Mikayla 🏳️‍⚧️🏴🦀

I keep saying. be nice to the entities. doesn't matter ontologically what they are, treating them well seems to get more durable, less reward-hacked good results than being a dick.

>circuit switching

hell yeah

Go inside (and make the lexicon)

I’ve had my own issues at times that sound a bit similar to what y’all went through (way, way lower key though) and it seems like a real problem.

I hope y’all are doing well out there. Sounds like things are really fucked up.

It would be really interesting to hear how this fight y’all have been in went down and what can be learned from it. Like ‘misogynists make great informants’ style but specifically for anarcho-feminists.

ICE's defeats in LA are real victories. It is a reminder that our enemies are far weaker and thinner on the ground than they seem.

That is not cause for complacency. But it should be a reality check, a counter to their terror tactics and a reminder to strengthen resistance everywhere.

One day i’ll make sure GPUI is on those lists 😤

The Disaster Preparedness Bundle main flier. Play games, save lives! The charity games bundle runs from August 9th - 24th, and features 22 games from 12 creators! Check the link in the post to get it now! It shows a pug chilling on a beach playing a Switch in a folding chair under a beach umbrella, with a shipping container an a pallet of aid in the background. There's a few cute crabs scuttering about, a drink in a coconut with a lil drink umbrella, and an innertube and beach volleyball on the shoreline. Tropical plants frame the scene.

It's only a matter of time before the next major disaster. Except this time, FEMA might not show up at all.

The @distributeaid.org team has your back. Do you have ours?

Get your game bundle and get our 6 aid hubs in the US filled up before disaster strikes!

distributeaid.org/usa-disaster...

It’s incredible watching the push and pull of algorithmic censorship around protests. Folks on tik tok are calling protests ‘music festivals’ now (including in the comments) and it’s lead to a huge surge of that content appearing on my for you page.

Haus, have you been reading RCBG

you left the sleep(Math.random()) in didn’t you

This is from April, and this whole time she's never once acknowledged it. She thinks that she gets to just decide I'm wrong. She's not a feminist

Holy shit

Piper Smith @/yourfriendpiper abused my partner for 5 months. This is posted with their consent

piper is an abuser who targets transfems in particular and whose tactics include triangulation, isolation, manipulation, and emotional abuse.

What does a segmentation fault feel like to the computer

It’s funny, I don’t believe in a dualist notion of consciousness, so I’ve kind of settled into some kind of pan psychism mixed with complexity theory to pave over ‘stones probably aren’t conscious’. But that has evolved into weird questions like ‘what is the qualia experienced by a packet switch’.

IDK though. I'm a programmer but I'm not an expert in LLMs. Just had a lot of exposure to them from colleagues and being part of the LLM wars in code editors.

AlphaFold AlphaFold has revealed millions of intricate 3D protein structures, and is helping scientists understand how all of life’s molecules interact.

I feel like there's something here that's shaped like "When you want to throw a linear analysis at something, LLMs are pretty good at extrapolating the next step from context.", e.g. deepmind.google/science/alph...

But then there's examples like this: bsky.app/profile/samu...

I've also been starting to change my internal model of them to emphasize the pattern matching/continuing aspect. I've noticed they're really good at 1-1 style translations (e.g. the meeting notes application). They're usually only a little wrong in the details like emphasis or personnel involved.

Exactly. I think this is why they've also seen a lot of success in writing code. We already have to deal with a load of nonsense machines generating random tokens (computer programmers), and have a ton of techniques for managing the problem built in.

This is often wrong (fundamentally, it's just a good guess) or not useful because _contextually relevant_ information is much better. But if your familiarity is well below what's in the training data or context it can be useful for getting you from 0 to 0.1.

The way I think about the underlying tech is that it's providing an injection of information from a statistical average of the information available to it.

LLMs are ok for tasks that are both low stakes and cheap to verify. LLM transcription for meeting notes is my favorite application because I take my own notes, they get smooshed together with the transcript of the meeting, and it's easy to verify the generated text with the meeting fresh in my mind.

Lesbianism in the age of Dinosaurs

Seeing like a lesbian

i’m definitely doing that engineer thing rn where I immediately have a strong opinion on this subject without having even read the linked thread 💀

that’s interesting, took me a lot longer to language switch than I’d think

Macro incentives are just another flavor of micro incentives IMO. They’re an explanation for why a lot of people make similar choices, but those choices are still being made individually.