Search results

FediThing 🏳️‍🌈 @FediThing@chinwag.org
Public

This is a really excellent non-fiction piece by @WeirdWriterRobert Kingett about a writing group with a tech bro:

sightlessscribbles.com/the-col

It is a distilled essence of the social and cultural damage AI/LLM is causing, how AI promoters are cynically destroying people's confidence in their own humanity, while simultaneously trying to ridicule and other people who point out that AI is bullshit. (And this isn't even mentioning the environmental consequences.)

The Colonization of Confidence., Sightless Scribbles

A fabulously gay blind author.

sightlessscribbles.com

0
0
0
0
0
0
Jaeyeol Lee @kodingwarrior@hackers.pub
Public

https://github.com/ComposioHQ/awesome-claude-skills/tree/master/skill-creator

Claude Skill 기능을 적극적으로 활용해보려고 하는데, skill을 만들 수 있도록 돕는 skill-creator라는게 있다. 이걸 좀 더 참고해서 어떻게 나한테 쓸만한걸 만들 수 있는지 한번 살펴봐야겠다.

A curated list of awesome Claude Skills, resources, and tools for customizing Claude AI workflows - ComposioHQ/awesome-claude-skills

awesome-claude-skills/skill-creator at master · ComposioHQ/awesome-claude-skills

A curated list of awesome Claude Skills, resources, and tools for customizing Claude AI workflows - ComposioHQ/awesome-claude-skills

github.com · GitHub

1
1
0
J❀y @lurkjay@mastodon.social
Public

Healthy radium cigarettes

This one was inspired by a tweet I saw a few years ago, something along the lines of “machine learning is the radium salts of the twenty-first century”. This joke is more relevant than ever; corporations are trying to squeeze “AI” into every role imaginable, no matter that it’s mostly useless and often harmful.

Full size (3000×3500, 1.13 MB): deviantart.com/lurkjay/art/105

Painting of a white pack of cigarettes on a dark blue background. The front of the pack says “LLM” in dark red slab-serif letters, underneath it is a line of dark grey sans-serif text saying “healthy radium cigarettes”.
0
0
0
0
0
0
mcc @mcc@mastodon.social
Public

One thing about "AI" is with the technology OpenAI has (large neural network plus manual tagging) you could've made the best search engine ever. There could be a Copilot where you describe what you wanted and it finds an example of it in the corpus of open source software. You could go from a fuzzy image description to a stock image. These would be better than buggy code and fucked up images. But they wouldn't do that because the *service* OpenAI provides is obscuring that the content is stolen.

Wulfy @n_dimension@infosec.exchange
Public

@mcc

Not sure when you've used 👉properly👈.

In my experience the more vocal opponent of AI is the further back in time their (lack of use) goes.
With the most ardent opponents having never used the models, yet having most empathic (and increasingly inaccurate) opinions.

Attached media, a public query from today, with sources dropdown at the bottom.

Approx 30% of web searches comes from the engines nowadays.

(Edit: Hahaha, insta blocked by poster, I guess folks don't like to be called out on saying patent provable falsehoods 🤡

The poster, made a comment exposing their ignorance of features of existing AI. This one has 33,000 followers, question is "How many others like them have zero idea about the systems they critique"?)

Chatgpt with sources
0
0
0
Radio Azureus @RadioAzureus@mastodon.social
Public

This is an interesting a well-crafted article

The title should be don't shove large language models down our throat.

True artificial intelligence is that what you have and your games, your simulators, etc which has been here for many many decades. This is large language model talk, a whole other Beast

Interesting read recommended

gpt3experiments.substack.com/p

0
0
0
@reiver ⊼ (Charles) :batman: @reiver@mastodon.social
Public

vibe coding

2/

A number of the people who attended demoed what they made — including my wife's friend.

It was interesting to see how vibe coding was enabling people without programming skills to create applications.

Are their applications as good as applications created by career software-engineers‽ — no, but that is probably OK. Their vibe coded applications seem to be good enough for their needs.

What is interesting, though, is —

@reiver ⊼ (Charles) :batman: @reiver@mastodon.social
Public

vibe coding

3/

What is interesting, though, is —

This (non-programmers using vibe coding to create applications) reminds me of something I noticed decades ago about spreadsheets —

People who are bright who don't know how to computer-program use spreadsheets to create applications

Are their spreadsheet-based applications as good as applications created by career software-engineers‽ — no, but they are good enough for their needs. And, that's fantastic!

0
0
0
@reiver ⊼ (Charles) :batman: @reiver@mastodon.social
Public

vibe coding

2/

A number of the people who attended demoed what they made — including my wife's friend.

It was interesting to see how vibe coding was enabling people without programming skills to create applications.

Are their applications as good as applications created by career software-engineers‽ — no, but that is probably OK. Their vibe coded applications seem to be good enough for their needs.

What is interesting, though, is —

0
0
0
0
0
0
Alessia Visconti @alesssia@mas.to
Public

Despite being extremely vocal against , I strongly suspect that too many juniors (not just students) are using it for reports and manuscripts. I have no problem in dedicating my time to edit a draft so that someone with less experience will improve their , but editing is really starting to annoy me.

How are other people in (including juniors) dealing with this? How to tag (or self-tag) someone's own work?

0
0
0
0
0
0
W3C Developers @w3cdevs@w3c.social
Public

⏰ The deadline to submit a proposal for the @w3cWorld Wide Web Consortium workshop on Smart Voice Agents (Feb 2026, virtual) is 27 Nov 2025!

Smart voice agents need clearer use cases, stronger , better -based interaction, and improved accuracy, , and multilingual support. Broader concerns include device coordination, , , regulatory gaps and emerging business models.

Don’t miss your chance to present your work and submit now: w3.org/2025/10/smartagents-wor

W3C Workshop on Smart Voice Agents - February 2026, Virtual on zoom
0
0
0
Glyph @glyph@mastodon.social
Public

LLMs are *absolute concentrated brain poison* for these folks. They try out the LLM to see if it can solve a few simple problems and then they extrapolate to more complex problems. Wrongly. They infer from social cues in their cohort, which are absolutely fucked by the amount of synthetic money (and maybe fraud?) driving a subprime-bubble type mania. They infer from the plausibility of its outputs, which are absolutely fucked because the job of these models is to produce plausible outputs.

C++ Wage Slave @CppGuy@infosec.space
Public

@glyph

I've said it before, but the antivenin for this brain poison is to boot up and try some really small . With a sufficiently small model, it's completely obvious that the machine has no understanding, no consciousness, no intelligence, and no mental model of the problem it's being asked to solve. That insight equips the user to interact with a larger and not be bamboozled by its plausibility.

ollama.com/search

Ollama Search

Search for models on Ollama.

ollama.com

0
0
0
0
0
0
Metin Seven 🎨 @metin@graphics.social
Public

All mentions of "AI" should be changed into "Brain Rot Inducer"…

"Do you want the Brain Rot Inducer to summarize this text for you?"

"The Brain Rot Inducer can auto-answer this e-mail for you!"

"Let the Brain Rot Inducer write this social media post for you!"

"Easily generate an image with the Brain Rot Inducer, and call it your own creation!"

0
0
0
Ayo @ayo@ayco.io
Public

What would probably increase the quality of responses is if they add hyperlinks to high quality sources from real . As in, not just mashed up footnotes, but actual words/phrases/sentences within their responses linked to real people/orgs websites. Obviously they should be correct reference links too. No need for web to die.

0
0
0
FediThing 🏳️‍🌈 @FediThing@chinwag.org
Public

If you've invented a useful technology or fun gadget which people genuinely want to use, you don't have to say things like "It's here to stay like it or not" or "You'd better get used to it" or "It's not going away".

I don't recall anyone from Nintendo parading around with the Wiimote screaming "YOU'D BETTER GET USED TO THIS WHETHER YOU LIKE IT OR NOT!"

If you DO have to use such phrases, maybe people don't actually want it or find it useful?

0
0
0
Greg Lloyd @Roundtrip@federate.social
Public

@nicoleNicole 🏳️‍⚧️ 🌈 @mjdMark Dominus

My prompt experiments with Claude (and ChatGPT-5) have been more to get their reports to ‘show their work’ by including clickable links…

Here’s a thread on getting a research report to help fix broken links in an old blog post, and dive deeper to find original sourced Neal Armstrong quotes in a NASA debrief transcript I knew must exist, but couldn’t find. federate.social/@Roundtrip/115

Screenshot of a 26 Aug 2012 blog post ‘Remembering Neil Armstrong…’ by Greg Lloyd, TractionSoftware.com after Armstrong’s 25 Aug 2012 death. The post starts with a quote by Armstrong, writing in ‘The New Engineering Century: ”I am, and ever will be, a white-socks, pocket-protector, nerdy engineer — born under the second law of thermodynamics, steeped in the steam tables, in love with free-body diagrams, transformed by Laplace, and propelled by compressible flow.” Lloyd remembers asking a question to Armstrong at an event for high school students in Spring 1996, just after the Gemini VIII mission. The post includes tributes to Armstrong, links to his New York Times obituary, NASA mission logs, an Armstrong NASA oral history interview, a link to the original 1996 Gemini VIII Mission Crew Debrief, and other nerdy references.

Greg Lloyd (@Roundtrip@federate.social)

Attached: 1 image @MissConstrue@mefi.social 🧵ChatGPT-5 as reference link fixer #LLM I agree. I have however been able to easily prompt and poke ChatGPT-5 to review a 13 year old blog post about Neil Armstrong (I gave it a URL) and suggest fixes for any broken links it found iin the post’s reference list. It found the broken links and made helpful suggestions, but I wanted more. 1/N https://tractionsoftware.com/traction/permalink/Blog1964

federate.social · federate.social

0
0
0
orizuru @orizuru@mastodon.world
Public

> This book is dedicated, in respect and admiration, to the spirit that lives in the computer.
> I think that it's extraordinarily important that we in computer science keep fun in computing.

mitp-content-server.mit.edu/bo

When your boss and all the tech influencers push you to offload your creativity and understanding to an . It's hard to remember this sometimes.

Structure and Interpretation of Computer Programs

mitp-content-server.mit.edu

0
0
0
Giacomo Tesio @giacomo@snac.tesio.it
Public

@amszmidt@mastodon.socialAlfred M. Szmidt

The way are voiding classic has been experimentally demonstrated years ago, when was caught distributing well known code with a permissive license and wrong author attribution.

Also, technically that incident demonstrated how Copilot's model itself is a derivative work of copylefted works: if a lossy compression of copyrighted material is still subject to authors' , encoding such compression as arrays of floats that can be executed by virtual processors with a dedicated architecture (so called "inference" engines) does't change its nature of derivative work. Similarly violating copyrights of millions of authors at once doesn't free you of such rights.
That's basically why and friends are so scared by current lack of sustainable business models for their : they need money to keep Judges away.
But anyway I still have to find a single person that debate with technical competence and in good faith the derivative nature of LLMs from the text corpora compressed in their models.

As for the Hacking License not being a Free Software license, it's debatable after a careful read since the only thing you cannot do with the software is to prevent others from enjoining the same freedom it grants you.

Yet I've never claimed it is Free Software because, sadly, I'm forced to move beyond Free Software by its own limit.

OTOH I'm proud that it's not an license as I'll never submit it to corrupted .¹

As for , it's a term designed to confuse free software values with corporate propaganda while marginalizing hackers: its a leaking abstraction designed to fool developers and exploit their naive groupthinking. Having been fooled myself. Never again. 😉

____

¹ The way OSI tried to the , with an over complicated process that doubled 's lobbyists' votes to exclude training data from the requirements confirmed my opinion about them.

@doctormo@floss.socialMartin Owens :inkscape:

GitHub CopyALot distributes GPL code with wrong attribution and a permissive license

GitHub CopyALot distributes Quake III Arena code with wrong attribution and a permissive license, suggesting to a developer to include such code in his product. Video from Armin Ronacher's twitter ...

peertube.opencloud.lu · PeerTube Luxembourg

0
0
0
Giacomo Tesio @giacomo@snac.tesio.it
Public

Still, are voiding the (and ) reciprocity.

That's why years ago I wrote the https://encrypted.tesio.it/documents/HACK.txt

It was designed with automated corporate of in mind: it's goal is to balance and , and it share with those that accept it much more than permissions, while being a stromger and an explicit shrink-wrap contract.

Unfortunately, it's not compatible with GPL, because GPL is much weaker.

The fundamental issue of Free Software, the one that let people create the narrative and permessive licenses to exploit programmer ideals and , was that , as an American grown up during , was too fond of the freedom-vs-communism propaganda to understand how lack of rules means the rule of the rich.

The problem is not commercial use of free software but commercial exploitation of free labour, as @doctormo@floss.socialMartin Owens :inkscape: correctly stated.

The Hacking License does not prohibit commercial use, but requires recipient to share their own with the users of any derivative or dependant work they create as a contractual binding.

It's modelled after the research of about Commons governance and the ethics based on the value of .
0
0
0
0
0
0
Corv @corv@social.tchncs.de
Public

Giving an LLM agent root access to debug your Linux system is like handing someone a spoon to eat spaghetti—technically possible, catastrophically messy.

Shannot solves this with secure sandboxing for AI diagnostics. LLM agents can read logs, inspect configurations, and run diagnostic commands in a locked-down environment with zero write permissions.

They get the visibility they need to help you troubleshoot, without the access to accidentally destroy your system in the process.

github.com/corv89/shannot

Claude Desktop is running Shannot MCP to diagnose a Linux VM autonomously but securely thanks to sandboxing
0
0
0
Mariya Delano @mariyadelano@hachyderm.io
Public

LLMs are a fundamentally useless technology because their applications (supposedly) boil down to humans not having to think for themselves or do their own writing / drawing / filming.

But if you can do it on your own - why would you need a robot to do it? It’s, at best, a novelty.

That’s why this shit only resonates with executives and capital owners. “Get things done with fewer people and expenses” is at least an actual pitch. “Get things done faster for yourself” isn’t.

The individual angle really works for things you already were trying to avoid doing because you’re either disinterested or don’t have enough time to do things right.

“Avoid your work” as a value proposition doesn’t work when you’re dealing with intellectual labor rather than commodities. Not large scale, not long term.

Sorry for the rant, I saw some Notion ads on the subway and got irritated 😅

0
0
0
0
0
0
Trey Hunner 🐍 @treyhunner@mastodon.social
Public

If you avoid using an LLM, what are your primary reasons?

Mastodon only allows 4 options in polls and I know there are many more possible concerns.

If it's something else, like the spread of disinformation, mental health concerns, etc. reply with that.

0
0
0
Tattie @Tattie@eldritch.cafe
Public

I studied Artificial Intelligence for four years, and I am not touching LLM AIs with a ten-foot pole.

It's not really about the insane electricity demands, the water usage, tho that's a good reason. It's not even, if I'm honest, about the disastrous effect on the sum of all human art and knowledge.

It's because a) I've studied enough AI to know it's a trick, a sort of linguistic illusion, and b) I've studied enough everything else to understand that I'm not immune to such illusions.

0
0
0
José A. Alonso @Jose_A_Alonso@mathstodon.xyz
Public

Lean4Physics: Comprehensive reasoning framework for college-level Physics in Lean4. ~ Yuxin Li et als. arxiv.org/abs/2510.26094v1

arXiv logo

Lean4Physics: Comprehensive Reasoning Framework for College-level Physics in Lean4

We present **Lean4PHYS**, a comprehensive reasoning framework for college-level physics problems in Lean4. **Lean4PHYS** includes *LeanPhysBench*, a college-level benchmark for formal physics reasoning in Lean4, which contains 200 hand-crafted and peer-reviewed statements derived from university textbooks and physics competition problems. To establish a solid foundation for formal reasoning in physics, we also introduce *PhysLib*, a community-driven repository containing fundamental unit systems and theorems essential for formal physics reasoning. Based on the benchmark and Lean4 repository we composed in **Lean4PHYS**, we report baseline results using major expert Math Lean4 provers and state-of-the-art closed-source models, with the best performance of DeepSeek-Prover-V2-7B achieving only 16% and Claude-Sonnet-4 achieving 35%. We also conduct a detailed analysis showing that our *PhysLib* can achieve an average improvement of 11.75% in model performance. This demonstrates the challenging nature of our *LeanPhysBench* and the effectiveness of *PhysLib*. To the best of our knowledge, this is the first study to provide a physics benchmark in Lean4.

arxiv.org · arXiv.org

0
0
0
Devorppa @Devorppa@aus.social
Public

- I recently stopped my windows based computer backing every document and photo to the because fuck and their - I never asked for their fucking cloud storage in the first place. Or their LLM that they incorrectly call AI - it is not AI and can’t even add up.

Anyway, as part of this process, Microsoft then automatically deleted all my files - *including* the versions on my own computer. Fuck Microsoft.

As it happens, I back up all my computer files on an external HardDisk on the first day of every month. As a result, I was able to copy the files and directories back to my computer with little disruption.

If you are going to disconnect from your cloud storage - be warned in advance … back up all your files and directories on an external Hard Disk first.

0
0
0
José A. Alonso @Jose_A_Alonso@mathstodon.xyz
Public

APOLLO: Automated LLM and Lean collaboration for advanced formal reasoning. ~ Azim Ospanov, Farzan Farnia, Roozbeh Yousefzadeh. arxiv.org/abs/2505.05758

arXiv logo

APOLLO: Automated LLM and Lean Collaboration for Advanced Formal Reasoning

Formal reasoning and automated theorem proving constitute a challenging subfield of machine learning, in which machines are tasked with proving mathematical theorems using formal languages like Lean. A formal verification system can check whether a formal proof is correct or not almost instantaneously, but generating a completely correct formal proof with large language models (LLMs) remains a formidable task. The usual approach in the literature is to prompt the LLM many times (up to several thousands) until one of the generated proofs passes the verification system. In this work, we present APOLLO (Automated PrOof repair viaLLM and Lean cOllaboration), a modular, model-agnostic agentic framework that combines the strengths of the Lean compiler with an LLM's reasoning abilities to achieve better proof-generation results at a low token and sampling budgets. Apollo directs a fully automated process in which the LLM generates proofs for theorems, a set of agents analyze the proofs, fix the syntax errors, identify the mistakes in the proofs using Lean, isolate failing sub-lemmas, utilize automated solvers, and invoke an LLM on each remaining goal with a low top-K budget. The repaired sub-proofs are recombined and reverified, iterating up to a user-controlled maximum number of attempts. On the miniF2F benchmark, we establish a new state-of-the-art accuracy of 84.9% among sub 8B-parameter models (as of August 2025) while keeping the sampling budget below one hundred. Moreover, Apollo raises the state-of-the-art accuracy for Goedel-Prover-SFT to 65.6% while cutting sample complexity from 25,600 to a few hundred. General-purpose models (o3-mini, o4-mini) jump from 3-7% to over 40% accuracy. Our results demonstrate that targeted, compiler-guided repair of LLM outputs yields dramatic gains in both efficiency and correctness, suggesting a general paradigm for scalable automated theorem proving.

arxiv.org · arXiv.org

0
0
0
Iván Sánchez Ortega @IvanSanchez@mastodon.social
Public

I'm in a internal group for high-profile FOSS projects (due to @leaflet having a few kilo-stars), and the second most-wanted feature is "plz allow us to disable copilot reviews", with the most-wanted feature being "plz allow us to block issues/PRs made with copilot".

Meanwhile, there's a grand total of zero requests for "plz put copilot in more stuff".

This should be significative of the attitude of veteran coders towards creep.

A screenshot of a github discussion titled "can't disable copilot code reviews"A screenshot of a github discussion titled "Allow us to block Copilot-generated issues (and PRs) from our own repositories"
0
0
1
MidsouthMouth.OctaviaKeats @fkaOctaviaKeats@wandering.shop
Public

Tell me, if *I* regularly "hallucinated" on the job, how long before I was separated?
AI up here being treated like a rich white man's nephew in the workplace!

"Researchers say AI transcription tool used in hospitals invents things no one ever said" from AP News







apnews.com/article/ai-artifici

Assistant professor of information science Allison Koenecke, an author of a recent study that found hallucinations in a speech-to-text transcription tool, works in her office at Cornell University in Ithaca, N.Y., Friday, Feb. 2, 2024. The text preceded by "#Ground truth" shows what was actually said while the sentences preceded by ""text"" was how the transcription program interpreted the words. (AP Photo/Seth Wenig)

Researchers say an AI-powered transcription tool used in hospitals invents things no one ever said

Whisper is a popular transcription tool powered by artificial intelligence, but it has a major flaw. It makes things up that were never said.

apnews.com · AP News

0
0
0
José A. Alonso @Jose_A_Alonso@mathstodon.xyz
Public

Mathematics with large language models as provers and verifiers. ~ Hieu Le Duc, Leo Liberti. arxiv.org/abs/2510.12829

arXiv logo

Mathematics with large language models as provers and verifiers

During 2024 and 2025 the discussion about the theorem-proving capabilities of large language models started reporting interesting success stories, mostly to do with difficult exercises (such as problems from the International Mathematical Olympiad), but also with conjectures [Feldman & Karbasi, arXiv:2509.18383v1] formulated for the purpose of verifying whether the artificial intelligence could prove it. In this paper we report a theorem proving feat achieved by ChatGPT by using a protocol involving different prover and verifier instances of the gpt-5 model working collaboratively. To make sure that the produced proofs do not suffer from hallucinations, the final proof is formally verified by the lean proof assistant, and the conformance of premises and conclusion of the lean code is verified by a human. Our methodology is by no means complete or exact. It was nonetheless able to solve five out of six 2025 IMO problems, and close about a third of the sixty-six number theory conjectures in [Cohen, Journal of Integer Sequences, 2025].

arxiv.org · arXiv.org

0
0
0
0
0
0