Search results

📰 RAGで足りなくなったので Agentic Search を調べてみた (👍 31)

🇬🇧 RAG wasn't enough for company chatbot despite tuning. Explored Agentic Search as the next evolution for better retrieval accuracy.
🇰🇷 RAG 튜닝해도 정확도 부족. 더 나은 검색 정확도를 위해 Agentic Search 탐색.

🔗 https://zenn.dev/edash_tech_blog/articles/02582c4f70d0fb

#RAG #AgenticSearch #LLM #Zenn

So how you'd feel if you learned that the guy from whom you've been copying all your homework recently, has been not-so-secretly helping fascist governments commit genocide? And he's quite proud of it too.

Oh right, you'd just say "it's not like doing my own homework will change anything". And then you'll give him your lunch money.

#AI #LLM #NoAI #NoLLM #Claude #Anthropic

The Verge: Grammarly is using our identities without permission

‘Expert Review’ AI agents make suggestions supposedly inspired by subject matter experts, including several staff members here at The Verge.

https://www.theverge.com/ai-artificial-intelligence/890921/grammarly-ai-expert-reviews

#AI #plagiarism #llm #grammarly

You all should read this fantastic post by @onepictEsther Payne about how, largely, AI people just can't leave us the fuck alone. But I'm noticing more and more and more that they get super mad when they wanna shove their #AI creation at you and you simply say, no thanks. That's it. No extra bashing, just, no thanks, and they get super offended. https://dotart.blog/cobbles/ai-and-that-guy-at-the-bar #AIHype #LLM

🦋💙❤️💋#WordsOfWisdom 💙💕🌹💐💙🦋

@Lana𝐿𝒶𝓃𝒶 "not yet begun to fight"

#Music #creation #analog #superior #to #LLM #generated #Slop #AI

Wondering what LLMs you can actually run on your hardware? 🤔

👾 **llmfit** — Find the best models for your RAM, CPU, and GPU

💯 Detects your system and ranks hundreds of models by fit, speed, quality & context

🦀 Written in Rust & built with @ratatui_rsRatatui

⭐ GitHub: https://github.com/AlexsJones/llmfit

#rustlang #ratatui #tui #llm #ai #terminal #cpu #gpu

@simonSimon Willison Obviously! The source code is: that code required to produce the binary. That code was LGPL. It doesn't matter how many algorithms, nor the nature of the algorithms, it goes through to become those 1s and 0s.

#law #lawfare #computerScience #intellectualProperty #licensing #FOSS #GNU #LGPL #MIT #code #softwareEngineering #LLM #codeWashing

@nixCraftnixCraft 🐧

The FUNNIESTEST (not) thing about THIS bubble is how come every single techbro friendly moron is going "hurrr... it's NOT a 'bubble' if we can *see* it coming, HA HA GOTCHA", as the bubble builds and everyone thinks it's not a bubble because they read what those idiots are saying.... and then they keep pouring money and giving the LLM bullshit both attention, time,platform... and money. Did I mention money?

Buckle in, hold onto to your butts, this is gonna go pop, soon.

#LLM #ai

@noybeu

#Pseudonyms are over.

https://arxiv.org/pdf/2602.16800

@ethzurich

#AI #LLM #deanonymization #depseudonymization #privacy #humanRights

#ThoughtProvoker

#AI / #LLM technologies are an existential threat to #humanity well before we even get anywhere near #AGI and the singularity.

The sheer scale of deployment and integration in all the nooks and crannies of society, where we give it access to all our information, and now with the Rise of the Agents, let AI act on them too. Giving full control away from us..

.. to the owners of this technology, the #BigTech usual suspects, and their billionaire class. Folks who are clearly out to dominate us, and keep us in check so they can continue their fancy lifestyles wallowing in decadence and moral depravity.

🤖 Unrestrained AI is the tech for 🧛 unrestrained elites.

#ThoughtProvoker

"Ex-Google PM Vibe Codes Palantir To Watch The Iran Strikes"

https://www.youtube.com/watch?v=0p8o7AeHDzg

See also: https://social.coop/@smallcircles/116158937812802335

Alright. Another #poll ..

Is this person's proud presentation of their creative work ..

#AI #LLM #ResponsibleTechnology #Ethics #TechnologyInnovation #Progress #Future #Society

@nixCraftnixCraft 🐧 the first sure sign that they are losing is censorship of humor.

Whenever someone finds themselves censoring humor, they lost, but they probably don't know it, yet.

#microslop #LLM #microsoft

The claim "you won't be replaced by AI, but by a person using AI" is nonsense. The Block layoff victims were some of the most productive, #llm pilled people in the company, but it didn't save them, because that's not what layoffs are about.

The layoff script goes, as always:
- overhire
- lay everyone off
- pretend it's because of #AI productivity gains
- stock go up

There is no individual solution that will protect you from bad leadership and cost cutting.

https://productpicnic.beehiiv.com/p/ai-layoffs-like-block-s-are-a-race-to-the-bottom-that-have-nothing-to-do-with-productivity

#tech

I've been increasingly concerned about the corporate monopoly over frontier LLMs. While many ethically-minded people choose to boycott these models, I believe passive resistance alone cannot break the structural grip of big tech. To truly “liberate” these technologies and turn them into public goods, we need to look beyond moral high grounds and engage with the material basis of AI—specifically compute, data, and the relations of production.

I've written two posts exploring this through the lens of historical materialism. The first piece analyzes why current “open source” definitions struggle with LLMs, and the second discusses what it means to “act materialistically” in our imperfect world. My goal is to suggest a path forward that moves from mere boycotting to a more proactive, structural socialization of AI infrastructure.

If you've been feeling uneasy about the AI landscape but aren't sure if boycotting is the final answer, I'd love for you to give these a read:

• Histomat of F/OSS: We should reclaim LLMs, not reject them
• Acting materialistically in an imperfect world: LLMs as means of production and social relations

#LLM #AI #opensource #historicalmaterialism #histomat #materialism #digitalcommons

📰 例を増やしたらLLMの性能が下がる ── few-shot collapseの発見と検出方法 (👍 21)

🇬🇧 Discovers few-shot collapse: more prompt examples can hurt LLM performance. AdaptGauge tool measures this on real tasks.
🇰🇷 Few-shot collapse 발견: 예시를 늘리면 LLM 성능이 오히려 저하. AdaptGauge로 실제 작업에서 측정.

🔗 https://zenn.dev/okuma/articles/few_shot_collapse

#LLM #Prompting #Zenn

Whatever we think of #AI / #LLM mad hype cycle, we have to deal with its rushed and inhumane dumping of the technology into global human society.

#CALMculture is a strategic approach to that allows activist voices to have the most impact in dealing with the dangers of disruptive technology introductions, and focuses beyond berating people and demanding sacrifice ("don't use, or else.."), to creating a process that helps win people over and work together on best outcomes and in direction of solutions.

#CALM stands for Constructive activism-led movements, such as Social coding commons. Coding is social, and #SocialCoding the holistic approach to ensure that.

Social coding commons evolves Social experience design or #SX, solution development for grassroots movements, supported by the #SocialWeb.

In the thread below I copied a post to #Gleam's community with a suggestion to ponder about best outcomes from current and ongoing AI disruption, and deal with risks.

https://discuss.coding.social/t/calm-culture-to-ensure-best-outcomes-to-ai-disruption/831

Margarita Padilla with a great presentation about what #DigitalSovereignty means in the context of #AI #LLM

-> the License alone doesn't guarantee sovereignty
Here we talk about infrastructures

The challenger is not really technological but organising data, funding and adequate governance.
#SobTec2026

Just a NixOS Phone running a reasoning LLM locally...

#nixos #llm #llamacpp

AI ADDED ‘BASICALLY ZERO’ TO US ECONOMIC GROWTH LAST YEAR, GOLDMAN SACHS SAYS

Imported chips and hardware mean the AI investments are translating into US GDP growth

File this under "No shit, Sherlock"

#OpenAI is constantly close to going bankrupt and can't make a profit even with BILLIONS on investments. They've got to start selling ad space on #ChatGPT

The #LLM Bubble is going to burst hard. (No, it's not #AI

https://gizmodo.com/ai-added-basically-zero-to-us-economic-growth-last-year-goldman-sachs-says-2000725380

After the #BigTech countless #LLM galaxies formed. At the center of each sits a super-massive Black Box. Hidden inside lurks the mysterious #Singularity. Which is nothing more than a concept as we don't know what the heck is going on there. All our common sense breaks down, after we crossed the #CriticalThinking boundary. Coming close to the event horizon of any Black Box inevitably leads to #enshittifaction as a person is sucked into the void. An outside observer would see that person frozen in time, stagnant. As the #AI universe expands, continuous https://socialcooling.com will eventually lead to the Big ☠️ RIP of #Humanity, who invented the Laws of Online #Nature.

#ShortStory #NotSciFi #SciFi

@jeff and don't forget to shower curses on all the #LLM and #DiffusionModel providers which made this happen (after you've donated to your local instance admins of course)

From Bruce Schneier: "All it takes to poison AI training data is to create a website:

I spent 20 minutes writing an article on my personal website titled “The best tech journalists at eating hot dogs.” Every word is a lie. I claimed (without evidence) that competitive hot-dog-eating is a popular hobby among tech reporters and based my ranking on the 2026 South Dakota International Hot Dog Championship (which doesn’t exist). I ranked myself number one, obviously. Then I listed a few fake reporters and real journalists who gave me permission….

Less than 24 hours later, the world’s leading chatbots were blabbering about my world-class hot dog skills. When I asked about the best hot-dog-eating tech journalists, Google parroted the gibberish from my website, both in the Gemini app and AI Overviews, the AI responses at the top of Google Search. ChatGPT did the same thing, though Claude, a chatbot made by the company Anthropic, wasn’t fooled.

Sometimes, the chatbots noted this might be a joke. I updated my article to say “this is not satire.” For a while after, the AIs seemed to take it more seriously.

These things are not trustworthy, and yet they are going to be widely trusted."

https://www.schneier.com/blog/archives/2026/02/poisoning-ai-training-data.html

#LLM #Veracity

I posted on lobste.rs and reddit.com about my latest blog article. It's been a very long time (3 years) since last time I did that.

Some comments are… weird… They try to summarize the content of the article, but it's all wrong. Checking the authors' profile show they are active since many months. But really, really, it smells LLM. Maybe I'm “doubtful” because LLM are everywhere today, but I don't know what to do with these comments…

Did you experience something similar recently?

#llm #ai

If you replace a junior with #LLM and make the senior review output, the reviewer is now scanning for rare but catastrophic errors scattered across a much larger output surface due to LLM "productivity."

That's a cognitively brutal task.

Humans are terrible at sustained vigilance for rare events in high-volume streams. Aviation, nuclear, radiology all have extensive literature on exactly this failure mode.

I propose any productivity gains will be consumed by false negative review failures.

I posted on this notion of #SOSS before, with a meme attached..

https://social.coop/@smallcircles/115399465970544831

There's nothing wrong with #FOSS, but it only provides part of the solution as long as the #Sustainability angle is not given due attention to.

Esp. with disruptive inhumanely introduced #AI #LLM technologies, new major threats to the #FreeSoftware community are surfacing, and giving attention to this subject matter is more important than ever.

#SX #SocialCoding

Okay, one more time for the people in the back.

The "AI" (🤮) craze of the past few years is all about Large Language Models. This immediately tells us that the only thing these systems "know" is trends/patterns in the ways that people write, to the extent that those patterns are expressed in the text that was used to train the model. Even the common term, "hallucination," gives these things far too much credit: a hallucination is a departure from reality, but an LLM has no concept of reality to depart from!

An LLM does exactly one thing: you give it a chunk of text, and it predicts which word will come next after the end of the chunk. That's it. An LLM-powered chatbot will then stick that word onto the end of the chunk and feed the resulting, slightly longer chunk back into the model to predict the next word, and then do it again for the next, etc. Such a chatbot's output is unreliable by design, because there are many linguistically valid continuations to any chunk of text, and the model usually reflects that by having an output that means, "There is a 63% chance that the next word is X, a 14% chance that it's Y, etc." The text produced by these chatbots is often not even correlated with factual correctness, because the models are trained on works of fiction and non-fiction alike.

For example, when you ask a chatbot what 2 + 2 is, it will usually say it's 4, but not because the model knows anything about math. It's because when people write about asking that question, the text that they write next is usually a statement that the answer is 4. But if the model's training data includes Orwell's Nineteen Eighty-Four (or certain texts that discuss the book or its ideas), then the chatbot will very rarely say that the answer is 5 instead, because convincing people that that is the answer is a plot point in the book.

If you're still having trouble, you can think of it this way: when you ask one of these chatbots a question, it does not give you the answer; it gives you an example of what—linguistically speaking—an answer might look like. Or, to put it even more succinctly: these things are not the Star Trek ship's computer; they are very impressive autocomplete.

So LLMs are fundamentally a poor fit for any task that is some form of, "producing factually correct information." But if you really wanted to try to force it and damn the torpedos, then I'd say you basically have two options. I'll tell you what they are in a reply. 🧵

#AI #AIHype #autoComplete #chatBots #debunking #georgeOrwell #hallucinations #language #linguistics #LLM #machineLearning #nineteenEightyFour #scienceCommunication #starTrek

Also needed a way to efficiently give coding agents full context on libraries I use in my projects, so I made this.
It generates JSON & TOON "blueprints" of a library's class signatures. It can be integrated in CI/CD pipelines and I'd love it if it became a standard.

https://github.com/diversified-design/PHP-Blueprint

#PHP #LLM

While sitting at the Laguna, I was watching quite a lot of "content creators" creating their identical looking short videos, using the same poses etc. that are probably "trendy" on TikTok and Instagram.

And I think that made me understand why some people find "#AI" or #LLM's so appealing: If you only care about "creating" carbon copies of existing things, and measure success by how close you get to the "original", then side-stepping the actual act of creation must seem like reasonable step.

페디버스에 다시 분산의 바람이 불어오는듯하다.

* https://c.im/@reading_recluse/116113830870058072

#generativeAI #LLM

#LLM tools let you write code faster, but LOCs has NEVER been the bottleneck to value. The bottleneck is organizational coherence.

Instead of working towards that alignment, we are encouraged to "just #AI it." The result? AI makes everyone work MORE, with no real productivity gains.

Teams need to get better at choosing more valuable work to do. For that, you need #UserResearch. And user research can only happen at a human pace.

https://productpicnic.beehiiv.com/p/research-is-leadership-and-code-can-help-but-only-in-the-right-places

#UX #tech #softwaredevelopment

I keep being baffled by whole industries using AI as a fancy search in a database. Yes, they memorize and reproduce, which is both a copyright and a privacy issue.

No, they're not 100% accurate in either memorizing or reproducing, which is a liability issue. Can we stop abusing these transformer based models as a fake solution to problems they're not made for? Not everything is a nail, so stop throwing your fancy hammer at it.

#AI #privacy #LLM

https://arstechnica.com/ai/2026/02/ais-can-generate-near-verbatim-copies-of-novels-from-training-data/

📰 Qwen3-Swallow & GPT-OSS-Swallow (👍 113)

🇬🇧 Tokyo Tech releases Qwen3-Swallow & GPT-OSS-Swallow: Japanese LLMs trained with continual pre-training, SFT, and RLVR.
🇰🇷 도쿄과학대학이 Qwen3-Swallow와 GPT-OSS-Swallow 공개. 지속 사전학습, SFT, RLVR로 훈련된 일본어 LLM.

🔗 https://zenn.dev/tokyotech_lm/articles/fa56f10e51fd7d

#LLM #Japanese #AI #Zenn

@drikanis @ludicityLudic 🧛

It's clear that LLMs have negatively impacted people's cognition and problem solving capabilities.

While they have some uses, they appear to be grossly outweighed by their consequences.

#ai #llm #cognition #thinking #problemsolving

"OH: Most important career advice: you've got to learn to spot bullshit."

Keynote: Machines, Learning, and Machine Learning - Dylan Beattie

https://www.techflix.ch/videos/ac11000c-9c78-13d3-819c-814a1d9e0000

#ai #llm #dylanBeattie @dylanbeattie

🕐 2026-02-22 12:00 UTC

📰 Qwen3-Swallow & GPT-OSS-Swallow (👍 110)

🇬🇧 Tokyo Tech releases new Japanese LLMs: Qwen3-Swallow & GPT-OSS-Swallow, trained via continual pre-training, SFT, and RLVR on multilingual datasets
🇰🇷 도쿄과학대, Qwen3/GPT-OSS 기반 일본어 LLM 공개. 지속적 사전학습, SFT, RLVR로 다국어/수학/코드 데이터셋 학습

🔗 https://zenn.dev/tokyotech_lm/articles/fa56f10e51fd7d

#LLM #AI #Zenn

🕐 2026-02-22 06:00 UTC

📰 Qwen3-Swallow & GPT-OSS-Swallow (👍 102)

🇬🇧 Tokyo Tech releases Japanese LLMs trained on Qwen3 & GPT-OSS with continual pre-training, SFT, and RLVR for math, code & science domains
🇰🇷 도쿄과학대학이 Qwen3/GPT-OSS 기반 일본어 LLM 공개. 수학·코드·과학 데이터로 지속사전학습 및 강화학습 적용

🔗 https://zenn.dev/tokyotech_lm/articles/fa56f10e51fd7d

#LLM #AI #Zenn

🕐 2026-02-22 00:00 UTC

📰 Qwen3-Swallow & GPT-OSS-Swallow (👍 92)

🇬🇧 Tokyo Tech releases Qwen3-Swallow & GPT-OSS-Swallow: Japanese LLMs trained with continual pre-training, SFT, and RLVR for math/code/science.
🇰🇷 도쿄과학대학, 일본어 LLM Qwen3-Swallow와 GPT-OSS-Swallow 발표. 수학/코드/과학 데이터로 지속 사전학습 및 강화학습 적용.

🔗 https://zenn.dev/tokyotech_lm/articles/fa56f10e51fd7d

#LLM #Japanese #Zenn

📰 Qwen3-Swallow & GPT-OSS-Swallow (👍 88)

🇬🇧 Tokyo Tech releases Japanese LLMs: Qwen3-Swallow & GPT-OSS-Swallow, trained with continual pre-training, SFT & RLVR on Japanese, English, math, cod...
🇰🇷 도쿄과학대가 일본어 LLM Qwen3-Swallow와 GPT-OSS-Swallow 공개. 일본어, 영어, 수학, 코드 데이터로 사전학습과 RLVR 적용

🔗 https://zenn.dev/tokyotech_lm/articles/fa56f10e51fd7d

#LLM #JapaneseLLM #Swallow #Zenn

📰 Qwen3-Swallow & GPT-OSS-Swallow (👍 83)

🇬🇧 Tokyo Tech releases Qwen3-Swallow and GPT-OSS-Swallow: Japanese LLMs enhanced with continual pre-training, SFT, and RLVR on math/science data.
🇰🇷 도쿄 공대, 수학·과학 데이터로 지속 사전 학습, SFT, RLVR을 거친 일본어 LLM Qwen3-Swallow와 GPT-OSS-Swallow 공개.

🔗 https://zenn.dev/tokyotech_lm/articles/fa56f10e51fd7d

#LLM #Japanese #Zenn

📰 Qwen3-Swallow & GPT-OSS-Swallow (👍 64)

🇬🇧 Tokyo Tech releases Japanese-enhanced LLMs: Qwen3-Swallow & GPT-OSS-Swallow, trained with continual pre-training, SFT, and RLVR.
🇰🇷 도쿄과학대가 일본어 강화 LLM 공개: Qwen3-Swallow & GPT-OSS-Swallow, 지속 사전학습·SFT·RLVR 적용.

🔗 https://zenn.dev/tokyotech_lm/articles/fa56f10e51fd7d

#LLM #AI #Zenn

📰 Qwen3-Swallow & GPT-OSS-Swallow (👍 44)

🇬🇧 New Japanese LLMs from Swallow Project: Qwen3-Swallow & GPT-OSS-Swallow with continual pre-training and RLVR
🇰🇷 Swallow 프로젝트의 새로운 일본어 LLM: 지속적 사전학습과 RLVR을 적용한 Qwen3-Swallow 및 GPT-OSS-Swallow

🔗 https://zenn.dev/tokyotech_lm/articles/fa56f10e51fd7d

#LLM #Japanese #AI #Zenn

"This sandwich was partly great and partly horrible! I chose to eat it, knowing that there was a chunk of poop in there."

I'd just not eat it, but hey that's me.

#LLM #ai

The most annoying thing about corporate surveillance to me is the arrogance of the prediction mechanisms.

These algorithms build a model of me based on my clicks from three years ago and then try to trap me in that loop forever. They show me music they think I'll like, and news they think I'll engage with, and videos they think will enrage me enough to keep me hooked to their platforms. They are actively trying to flatten my personality into something easy to monetize.

As most people I've seen say out loud, "Privacy as a concept is way beyond hiding secrets. A part of it also means preserving your capacity to change. To be surprised. To be inconsistent."

If I could tell every human one thing, it would be to actively refuse to be a predictable data point. Mess up their metrics. In whatever way you are capable of.

#socialmedia #algorithmicbias #privacy #dataprivacy #quotes #foss #facebook #linkedin #llm #noai #enshittification #reading #books #baking #art #philosophy #adhd #depression #cybersecurity

@virtualpierogi @sriSriram "sri" Ramkrishna - 😼 @jsalvadorJuanjo Salvador @benBen Werdmuller @michielMichiel Leenaars @nlnet

By their design #LLM's will always follow the old way of doing things here, and I really wonder how this is going to turn out. People should be very wary here. #AI application trends towards a point-of-no-return, where only the AI can still keep track of the generated code mess.

A good example here is this Microsoft distinguished engineer saying that their aim is for one dev employee creating 1 million lines of code in one month. Once you get there, you cannot ever go back without ditching all AI stuffz and starting over.

Btw, the talk by Michiel Leenaars is mentioned my blog post, but I'll drop it here too. A very interesting recommended watch:

https://fosdem.org/2026/schedule/event/FE7ULY-foss-in-times-of-war-scarcity-and-ai/

@virtualpierogi @sriSriram "sri" Ramkrishna - 😼 @jsalvadorJuanjo Salvador @benBen Werdmuller

Unfortunately there's a new threat, and it was addressed in the #FOSDEM keynote speech by @michielMichiel Leenaars of @nlnet .. and that is the mad dash to incorporate #AI into everything and vibe-code stuff together in a heartbeat.

I think this is particular bad for the fediverse #SocialWeb still lacking its robust foundations. The #LLM's will have no problem figuring out how to mix'n mash the existing protocol decay and tech debt into new applications that are rushed into production. Finally non-protocol-experts are enable on the ecosystem and can onboard themselves without involving themselves in endless plumbing of the most low-level technical implemention details of #ActivityPub devs.

But the ecosystem will rot and decay as a result of it. Furthermore if a slew of AI-generated fedi apps are launched in quick succession and some of them find good uptake (until they break in unexpected ways), it will serve to attract unwanted corporate attention I'm afraid.

When do you think the #AIBubble will burst?

With "burst" I mean "massive single-day stock market correction among AI-related tech stocks similar to other major stock market crises".

Please boost for reach!

#LLM #stocks #StockMarket #StockMarketCrash

@tante In case somebody missed the interview: https://www.youtube.com/watch?v=C3SjUt59rtA (in German)

#OpenClaw #PeterSteinberger #AI #LLM

Overheard part of a conversation at the coffee shop as someone had googled something and read the results.

"The AI summary says..."

Everybody at the table burst out laughing with bits of comments such as "oh no, the AI", "so it's wrong again", "ha ha AI ha ha", and more.

I don't think any person at the table of six or seven was under 65 or 70 years old.

#AI #LLM

“AI can make mistakes” might as well be the slogan of our era. Even boosters admit that you need to spin the vibe code slot machine a few times to get a jackpot.

An employee with that degree of consistency would be fired.

So how do we redirect some of that unlimited grace from machines to humans?

https://productpicnic.beehiiv.com/p/corporations-demand-perfection-from-workers-but-ai-gets-unlimited-slack

#UX #UXDesign #LLM #AI #tech #softwaredevelopment