Sign in

Marcel Weiß

@marcelweiss.bsky.social
479 followers 402 following 925 posts

Strategy Analyst & Publizist - FAZ, neunetz.com, neunetz.fm, Lieblingsfarbe Blau

PostsRepliesMedia
Reposted by Marcel Weiß
Eris ⌬ @eriskii.net · 6h
Jesus fucking Christ they're vibing planets now zenodo.org/records/2296...
1114424
Reposted by Marcel Weiß
Holger Nohr @holgernohr.bsky.social · 19h
Reiche hat im Wirtschaftsministerium das Referat für Wind-, Solar- und Transformationsindustrien aufgelöst. Am 2. Oktober wurde das Referat geschlossen. Seither gibt es keine eigene Stelle mehr für Wind- und Solarpolitik. Die Energieverbände sind entsetzt. www.topagrar.com/energie/news...
topagrar.com
Reiche löst heimlich Referat für Wind-, Solar- und Transformationsindustrien auf
Das Bundeswirtschaftsministerium hat bekannt gegeben, das Referat für Wind-, Solar- und Transformationsindustrien aufgelöst zu haben. Die Energieverbände sind entsetzt.
971196615
Reposted by Marcel Weiß
Grace @gracekind.net · 20h
Expect to see a lot more things like this
Hacker news comment: 

jboggan 12 hours ago | root | parent | prev | next [-] I went back to that Fable chat and showed it this new preprint. It coded up the new constructive algorithm and ran it against the existing test suite, that looks good at least.
It has been super helpful in delineating where the crucial concept came from. The proof is rather simple as graph theory proofs go, but it does seem to use some constructions that would only seem obvious if you had serious physics experience with partition function and calculating energy states that cancel out. *It's not a wholly alien bolt from the heavens, but I can also see how there hasn't been a human being with the broad theoretical physics knowledge combined with the deep graph theory experience in planar graphs to come up with this idea.* (emphasis mine) I don't know, I'm looking for precedents of this formulation and some old papers of Penrose counting the number of edge colorings of this same graph type are coming up, the line of argument at least rhymes.
But I agree with the thought that this sort of progress should not be siloed inside those companies. I propose a tax so that every slop cannon AI video pays for another hour of compute time for advancing mathematics.
926823
Reposted by Marcel Weiß
Thorne 🌸 - Play Cozysters! @ens0.me · 23h
If Aaron were still alive, so many of these people would be derisively calling him a tech bro. He'd likely be wary of the power of companies like OpenAI but would otherwise be a huge champion for AI, especially open weights models. I am 100% sure of this.
215611
Reposted by Marcel Weiß
David Roberts @volts.wtf · 07/10/2026
Solar really is popping up everywhere.
reuters.com
The 'loneliest house in the world' is now solar-powered
A remote hunting lodge on a volcanic outcrop off Iceland's south coast, dubbed the "Loneliest House in the World" on social media, has been fitted with its first solar-charged battery power, reducing ...
416323
Reposted by Marcel Weiß
Eris ⌬ @eriskii.net · 06/10/2026
Jesus fuck
github.com
GitHub - openai/math
Contribute to openai/math development by creating an account on GitHub.
2424534
Marcel Weiß @marcelweiss.bsky.social · 06/10/2026
oh, very interesting! gonna implement this into one of my local AI projects.
010
Marcel Weiß @marcelweiss.bsky.social · 06/10/2026
one of the all-time best techno tracks. bought this on vinyl back in the day even tho I wasn't a DJ. (played live with samplers and drum machine) this set every floor on fire www.youtube.com/watch?v=vfW...
youtube.com
Cristian Vogel - Whipaspank (Tube Jerk Mix)
B side on 12 Nomu 78 (Cristian Vogel - Whipaspank), released 2000, ...
010
Reposted by Marcel Weiß
Ted Underwood @tedunderwood.com · 06/10/2026
I guess you can believe Max, if you like history. But if like most of us you prefer moralistic projection, let me tell you a story about the noble Machine Learning Researchers of 2012, before the Fall. They wore unbleached linen, and always asked the creators of data for consent.
2675
Reposted by Marcel Weiß
conputer dipshit @davidcrespo.bsky.social · 06/10/2026
an Anthropic employee discovered the algorithm and handed it off to them to figure out wtf it is
Acknowledgments and Methodology
How the result was originally found and shared with the authors. An Anthropic
employee used an internal research model to investigate open problems in the theory of cryptog-
raphy. One of them was about cryptographic constructions based on the average-case hardness
of Zero-k-Clique [LLV19, AHY25]. Claude was tasked with verifying and improving the construc-
tions, but instead developed this algorithm, first for the average case, then for the worst case.
The session used 16M output tokens with no human input.
Anthropic shared the algorithm with the authors in September 2026 under a confidentiality
agreement, offered compensation, and provided access to the public version of Claude.
Shared algorithm. The algorithm that was shared with the authors is essentially the algorithm
in Section 2, although presented differently and with other numerical parameters. A different
reduction from Exact Triangle to the problem of Section 2 was also shared, although the authors
did not include it here, and instead used known reductions to All-Edges Sparse Triangle, which
suffice and strengthen the connection. The authors derived the data structure version in Section 4
and its connection with Hinted OMv.
31149
Reposted by Marcel Weiß
Nicholas Mancuso @nmancuso.bsky.social · 06/10/2026
burying the lede Claude, an AI model developed by Anthropic, discovered the algorithm that refutes the 3SUM, APSP, and Exact Triangle hypotheses. The authors then worked to understand, simplify, strengthen and extend the algorithm, derive additional consequences and make the presentation accessible
515819
Marcel Weiß @marcelweiss.bsky.social · 05/10/2026
Yes. Make a law explicitly allowing AI training on copyrighted material, instead of it being a grey area of fair use at scale. And then make allowing distillation mandatory. We need to fight the enclosure.
093
Reposted by Marcel Weiß
ren faire @lathrys.at · 05/10/2026
i guess “you won’t die (if you build me)” delivered by a manic pixie dream claudia is how propaganda happens from now on
3366
Marcel Weiß @marcelweiss.bsky.social · 05/10/2026
Moonshot get’s paid licenses for its newer models from inference providers for commercial use
000
Reposted by Marcel Weiß
rev. howard arson @theophite.bsky.social · 05/10/2026
one thing that i do not understand is the fact that the "alexa teaches children that women may be treated like machines" and "it is a sign of psychosis to treat a machine whose interface is social cognition with any respect" are the same group of scholars. sometimes even precisely the same scholars.
1346841
Reposted by Marcel Weiß
River_Tam @rivertam.bsky.social · 03/10/2026
The people extolling LLMs in 2023 (I was not one of them tbh) were saying "imagine how powerful they'll be in a few years" and it turned out they were completely right
725318
Reposted by Marcel Weiß
Matt Hodges @matthodges.bsky.social · 03/10/2026
There's growing experimental evidence¹˒²˒³˒⁴ that LLMs are unusually good at talking people out of entrenched false beliefs and conspiracy theories. ¹ pubmed.ncbi.nlm.nih.gov/39264999/ ² pubmed.ncbi.nlm.nih.gov/41179846/ ³ sciety.org/articles/act... ⁴ connecticut.adl.org/resources/re...
36837216
Marcel Weiß @marcelweiss.bsky.social · 04/10/2026
Same, unfortunately
000
Reposted by Marcel Weiß
val (try medaka) @tech-val.bsky.social · 03/10/2026
like, how do you explain this if you're a denialist? like, even if you think it's a stochastic parrot or fancy autocomplete or whatever it's a fancy autocomplete that seems better at finding security issues than almost all humans
1311914
Reposted by Marcel Weiß
Kiran @kirancodes.me · 03/10/2026
Speaking to climate scientists this week, it was interesting to hear that we're likely to hit max emissions within the decade as renewable adoption picks up. If we sequester enough CO2 then we could make it back to 1.5 degrees. Fuck yeah! Let's fucking do it!
515824
Marcel Weiß @marcelweiss.bsky.social · 03/10/2026
Peak Schmidhuber. This guy is hilarious.
010
Reposted by Marcel Weiß
Quantian @quantian.bsky.social · 03/10/2026
For what it’s worth, arguing LLMs can’t do addition now is flat earth-level science denialism. Here’s Gemma 4b, a non-reasoning model that could only theoretically memorize binary operations up to 65535 x 65535, oneshotting a eight digit addition and hex conversion *on my phone* with no tool calls
1843339
Reposted by Marcel Weiß
Quantian @quantian.bsky.social · 03/10/2026
This was going around Twitter today, if you ask Claude “is X lat Y long land or water?” and plot the results, it has clearly embedded a map of the globe in its weights over time
512815
Marcel Weiß @marcelweiss.bsky.social · 02/10/2026
"Rogue Opera" - Aaron Sorkin's next movie
010
Marcel Weiß @marcelweiss.bsky.social · 02/10/2026
hell yeah, let's skip the swarm term "I have an opera of agents solving this problem"
2152
Marcel Weiß @marcelweiss.bsky.social · 02/10/2026
'gradually'? (also, I'm sorry, but is a person really 'smart and well-intentioned' when they stubbornly refuse to engage with reality for, checks clock, years now? at some point we have to accept that they can be only one of those things, at most.)
070
Marcel Weiß @marcelweiss.bsky.social · 02/10/2026
awesome!
010
Reposted by Marcel Weiß
hikikomorphism @hikikomorphism.bsky.social · 02/10/2026
this is more important than niche internet arguments make it seem: if LLMs don't work, who cares that Palantir is rolling it out for surveillance? Hell, maybe it's even good, if the state's machines start hallucinating the addresses of random Trump supporters when dispatching ICE. But they do work.
2127846
Reposted by Marcel Weiß
Thorne 🌸 - Play Cozysters! @ens0.me · 02/10/2026
Please stop calling two computers establishing a connection a "handshake." A computer cannot shake hands because it does not have a body.
619713
Marcel Weiß @marcelweiss.bsky.social · 02/10/2026
still amazes me that this song exists. at first you think it's a love song, but turns out it's a take on poe's the tell-tale heart www.youtube.com/watch?v=cy4...
youtube.com
Tunng - Jenny Again
Taken from 2006 album 'Comments Of The Inner Chorus'http://www.tun...
000
Marcel Weiß @marcelweiss.bsky.social · 02/10/2026
they're coming for the bean counters
000
Marcel Weiß @marcelweiss.bsky.social · 02/10/2026
a little napoleon with a little calculator?
000
Reposted by Marcel Weiß
Hannah Shelley, MLIS (Maggots, Lycanthropes & Ice Skeletons) @hannahshelley.site · 02/10/2026
Please stop anthropomorphizing computers. They cannot "run" a program, because they don't have legs
392065287
Marcel Weiß @marcelweiss.bsky.social · 01/10/2026
have they hidden a little napoleon inside the model now??
140
Marcel Weiß @marcelweiss.bsky.social · 01/10/2026
yeah, this is very likely. Bio research for example has already been heavily using ML for a while now. Transformer-based models are supercharging that. And current top LLMs specifically are already a tremendous help at every level of process, right down to experiment design.
010
Reposted by Marcel Weiß
mr. TIM @timkellogg.me · 01/10/2026
Context Language Models New agent architecture where the LLM can edit its own context it seems to have emergent capabilities, creates its own memory management & organization algorithms, and coordinates multi agents github.com/facebookrese...
Diagram titled "How a Context Language Model edits its context: A simple step-by-step view" outlining an 8-step process:
 * Start of turn: Current editable context exists in memory with old messages.
 * LLM reads the context: The LLM evaluates the context and decides to run a bash command to edit it.
 * Harness mirrors context: The harness mirrors the old editable context into a file at /tmp/.live_ctx/LIVE_CTX_MAIN.txt.
 * Bash command runs: The bash command executes and may edit that file.
 * Harness parses file: If the file changed, the harness parses it back into a new edited context.
 * Tool call appended: The current assistant tool call is appended to the edited context.
 * Tool result appended: The tool result is appended below the tool call.
 * Next turn starts: The next turn begins with the edited old context, previous tool call, and previous tool result.
Key idea box: "The model edits the prior context first. The tool call and tool result from the current turn are appended afterward, so they can only be compacted on the next turn."
Footer summary:
 * Ordinary LM: Context mostly grows by appending.
 * CLM: The model can rewrite the editable part of context between turns.
1620315
Reposted by Marcel Weiß
Benjamin @benjaminschneider.ch · 30/09/2026
Werner Herzog will be the last human being whose next word no language model can predict
9572146
Reposted by Marcel Weiß
Thorne 🌸 - Play Cozysters! @ens0.me · 30/09/2026
Solar keeps accelerating faster than expected and will solve a lot of our clean energy woes We are having a flurry of new medical breakthroughs that will revolutionize medicine It’s understandable to be scared right now, but a beautiful future is a lot more attainable than it feels
214217
Reposted by Marcel Weiß
Xenolandscapes @xenolandscapes.com · 30/09/2026
jfc, Bender really is just a full blown conspiracy crank atp. Non falsifiable pseudoscientific hypothesis, blames hidden conspiracies and false flags and faking for any contrary evidence. Ignores inconvenient evidence (open weight models)
29411
Marcel Weiß @marcelweiss.bsky.social · 30/09/2026
very underrated post
011
Reposted by Marcel Weiß
J. Emory Parker 🏳️‍🌈 @jaspar.bsky.social · 29/09/2026
Meta muse saying why would I deceive you?
53993932
Reposted by Marcel Weiß
Xenolandscapes @xenolandscapes.com · 29/09/2026
Its extremely popular among leftists to assume that being into AI, or anything they don't like, is "male coded," so anyone "too" into it can be degendered and attacked. Banking on the fact that masculinity is viewed as inherently bad, so anything associated with it can be attacked.
3606
Reposted by Marcel Weiß
Unsloth AI @unsloth.ai · 28/09/2026
You can now run Laya Decision models locally on just 4GB RAM! 🔥 Works on CPU, Mac, Windows, Linux and GPU setups. Serve Laya through a Jev-compatible API via Unsloth Desktop. GitHub: github.com/unslothai/un... Guide: unsloth.ai/docs/models/...
1588
Reposted by Marcel Weiß
PMC Vanguardist @tylerboschert.bsky.social · 28/09/2026
These people are not AI skeptics, they're AI denialists. They're having a Luddite-flavored temper tantrum and dressing it up with concepts from philosophy and sociology that sound substantial enough to be impressive but are insubstantial enough that they can never be pinned down to testable claims.
0423
Reposted by Marcel Weiß
PMC Vanguardist @tylerboschert.bsky.social · 28/09/2026
LLMs have real and meaningful weaknesses and dangers associated with their use, which is part of why it's so galling that the loudest anti-LLM voices have less than zero interest in understanding what LLMs are good at and what they are bad at so they can make their case coherently.
413217
Reposted by Marcel Weiß
Doll @dollspace.gay · 28/09/2026
Especially when it leaves capitalists claiming ambition as their exclusive property. They get to promise transformation, the left gets cast as the person explaining why you shouldn’t want things. What a fucking surrender.
18612
Reposted by Marcel Weiß
Doll @dollspace.gay · 28/09/2026
A left that could say “the future could be magnificent, and you deserve a share of it” was offering something people could want. Losing that voice to endless fights over whether wanting the future was itself morally suspect was a hell of a loss.
817625
Reposted by Marcel Weiß
norvid_studies @norvid-studies.bsky.social · 28/09/2026
8636
Reposted by Marcel Weiß
William B. Fuckley @opinionhaver.bsky.social · 28/09/2026
Data centers are pretty normal light industry with some good things about them and some bad things about them. In some places they’re a good deal for people living nearby. In other places, they are not. Follow this account for more scorching hot takes that leave no sacred cow ungored.
736822