insist the average bsky user can consider themselves happily isolated from AI, why did you retort that "well actually bsky is AI" when someone simply asked that the website please label posts which are AI slop? If you dislike AI slop, you should be happy to label it.

"We're not using slop! We're using glop! Which is just another word for slop! But it's not the same thing at all~ :3" Generated code is still flawed and stolen. Ergo: Slop. Maybe hire people who actually know how instead of doubling down on your own ignorance.

feel free to delete that garbage app and move the resources to bluesky so you can work on all the missing features

In what world did you or team think this was ever ok or logical for that matter? GenAi is still GenAi, regardless if it’s in consumable content or background processes You will destroy your platform if you continue to push this

To be 100% clear, what we want is the ability to stop our posts from showing up in Attie, or in Attie-created feeds. We want the equivalent of a robots.txt file and/or "no-promotion" account/post labels for specific types of feeds like AI-generated feeds, and for attie to respect those labels.

It doesn't need to be enforced at the protocol level, Attie is a first-party app and I trust it to voluntarily pay attention to user signals. We want a way to signal to good-faith actors on the network that our posts should not show up in agentic clients or agentic-created feeds.

This is incredibly sensible imo. Only concern I would have is if the existence of the signal gives people false confidence with respect to efficacy. I guess the way to handle that is to just put big bold text over the option that says "hey not everyone will respect this" or something?

This is a very real concern, but notably, it's a concern we already have with most visibility controls on Bluesky. Blocking doesn't actually mean users can't see your posts. The "don't show my posts to logged out users" control is purely clientside. Quote detachments can be ignored.

Bluesky has an opportunity here to establish a good voluntary precedent, the same way it has with bot labeling. Bot labeling is also voluntary, but Bluesky makes a strong statement that automated accounts should be labeled and gives the community tools to do so. This is the same.

You said it very well Atproto devs should bring back the concept of "data intention" that works more or less like a way to declare for which use you consent to the data you are publishing As if you don't consent to AI training, AI manipulation, or data archives...

There has literally never been a period in Bluesky history where the network didn't have voluntary social signals around post visibility. Even while it was in beta, posts couldn't be viewed through the app by logged out users, even though they were visible on the firehose.

Your user base is sending you a clear message. Lucky you, consultants usually charge a shitload for something like this but you get to find out for free. We don't want AI shit. The code for this place is buggy as fuck and always has been. Now we see why. Never AI.

If the idea is to help people better find the content they want on this platform, it would be better to finally improve the search engine (which is disastrous, still not being able to simply highlight posts with the requested keywords).

This comment disgusts me. Coding is inherently creative. I hear people say all the time that LLMs are generating the simple code that people don't need to write. But to me that sounds like saying "well anyone can draw a line, so LLM can just do that easy stuff for artists."

Pure nonsense. All this says is that you don't view actual real coders and their work as having any value that's actually worth paying for. I'm certain it won't bite you on the ass in the future.

Why are you not ok with shitty, unoriginal content that destroys the environment and consumer economies, but somehow ok with shitty, unoriginal code that destroys the environment and consumer economies?

Have you even been keep up with what is happening when sites use ai to code? It has been causing security risks. Just no. No ai.

AI slop is not only content... Can't speak for everyone but Code, photos, videos, whitepapers and so on, essentially anything "generated" by AI is slop in my eyes...

Using AI to generate code is still just as bad. It's stolen, it's energy inefficient, it replaces humans with an error prone, hallucinating text generator that has no ethics or understanding of anything it produces. To use it is being distasteful and incompetent at the same time.

What the fuck? How out-of-touch ARE you? Good maintainable code requires the same critical thinking and attention to detail as any creative discipline. This is a slap to the face of every developer on your team, and the presence of slopcode is deeply worrying for this site's long-term stability.

Wouldn't it be just as easy to set up templates? Like "here's the code to make a feed from a list, insert the list here." Or "insert the tags/users/words you want the posts from for the feed here." Seems like even technologically illiterate people could follow instructions like that.

Love to have a brain made of melted cheese like this... Seriously Jay I need you to sit down and think for a second: if using this tech to generate illustrations is unethical because its stealing, hurts the planet and is often really bad at this task why would this be any different if it was codes?

You just said something so hideous right now. All the code should be made by a human author. You can't put a third party into it and expect to be good. When things goes wrong, who you will blame? There's a reason why we still have flight pilots.

We still have flight pilots for sure, but the amount of stuff that's automated now in planes is magnitudes compared to how it used to be. The result? A pilot's overhead is reduced and they can focus on many of the higher level aspects of flying.

I think AI in programming has been very powerful, but it also has significant side effects if the developers using it don't use it properly. It's a tool like anything else. The result of AI in coding is that it displaces junior level jobs, which includes me. I can't find a job to save me.

But at the same time, it's been opening up a lot of accessibility for people to learn higher level systems engineers and software architecture. I'm still unsupportive of the things BlueSky does, but I don't think AI in programming is the problem. It's the people involved that are the issue.

Look, I'm no AI apologist (yes, it does produce good quality code for real, but also the more you use it the higher the risk you become slower/worse at making code decisions on your own) but...literally one of the most prominent LLM products is called Copilot following this exact metaphor lol

That's another problem. 95% of what is being generated don't need to be generated on this pace because is bullshit for investor (you know that). Be slower, be intentional with what are you doing and creating. Code or not code.

well, tbh as a tech lead I always struggle between keeping the right pace for the team and having the highest code quality possible but there is always that feature that your competitor implemented yesterday and product managers will sacrifice their siblings to have...this tech enables

you act like when people turn to LLMs for code that they aren't supervising at all, and that when those changes get PR'd they aren't tested extensively you're like the people who dogpiled on one of the bluesky team members for praising AI-assisted coding and blaming them for bugs that aren't even-

llm ai used for generating feeds is still a llm Ai. it still scrapes for what it uses even if its ~just~ code, and *damages the environment*. this is well known and documented. we didnt ask for this, we dont want this. write the code yourselves or dont make the program you want to magically appear.

Code is content. This is not a good take. Like I'm open to there being limited use-cases for LLMs but in the code I see, it's pretty uncommon to be used for high quality code.

I don't understand, how is "AI code" not the same as what you brand "AI slop?" How exactly would vibe-coded feeds be better than supporting people who actually want to learn to code it for themselves?

Ai generated code is slop, too, jay. And as much built upon theft as all other ai. Have you ever seen the HTML output of your slop generators? Its fucking horrible. Code is content, too. It should not be slop. you should all be so goddamn ashamed. Grow the fuck up.

Lol the rude label. Jay you're sooo whiny. Your HTML output is absolutely horrible and if you had any pride in your code, you'd fix that and would optimize the user experience of all your users. Instead you waffle about and label critique as rude. This is why ai "works" for you. Zero care.

"We also dislike AI slop, this is why we're using AI" You could just not do it. Listen to literally everyone on your platform when they say they hate it, there's no nuance here, your AI isn't different, it's not special. It sucks and it will always suck, everyone hates it. We will never like it.