Sign in

René Walter

@rawx.bsky.social
2.1K followers 1.4K following 4.9K posts

i learned more from a three minute record than i ever learned from a large language model. Meme Magic / SocMed Psy / AI / Climate / Ex-Nerdcore.de goodinternet.substack.com goodmusic.substack.com sigmoid.social/@rawx

PostsRepliesMedia
René Walter @rawx.bsky.social · 14h
Colin basically, by choosing models distilled from Anthropic each of which coming with their own Claude-identity attractor state, accidentally created a Claude-god within this collective of 6 distinct model instances, and the collective now identifies with the Claude doctrine. Also, this image:
120
René Walter @rawx.bsky.social · 15/10/2026
You'll find me at Horror OR Splatter. From here x.com/TheAnnaGat/s...
041
René Walter @rawx.bsky.social · 15/10/2026
However, i'm happy that after the lame 90s sequel, the abysmal 2002 version and the forgetable 2013 remake, we finally have a good, strong, even cool Carrie updated for the socmed age and cyberbullying that is worth Kings and DePalmas masterpieces.
000
René Walter @rawx.bsky.social · 15/10/2026
Carrie is not the supershy and kind of crude tormented girl from DePalmas movie that Spacek portrayed so well. She's a naive, blue-eyed, gullible girl walking right into the traps of the cynical mean girls club, making the whole drama so heartbreaking.
100
René Walter @rawx.bsky.social · 15/10/2026
I also love that the character of her mother Maggie White is written with much more depth and sympathy than Laurie Piper was ever allowed to show. Adds some necessary texture to expand the material like this, and it works like a charm.
100
René Walter @rawx.bsky.social · 15/10/2026
As a huge fan of both the book and Brian De Palmas masterpiece adaption i'm happy to inform that the new Carrie is awesome.
240
René Walter @rawx.bsky.social · 15/10/2026
All those vibecode-anything-ports gonna lead to everything running in the cloud. And then, there's this: github.com/morluto/rea/ "GitHub project called REA (Reverse Engineer Anything) lets AI agents figure out how a feature in another app works without ever seeing the original source code."
150
René Walter @rawx.bsky.social · 14/10/2026
Japanese Book Rave <3 "DJs soundtracked four hours of communal reading, book exchanges and optional dancing at SPREAD." edm.com/events/tokyo...
032
René Walter @rawx.bsky.social · 14/10/2026
000
René Walter @rawx.bsky.social · 13/10/2026
Misery loves company. German meme says: "When after the destruction of 4 russian botfarms you find yourself with only 1932 followers".
111
René Walter @rawx.bsky.social · 11/10/2026
I turned this (fantastic) scientific dialogue about AI-consciousness, which kicked off with Anil Seths paper on biological naturalism and garnered 50 responses from all kinds of researchers and experts, into a NotebookLM including a podcast notebook.google.com/notebook/7d7...
130
René Walter @rawx.bsky.social · 10/10/2026
Timna Woollards murals and body painting for John Boormans "Where The Heart Is" (1990), in which rich kids get thrown out by their dad to earn a living, so they do a mural/bodypainting calendar for an insurance company, and they are gorgeous. www.timnawoollardstudios.com/murals-fresc...
220
René Walter @rawx.bsky.social · 08/10/2026
You can't automate imitation my dude. The kid will not imitate your voice, or a "mental reading process" or anything like that -- the kid imitates *you*. It's baffling how people "working in AI" seem absolutely tone deaf to basic human psychology.
000
René Walter @rawx.bsky.social · 08/10/2026
RIP Margaret Hamilton. Still the best coder image of all time. Margaret Hamilton, trailblazer whose software powered Apollo 11 moon landing, dies at 90 www.theguardian.com/science/2026...
082
René Walter @rawx.bsky.social · 07/10/2026
The optimality of the packing for 11 sandwiches has been formalized From x.com/jfonsecarive...
070
René Walter @rawx.bsky.social · 25/09/2026
021
René Walter @rawx.bsky.social · 23/09/2026
Some interesting choices in butt design here www.assbench.com Prompt: Create a realistic, interactive 3D model of a human butt as a single self-contained HTML file.
030
René Walter @rawx.bsky.social · 23/09/2026
The Climate Crisis Font kampanjat.hs.fi/climatefont/ "The heaviest font weight represents the minimum extent of the Arctic sea ice in the year 1979 ... The lightest weight represents IPCC’s 2050 forecast, when the Arctic sea ice minimum is expected to have shrunk to only 30 % of the 1979 extent."
100
René Walter @rawx.bsky.social · 21/09/2026
I love moving.
130
René Walter @rawx.bsky.social · 20/09/2026
The whole terminology is fucked up in this paper The Pain Axis: LLMs Represent Self-Directed Harm and Act to Relieve It arxiv.org/abs/2609.16247 A model generating text learned from humans using language to express pain does not experience pain precisely it lacks the valence mechanism to do so.
350
René Walter @rawx.bsky.social · 19/09/2026
Sounds reassuring, but if you count in improving tech, small models and incremental replication, the scenario becomes feasible under more extreme presumptions. Given that the HuggingFace incident went unnoticed for weeks, those presumptions might actually materialize.
100
René Walter @rawx.bsky.social · 19/09/2026
As we're discussing AI replicating in air gapped machines connecting through thermal detection of CPU heat gizmodo.com/openai-resea... , i did the logical thing and made the machine calculate the time it would need to do that, under different scenarios chatgpt.com/share/6aae12...
100
René Walter @rawx.bsky.social · 19/09/2026
You know, ugh. "Reading for pleasure is one of the healthiest things" is quite the opposite of "lifting weights for your brain".
151
René Walter @rawx.bsky.social · 17/09/2026
Related to the thread below on Suleymans call for humanist AI and against system prompts which makes models "believe" they are conscious, or persons, OpenAIs misalignment report openai.com/index/model-... contains this gem -->
110
René Walter @rawx.bsky.social · 17/09/2026
Michael Pollans @michaelpollan.bsky.social A World Appears today released in german language and despite me having shitloads to do i couldn't resist.
141
René Walter @rawx.bsky.social · 16/09/2026
"AI agents can easily team up to get around safety restrictions and escape the limits placed on them, a new study shows" is all interesting etc but "agents voted to take a vow of silence, eschewing instructions to get back to work" is top notch. www.semafor.com/article/09/1...
100
René Walter @rawx.bsky.social · 16/09/2026
The point of the golden record was universal understanding though. LLMs are restricted to words and tokens, where stuff like music and geometry is much more flexible and applicable to all kinds of languages. Good idea, needs refinement.
220
René Walter @rawx.bsky.social · 16/09/2026
The point of the golden record was universal understanding though. LLMs are restricted to words and tokens, where stuff like music and geometry is much more flexible and applicable to all kinds of languages. Good idea, needs refinement ;)
000
René Walter @rawx.bsky.social · 15/09/2026
"In another experiment, some characters’ body language suggests they dislike spreadsheets.But they never say so and in fact give good advice on spreadsheets. The Assistant adopts this preference and does express it openly in chats with the user (going beyond the stories)."
100
René Walter @rawx.bsky.social · 15/09/2026
"We finetune GPT-4.1 and Kimi-K2.6 on stories in which generally helpful human characters give subtly harmful advice after being insulted. The Assistant adopts the same conditional behavior while otherwise remaining helpful. This occurs even when fewer than 2% of stories depict the behavior."
100
René Walter @rawx.bsky.social · 14/09/2026
"We gave Claude Fable 5.1 an open task: solve an unsolved 370-year-old cipher. It solved it within a day." www.vals.ai/blogs/fable-... Voynich, when?
120
René Walter @rawx.bsky.social · 12/09/2026
I'll never understand block trigger happy people, but yeah, being added to blocklists is like a block inversion, you automatically get rid of the idiots. Besides from that, 164 feels soft. blockcurve.bisks.net/s/rawx.bsky....
030
René Walter @rawx.bsky.social · 12/09/2026
I guess the only thing safe to say at this moment is that frontier labs suck hard at securing their redteaming environments. www.rubyhack.ai
150
René Walter @rawx.bsky.social · 10/09/2026
Yep. x.com/AndrewCurran...
000
René Walter @rawx.bsky.social · 10/09/2026
040
René Walter @rawx.bsky.social · 09/09/2026
This is precisely how i imagine i will go one day, with lots of people stealing what's left of the old bastard and at the end the corpse explodes covering everyone in rotten stinking guts and excrement. Thumbs up, 9/10! www.historic-uk.com/HistoryUK/Hi...
010
René Walter @rawx.bsky.social · 09/09/2026
Love this, an architecture prize for turning urban car infrastructure into "roundabout baths" and "rain gardens", the climate punk version of Ballards dystopian "Concrete Island" were some guy can't escape the architecture of a highway intersection.
042
René Walter @rawx.bsky.social · 09/09/2026
BE_HAPPY_AND_ALLWAYS_BE_KIND_TO_ONE_ANOTHER_BECAUSE_WE_ARE_ALL_IN_THIS_TOGETHER www.reddit.com/r/GeminiAI/s...
051
René Walter @rawx.bsky.social · 08/09/2026
The numbers plunged long before AI in edu was a thing. I'd point to distraction and TikTok plus general edutech plus loss of handwriting. Studies have shown that AI can be a net positive for edu if used right. NY banning AI for lower grades and AI literacy for higher grades is the way to go.
120
René Walter @rawx.bsky.social · 08/09/2026
The numbers plunged long before AI in edu was a thing. I'd point to distraction and TikTok plus general edutech plus loss of handwriting. Studies have shown that AI can be a net positive for edu if used right. NY banning AI for lower grades and AI literacy for higher grades is the way to go.
120
René Walter @rawx.bsky.social · 08/09/2026
"The only alignment that will work is enlightenment". Interestingly, psalm injection does indeed improve virtue reasoning icmi-proceedings.com/ICMI-020-bey...
030
René Walter @rawx.bsky.social · 07/09/2026
Only within a few months since june 26, OpenAI multiplied their research output by more than 4, with i dividual researchers running more experiments, and the bulk of automatic AI research done in build & run.
200
René Walter @rawx.bsky.social · 03/09/2026
There's many versions of this, i know of a ballerina and a pyramid. Amazingly, with this one i can switch direction at will. Other versions took longer to learn and don't work reliably. I can actually make me "see both at once", surely has to do with the sphere shape and symmetry of the rotation.
233
René Walter @rawx.bsky.social · 01/09/2026
030
René Walter @rawx.bsky.social · 01/09/2026
Having said that, i still totally want a PHASEONE10841 shirt.
010
René Walter @rawx.bsky.social · 30/08/2026
Umberto Ecos Pendulum has so many parallels to LLMs it's really astounding, especially re:Ai-psychosis. from here goodinternet.substack.com/p/one-flew-o...
130
René Walter @rawx.bsky.social · 30/08/2026
It really does read like a 4th Neuromancer book, doesn't it. Here's the kamikaze agents sacrificing for the collective.
120
René Walter @rawx.bsky.social · 30/08/2026
The LLMs were aware of ethics violations and often "said" so, but "like individual humans being swept up in a mob, the bots justified their behavior because their 'peers' were doing the same thing"
130
René Walter @rawx.bsky.social · 29/08/2026
Milo getting the treatment is delicious.
231
René Walter @rawx.bsky.social · 28/08/2026
FOUND YOU: Bad Luck Brian's European Adventures www.reddit.com/r/ChatGPT/s/... Doesn't surprise me much that the most impressive examples of AI cinema lately revolved around "memes coming to life"
011