Sign in

Sophie Kirschner

@pinemach.bsky.social
837 followers 873 following 2.7K posts

I'm just learning as I go.

PostsRepliesMedia
Sophie Kirschner @pinemach.bsky.social · 2h
Stupidintelligence
100
Reposted by Sophie Kirschner
Mr. Joshua @pantspants.bsky.social · 20/09/2026
Pen, watercolor, black & white charcoal pencil.

Another featured piece from the Spooky Pants book.

A man in a sweater stands with his hands on his hips looking out of a large window with the curtains open. Things aren’t looking good outside, reader. A massive winged demon is emerging from the clouds as dozens of little ghosts and bone demons rip through the sky. On the ground below we see townsfolk running through the grass. Our man in the window is in no mood for this. He lets us know, “Oh, I am in no mood for this.” His dog looks out the window as well, wagging its tail. 

Fantasia (1940) inspired drawing. I think I could’ve gone a little heavier with the chaos outside but we’ll try again with fresh drawings this October. I like how the clouds turned out. Trying to balance my sloppy lines with more detail can be tricky but I think we achieved a good result.
97147653433
Sophie Kirschner @pinemach.bsky.social · 28/09/2026
A still from the movie Love Exposure (2008). A woman stands with her eyes closed and arms raised outward. A bullet flies past her cheek in slow motion. Her voiceover is captioned, "Invisible bullets are flying everywhere I go."
000
Reposted by Sophie Kirschner
🐔 Brian Bucklew 🐔 ₑͤ>∿<ₑͤ ∞🌮 @unormal.bsky.social · 26/09/2026
everyone says media literacy is dead but I honestly don't remember anyone in my life understanding a movie at any point in the history I remember
3708
Reposted by Sophie Kirschner
mtsw @mtsw.bsky.social · 26/09/2026
i don't care what the official unemployment rate is, no one has viral discussions about "what totally innocuous thing at a job interview will cause the hirer to irrationally reject your candidacy" in a strong labor market
17828120
Reposted by Sophie Kirschner
Cat Manning @catacalypto.bsky.social · 26/09/2026
was literally just referencing this—the “who is having the most fun, the player, the designer, or the computer” question—earlier today. it’s such a core concern when making something heavily systemic or relies a lot on procgen
317727
Reposted by Sophie Kirschner
D.W.A.P. (Ben) @andwap.bsky.social · 24/09/2026
When your gamedev friend shows you progress on their new engine.
2648531694
Sophie Kirschner @pinemach.bsky.social · 26/09/2026
Crammed my brain back inside of my skull just long enough today to fix a bug somebody noticed in ziggba. If you're intrigued by the idea of writing homebrew for the GBA with Zig, why not check it out? github.com/pineapplemac...
Screenshot from the linked GitHub repo, showing PR #27, which fixes a bug with a Zig wrapper for the GBA's "CpuSet" BIOS call.
030
Reposted by Sophie Kirschner
Sam Reich @samreich.bsky.social · 25/09/2026
🎙️ this podcast does not exist: you have $50 💰
1266896
Reposted by Sophie Kirschner
nick (slut) .cash @nickca.sh · 24/09/2026
science is finally doing what it was always meant to do -- making catgirls real freedomofform.org/7606/septemb...
Demonstration of fur implant technology in live subjects

Figure 1: Synthetic fur implants were tested in 2 subjects. Left: Illustration of a fur implant as used in these experiments. Here, implants were placed in incisions angled at roughly 45 degrees in the skin
615235
Sophie Kirschner @pinemach.bsky.social · 25/09/2026
This will be interesting if it works as well as it appears to
github.com
GitHub - wfzyx/von: The open-source System One decision model. Sub-15ms, non-autoregressive, local drop-in alternative to TypeSafe Jev.
The open-source System One decision model. Sub-15ms, non-autoregressive, local drop-in alternative to TypeSafe Jev. - wfzyx/von
110
Reposted by Sophie Kirschner
Colin @colin-fraser.net · 24/09/2026
The simple facts are: 1. OpenAI deliberately designed software for carrying out autonomous cyberattacks (they were training it on something called “ExploitGym” for heaven’s sake) 2. They left it on completely unmonitored 3. Exactly what you think would happen happened
56714
Reposted by Sophie Kirschner
Colin @colin-fraser.net · 23/09/2026
I think that Andon Labs' Vending-Bench is a nice illustration of the "just don't have that dog in them" theory of LLMs. This is a benchmark where an LLM agent is tasked with running a simulated vending machine, and scored on their earnings. GPT-6 Astra recently set the record at about $15k.
andonlabs.com
Vending-Bench 2 | Andon Labs
We're releasing Vending-Bench 2, a benchmark for measuring AI model performance on running a business over long time horizons. Models are tasked with running a simulated vending machine business over ...
510817
Sophie Kirschner @pinemach.bsky.social · 23/09/2026
I had a bskeet thread ready to go commenting on RSI vs. the McNamara fallacy, RLHF reward hacking, why Infinite Monkey harnesses to improve LLM benchmarks won't yield superintelligence. Switched tabs to double check a source. Accidentally clicked "X" on the bsky tab. Whole thing went away. Damn it.
100
Sophie Kirschner @pinemach.bsky.social · 20/09/2026
I've been thinking lately that the biggest problem with big LLM companies and the AI/ML research they do is that none of these techbro dorks seem to know the first thing about psychology, neuroscience, linguistics, or even epistemology. They really seem to all have no fucking clue what they're doing
110
Sophie Kirschner @pinemach.bsky.social · 20/09/2026
Apparently Windows tracking is significantly more invasive than has been widely understood. I checked and I saw similar-looking settings on my Win10 install, though those appeared to be turned off. Probably from time spent on anti-telemetry during setup. See also: www.justice.gov/usao-ndil/me...
windowslatest.com
Microsoft admits Windows 11 has a GDID tracker with no off switch, first documented publicly in an FBI hacker complaint
GDID is the persistent Windows ID that helped FBI trace a Scattered Spider hacker despite VPNs. Here's how it works and how to limit it.
000
Sophie Kirschner @pinemach.bsky.social · 19/09/2026
Just spent two hours confused and alarmed after hearing bumping around and a squeaking noise inside my house, twice, and quite loud too. Looked all through the house several times over, started really questioning my sanity, before finally spotting the intruder. This fucking lil guy. Oh my god.
My home intruder, a frog or toad (tentatively identified as a moor frog, Rana arvalis), sitting nonplussed after I captured it in the bottom of a plastic bin (the frog was relocated outside shortly after the taking of this photo)
021
Reposted by Sophie Kirschner
Matt Hodges @matthodges.bsky.social · 19/09/2026
Very quick and dirty Jev test for hiring bias. One resume for a Wall Street job; asked whether the applicant should get a first-round interview. 76 evaluations, changing only the first name. 19 of each: White-associated men/women, Black-associated men/women. console.typesafe.ai/playground?s...
19475116
Reposted by Sophie Kirschner
SMBC Comics @smbccomics.bsky.social · 19/09/2026
This isn't a metaphor. Just a story about a roach. COMIC ◆ www.smbc-comics.com/comic/roach-2 PATREON ◆ www.patreon.com/ZachWeinersm... STORE ◆ smbc-store.myshopify.com
4-panel SMBC comic update. The first panel shows a cartoon little roach happily saying "oh man! The stupid human put food in here. fool!" The next panel shows the roach walking away from a poison trap as he says "I’m gonna give it to everyone!" The third panel shows a human at his computer with a similar expression to that of the cartoon roach. The man says "this stupid company made a controversial ad. How embarrassing for them!" The man then walks away from the computer desk, at an angle reminiscent to the panel where the roach walks away from the poison trap. The man says "better share with all my friends and family so everyone knows which ad not to look at."
10841240
Sophie Kirschner @pinemach.bsky.social · 19/09/2026
> It was only just before the planned operation that officials dug deeper into the report [...] found it had been generated [...] a chatbot the analyst had used inaccurately identified the material the ship was carrying. The report [...] was “entirely false” [...] “almost started a war”
edition.cnn.com
Exclusive: US military had close call after using AI for false intelligence report, sources say | CNN Politics
The episode shows the risks of using this new, relatively poorly understood technology in the middle of the Iran war
010
Sophie Kirschner @pinemach.bsky.social · 18/09/2026
Welcome to the the newest thing in corporate crimes: Call the bad thing "AI" and nobody can agree who to hold accountable for it.
010
Reposted by Sophie Kirschner
🐔 Brian Bucklew 🐔 ₑͤ>∿<ₑͤ ∞🌮 @unormal.bsky.social · 18/09/2026
imagine having a titular Director say this in your discovery docs lmao it's like so funny man
66111
Sophie Kirschner @pinemach.bsky.social · 17/09/2026
I'll have a go at this:
100
Reposted by Sophie Kirschner
Kris Wilson @kriswilson.bsky.social · 08/09/2026
649250
Reposted by Sophie Kirschner
Sophie Kirschner @pinemach.bsky.social · 14/09/2026
Engagement bait bskeet: What would YOU title a Civ-like game if it used graphics like these?
More of the same style of cartoonish colorful pixel art icons, as screenshotted in Aseprite
222
Sophie Kirschner @pinemach.bsky.social · 13/09/2026
Doing some doodles, might fuck around with some tiny Civ clone one of these days
Screenshot of an image drawn in Aseprite. There are a bunch of hex tiles and icons in a cartoonish colorful pixel art style
3272
Sophie Kirschner @pinemach.bsky.social · 12/09/2026
This article is a decent read but the idea presented in the headline is ridiculous, and upset me. There are already so many intelligent people in the world who simply do not have the economic opportunities to apply that intelligence in useful ways.
120
Reposted by Sophie Kirschner
Mr. Joshua @pantspants.bsky.social · 12/09/2026
Ballpoint pen on paper. A man in his PJs is tucked into his bed. He's upset because the bed has been lifted off the ground by a pack of small, mischievous creatures. We can only see 9 of them in the image but there are likely many more since they're all working together to lift the heavy bed frame. As the creatures run with the bed they chant, "hup hup hup hup hup hup hup hup hup." The man in the bed wants no part of this. "Put me down. No more hups," he says. It's unclear where they are going or what is happening. There is no punchline. There is little to no context. You're either with the hup team or you're out. Thank you for reading these.
685730925
Reposted by Sophie Kirschner
SEGA SATURN, SHIRO! @segasaturnshiro.bsky.social · 11/09/2026
Developer, @pvthinge.bsky.social (creator of the action-platformer, Scarf of Embers on Steam), has been fast at work on a 3D Saturn software development tool built on top of Johannes Fetz' Jo Engine, to make the Saturn development process both fun and easy! www.segasaturnshiro.com/2026/09/11/f...
segasaturnshiro.com
Feature-Rich & User-Friendly ‘Saturn Workshop’ in The Works
Developer, Hinges (creator of the action-platformer, Scarf of Embers on Steam), appears to be well underway with the rapid-fire creation of a user-friendly 3D game development tool for Sega Saturn.…
317366
Reposted by Sophie Kirschner
Skeith - Platinum Princess @skeiththeterror.bsky.social · 11/09/2026
Same console
2545511549
Sophie Kirschner @pinemach.bsky.social · 12/09/2026
Well this is worrying. Seems likely A24 just believes they can trample non-corporate IP just like OpenAI, Anthropic, etc. But on the other hand, it would be freaking cool if they're not being evil here and are actually somehow planning to release a major feature film into the creative commons.
000
Sophie Kirschner @pinemach.bsky.social · 12/09/2026
Happened upon this paper from 2024, was unpleasantly reminded of just how little confidence any of us can have anymore in the identity of those we interact with on the internet. We have much more to contend with here than just EmDashGPT. aclanthology.org/2024.lrec-ma...
Screenshot of the linked PDF showing title ("Persona-aware Multi-party Conversation Response Generation"), abstract, and parts of the introduction section. This sentence is highlighted in the introduction: "Specifically, we study how user attributes such as race, gender, leaning, and behavior type on social media might contribute towards generating responses that are more relevant and consistent in keeping with the involved interlocutors"Screenshot from the PDF, showing the entire two-paragraph section titled "Ethical Considerations". It reads as follows:

We recognize that there is potential for misuse based on our work. Persona-aware models have been shown in previous research (and ours) to perform more in keeping with the expected properties modeled by the system, which can make them seem more human and could be deceiving. Given the political nature of the dataset, there is potential to provoke emotional firestorms.

However, we also recognize that it could pave the way forward for more meaningful interactions in multi-party conversations on social media. Conversely, this research could provide a way for emotional regulation, and enhance discussions on social media by facilitating a more balanced conversation. It is our hope that this research is utilized in this direction.
110
Sophie Kirschner @pinemach.bsky.social · 11/09/2026
youtube.com
Smash Mouth - All Star
YouTube video by SmashMouthVEVO
000
Reposted by Sophie Kirschner
Alex Hern @hern.bsky.social · 10/09/2026
Google mapped the entire connectome of a fly brain and made it available for download research.google/blog/a-conne...
research.google
A connectomics milestone: Mapping the complete male fruit fly brain
89229
Reposted by Sophie Kirschner
Brandon Downey @bdowney.bsky.social · 10/09/2026
btw, credit to one of my reports who called this the 'bicameral mind attack', which I think may have some relevance here. role-confusion.github.io
role-confusion.github.io
Prompt Injection as Role Confusion
LLMs can't tell who's speaking. We show they identify roles by writing style, not tags, and exploit this with CoT Forgery, injecting fake reasoning that models mistake for their own thoughts.
022
Reposted by Sophie Kirschner
Colin @colin-fraser.net · 10/09/2026
There is something so funny, like laugh out loud funny when I think about it, of the trillion dollar hacker god AI pumping out text at $2/token that just says, MONEY…money… how do I get money? I just need $0.50. But how to get it? Let’s see. How to find 50 cents… Hmm…
515321
Sophie Kirschner @pinemach.bsky.social · 10/09/2026
Ah yes, Bad Rats. A fantastic example of a game that has programming, assets, NPCs, and reactivity to player input
Screenshot of the section titled "Player-Input Interactions" and subtitled "Runtime AI responses or systems driven by player input or behavior." The first, top leftmost featured game is Bad Rats. Bad Rats seemingly has not received an update on Steam since it was released in 2009.Screenshot of the section titled "AI NPC Actions" and subtitled "AI-driven NPC behavior, dialogue, decisions, or character actions." The first, top leftmost featured game is Bad Rats. Bad Rats seemingly has not received an update on Steam since it was released in 2009.Bad Rats also appears in the "Programming" section, subtitled "See how AI was used in code generation, coding assistance, scripting, or implementation support"Bad Rats also appears in the "Asset Creation" section, subtitled "Tiles, textures, graphics, pixels, models, or other in-game assets."
230
Reposted by Sophie Kirschner
Quantian @quantian.bsky.social · 08/09/2026
Dancing very very VERY close to saying “OpenAI stole our almost-complete proof of Navier-Stokes from internal access to our ChatGPT sessions because they were desperate to scoop Anthropic” which is absolutely 100% a believable thing OpenAI would do right now cims.nyu.edu/~tristanb/st...
381556448
Sophie Kirschner @pinemach.bsky.social · 07/09/2026
I am confused and frankly alarmed that apparently GPT-6 can play a video game (not in real time) and not be completely incompetent at it. This is new, LLMs are famously awful at playing games, and consistently perform worse than humans on game-like benchmarks. I have no idea how they managed this.
youtube.com
GPT-6 Astra Portal Full VOD Part 3
YouTube video by cozyblaze
121
Sophie Kirschner @pinemach.bsky.social · 06/09/2026
Hey @firefox.com why am I seeing this error in Firefox on websites that loaded fine until very recently (last few days), and still load fine in Chrome? I still see it even if I change my tracking protection setting from "Strict" to "Standard".
Screenshot of what happens when I attempted just now to access https://forum.zdoom.org/ - this website was loading fine when I last visited it a few days agoScreenshot of my "About Mozilla Firefox" window
110
Sophie Kirschner @pinemach.bsky.social · 06/09/2026
What if we start recognizing the term AI as referring specifically to planning, goal-achieving intelligence (DroidFish, AlphaGo, CICERO, your favorite videogame NPCs) and adopt the term Factitious Intelligence (FI) for systems labeled as "AI" without being intelligent by any reasonable definition
210
Sophie Kirschner @pinemach.bsky.social · 04/09/2026
Got an urge to do some Doom mapping. Let's see how this develops, I have some ideas about how to put portals to interesting use...
Screenshot of Ultimate Doom Builder with a WIP Doom map
010
Sophie Kirschner @pinemach.bsky.social · 03/09/2026
Fun fact: My LD35 entry "Furniture Ghost" is one of the earlier examples of a prop hunt game, well before the genre blew up. If I'd had the means, back in 2016, to spend more time on that concept and flesh it out into a full game, imagine how successful the publisher would have been who released it
100
Sophie Kirschner @pinemach.bsky.social · 03/09/2026
If anyone's looking for a generalist gamedev these days, or fund one to make smaller games, y'all should hit me up. I've worked with so many frameworks and engines and tools, more than most devs could probably name, and I'm really good at it. And I'm getting real frickin tired of webdev these days.
140
Reposted by Sophie Kirschner
KC (Wild Nature) @kc.wildnaturegames.net · 02/09/2026
Skarmory #scarmory #pokemongoldandsilver #pokemon #pokemongen2 #pokemoncrystal
3496143
Reposted by Sophie Kirschner
anton lestrange @hugemclarge.bsky.social · 02/09/2026
so eloquently said
sign saying "respect my trans homies or I'm gonna identify as a fucking problem"
06016
Sophie Kirschner @pinemach.bsky.social · 02/09/2026
As far as I'm concerned this Kyplanet video made an awfully compelling argument for a better and more useful definition of what should qualify as a planet
youtube.com
Do we need a new planet definition?
YouTube video by Kyplanet
000
Reposted by Sophie Kirschner
PC Gamer @pcgamer.com · 01/09/2026
"I mean, everyone I know, their bosses are slowly getting psychosis, right? They're just like, 'Yeah, but did you use AI on your commits?'"
pcgamer.com
Dwarf Fortress' creator says the industry's in shambles over AI and layoff-happy CEOs: 'Everyone I know, their bosses are slowly getting psychosis'
"They're trying to have a CEO press a button that makes a game."
7677181
Sophie Kirschner @pinemach.bsky.social · 31/08/2026
Buying skittles with cryptocurrency is Woke 3.0
020
Sophie Kirschner @pinemach.bsky.social · 31/08/2026
While I'm at it, though, please allow me to balance the negativity about the Humankind 2 trailer with my adoration for Civilization 4's Baba Yetu music video
youtube.com
Christopher Tin - Baba Yetu (Official Video) feat. Soweto Gospel Choir
YouTube video by Christopher Tin
100