wizardry variants kib

Unfortunately the working turntable I have doesn't fit inside my cabinet and I'll have to be on the lookout for a smaller one but otherwise I've got my antique stereo cabinet working again and am celebrating with some classics

The chocolate vaguely carved out like the art on the pouch

The chocolate put back in the ripped up pouch with a zip tie around it

Wasn't that hard really! Ready for trick or treaters!

A pumpkin shaped chocolate with part of a jack o lantern face carved out. The peanut butter layer is several mm deeper than the packaging suggested

The art is deceptive about how deep the chocolate layer is

Pumpkin reeses packaging with a pumpkin shaped chocolate with a jack o lantern face cut out. In tiny text it says "decorating suggestion"

A pumpkin (or blob) shaped chocolate with no face carved

This bag of Reese's cups turned out to be a mixed bag, and has these pumpkin shaped ones.

Except, it doesn't have a face like the picture.

That's a "decorating suggestion"

So I'm pulling out the xacto knife

It's amazing how many AI articles state as fact things just being said by AI companies who want to sell you their AI software

One time I was hanging with some nerd friends and this guy we didn't know well was also there and he tried to insist that Uranus and Neptune were never in the Sailor Moon anime

There were like five of us yelling at this man

edited comic panel. Batman is breaking a lump-with-handle cow tool over his knee and saying "this is the tool of the cow. we do not understand it. we will not use it."

it's not FOR us

YouTube video by ohnoitisnathan14 Toto Coelo - Dracula's Tango (Sucker For Your Love)

It's that time of year

www.youtube.com/watch?v=ROtq...

They spent a long time saying more users wouldn't fix their problems then at the last second before dying went "we need to push for more users"

And like I said somewhere in here, if Cohost had continued to grow they'd be fucked because they'd HAVE to get a second moderator and HAVE to pay them $94k

SA mods didn't cost $94k per so that makes a difference lol

Cara did this, they didn't limit their hosting (which they could have done) and blew up with new users and their next hosting bill was ABSURD.

When you see people trying to flee some larger site you need to clamp down on new users, but I understand why you don't want to because more users = more revenue. Unfortunately it also means more expenses so if you don't do the math carefully you can fuck yourself over.

Which I think is what Bluesky did, and what I think Cohost did. You could tell from the financial reports Cohost got a boost in subs when Tumblr drove a bunch of people off in drama and Twitter doing the same at some point.

Yeah and you could pay money for all sorts of other dumb shit people liked paying money for. I think this is also why Pillowfort survived its early years, you had to either pay $5 or have an invite from someone.

The worst thing a platform can do is let itself balloon in size too fast

It's basically just that Blum doesn't CARE if the movies are good, they just need to be cheap. Which means most of them are slop. But sometimes you get a Soft & Quiet that's good AND cheap to make.

Blumhouse works in bulk, they produce 500000 movies and a few of them might genuinely be good

And because Cohost paid everyone the same, the second you added another employee to handle that huge new amount of users, you've undone all the gains you had, because that's another $94k/yr

It's a losing battle and that's why large scale social media sucks (large = not a 10 person fedi server)

Like even with cohost...their monetization model would have required them to probably double or triple the userbase

And even if they did and even if most of those people became paid subscribers, adding more people would make hosting cost more and...you need another mod at that point!

And that's the bad part like it DEFINITELY isn't going to pay for itself if you do it ethically

But even most of the places doing horrifically unethical shit aren't actually getting profitable with it

It doesn't scale well at all

Social media is never profitable, and can only be self sustainable if you don't intend for it to be your day job. Or if you do a very good job it can maybe be one person's day job.

If you aren't taking nonstop VC funding that's going straight down the drain, then it's all very short lived.

Unless you're one person running a fedi server on beer money. Or something like Pillowfort where the CEO doesn't take a salary.

It is. I mean...reddit just became profitable like ten seconds ago. I don't know if Twitter was ever profitable. People assume these businesses must be profitable because they've existed so long but they're just endlessly spending a nonstop supply of VC funding

And then there's generative AI which I have yet to find a use for that doesn't just disgust me lmao

So when I say I'm anti AI I don't mean I am disgusted with the idea of being able to search my photos for "Little My" and get pics of my dog. But I am disgusted knowing that's going in a database somewhere and we can't just have nice sensible things.

The problem comes down to how interconnected it all is. I know that I'm training Google's shit when I use these features. I know it's training off my dog photos whether I want it to or not. I know they're using those captchas of theirs to unleash poorly tested cars that will drive straight into kids

Now, it'll probably also miss a cat pic or two. Well. That's fine, I'll still have deleted MOST of the cat pics in my review, I'll find the ones it missed on my own but it helped me get through the bulk of it.

That's a totally good use of this tech!

Let's say I have a huge photo library and I want to delete every photo of a cat. I'm done with cats.

AI is pretty good at telling if there's a cat in a pic. It might mess up and show me one of my precious dog pics, but I'm deciding what to delete. If I let it decide it'd be bad! I'd lose dog pics!

ML/LLM AI stuff's genuinely good utility is in its ability to filter information under two conditions:

1) you're filtering data for human review, not using it verbatim

2) false positives and false negatives do not cause any real problems (heavily tied into #1)

The thing is if the AI accidentally misidentifies something else as my dog, ah well, it doesn't matter much. If it fails to find the dog photo I wanted, ah well, I'll dig and find it.

You can't "ah well" for the police state using it to ID people.