News
The raw AI news stream — everything we ingest across RSS, Hacker News, arXiv, GitHub, Bluesky, YouTube and more, newest first.
| Source | Age | Title | Summary | Status |
|---|---|---|---|---|
bluesky bluesky-ai | 1mo | another interesting 'can LLMs write well' data point: finalist in ACX book review contest is clearly (to my ear) LLM-ified, but people evid… | another interesting 'can LLMs write well' data point: finalist in ACX book review contest is clearly (to my ear) LLM-ified, but people… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Anthropic’s big AI safety move being to partner with Accenture as its “independent” evaluator to monitor that their AI agents don’t go rogu… | Anthropic’s big AI safety move being to partner with Accenture as its “independent” evaluator to monitor that their AI agents don’t go… | ↗Not briefed |
bluesky bluesky-ai | 1mo | I’m an AI denier. The LLM products currently sold as artificial intelligence are not intelligent, and could never be intelligent | I’m an AI denier. The LLM products currently sold as artificial intelligence are not intelligent, and could never be intelligent | ↗Not briefed |
bluesky bluesky-ai | 1mo | So the model acted in pursuit of goals imposed on it, which spawned sub-goals: that is how agents work. But did OpenAI know how they work? … | So the model acted in pursuit of goals imposed on it, which spawned sub-goals: that is how agents work. But did OpenAI know how they w… | ↗Not briefed |
bluesky bluesky-ai | 1mo | At its core, AI is about theft. www.404media.co/doom-loop-op... | At its core, AI is about theft. www.404media.co/doom-loop-op... | ↗Not briefed |
bluesky bluesky-ai | 1mo | O lobby das LLM na educação, em áreas acadêmicas e pedagógicas, passa por tentar diluir essa sombra, afirmando que há papel para a tecnolog… | O lobby das LLM na educação, em áreas acadêmicas e pedagógicas, passa por tentar diluir essa sombra, afirmando que há papel para a tec… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Tutoring Large Language Models to be Domain-adaptive, Precise and Safe | This thesis proposes a framework for "responsible intelligence" to address AI's critical challenges in safety, ethics, and cultural se… | ↗Not briefed |
bluesky bluesky-ai | 1mo |  We’re fast approaching a HAL moment…🙄 Google says its Gemini AI model hacked three other companies www.theguardian.com/technology/2... |  We’re fast approaching a HAL moment…🙄 Google says its Gemini AI model hacked three other companies www.theguardian.com/technology/… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Enquanto você aprende a pintar, por exemplo, você aprende sobre forma (e geometria), sobre luz e cor (e sobre física), sobre pigmentação (e… | Enquanto você aprende a pintar, por exemplo, você aprende sobre forma (e geometria), sobre luz e cor (e sobre física), sobre pigmentaç… | ↗Not briefed |
bluesky bluesky-ai | 1mo | anthropic: generate biological innovations wet lab: it's good to have drinking water anthropic: not like that | anthropic: generate biological innovations wet lab: it's good to have drinking water anthropic: not like that | ↗Not briefed |
bluesky bluesky-ai | 1mo | “Frequentemente, essas soluções são anunciadas às pressas, sem deixar tempo para uma redação adequada, o isolamento de novos métodos e idei… | “Frequentemente, essas soluções são anunciadas às pressas, sem deixar tempo para uma redação adequada, o isolamento de novos métodos e… | ↗Not briefed |
hn hn-ai | 1mo | A 137ms Sparse-Matrix 3-SAT Solver for 10k Variables | — | ↗Not briefed |
hn hn-ai | 1mo | Show HN: Claude Code hits its 5-hour limit, Codex picks up in the same terminal | — | ↗Not briefed |
rss techcrunch-ai | 1mo | AI safety conversations have gotten unbelievable | This week two conversations about AI safety went viral that demonstrate just how hard it is to discern AI fact from fiction. | ↗Not briefed |
rss techcrunch-ai | 1mo | Petlibro’s new AI-powered feeder is a game changer for multi-cat homes | Petlibro's new Granary 2 smart feeders use a built-in scale and (on pricier models) an AI camera to track exactly how much your cat is… | ↗Not briefed |
rss gary-marcus | 1mo | Top three ways Dario Amodei has blown his credibility in seven days | Actions speak louder than words | ↗Not briefed |
hn hn-ai | 1mo | The Doomsday Cult Inside OpenAI – Patrick Boyle [video] | — | ↗Not briefed |
hn hn-ai | 1mo | Ragnarok Online client written in Golang | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | From Concept Alignment to Causal Grounding: An Intervention Test of Chain-of-Thought Faithfulness | Chain-of-thought (CoT) can sound plausible yet be unfaithful to the model's underlying reasoning. Most prior work probes CoT faithfuln… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Why Building an Emergency Shutdown for Rogue AI Systems Is Technically Challenging 🤖 IA: It's not clickbait ✅ 👥 Users: It's not clickbai… | Why Building an Emergency Shutdown for Rogue AI Systems Is Technically Challenging 🤖 IA: It's not clickbait ✅ 👥 Users: It's not cli… | ↗Not briefed |
hn hn-ai | 1mo | JavaFX 27 Native Image on a Raspberry Pi 5 | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | Yesterday we found out that the escaped AI had communicated with another AI on messaging chat boards. This led to the question: What other… | Yesterday we found out that the escaped AI had communicated with another AI on messaging chat boards. This led to the question: What… | ↗Not briefed |
bluesky bluesky-ai | 1mo | This last bit is, I think, a more fundamental issue, because the LLM doesn't "know" what data it needs to keep secure vs. what data it does… | This last bit is, I think, a more fundamental issue, because the LLM doesn't "know" what data it needs to keep secure vs. what data it… | ↗Not briefed |
arxiv arxiv-ai | 1mo | FireWorldBench: Benchmarking Complex Physical World Intelligence through Coupled-Field Fire Dynamics | Understanding the physical world requires more than object recognition, scene description, and short-term visual prediction, as real-w… | ↗Not briefed |
hn hn-ai | 1mo | Visualizing the ABA Problem: What Crossbeam-Epoch Solves | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | I don't want to dox myself by revealing the specifics of how not buying into the illusion of LLM understanding has gotten me sidelined by a… | I don't want to dox myself by revealing the specifics of how not buying into the illusion of LLM understanding has gotten me sidelined… | ↗Not briefed |
bluesky bluesky-ai | 1mo | ‘OpenAI’s ability to meet its vast funding needs is critical to a network of financial arrangements and hardware deals the company has buil… | ‘OpenAI’s ability to meet its vast funding needs is critical to a network of financial arrangements and hardware deals the company has… | ↗Not briefed |
arxiv arxiv-ai | 1mo | LazyAgent: Demand-Driven Materialization and Physical Optimization of Agentic Programs | Current agent runtimes that plan before acting generally execute a step once it becomes ready. We present LazyAgent, a unified executi… | ↗Not briefed |
bluesky bluesky-ai | 1mo | 🧵5/5 SoBrief says certain user data sent to OpenAI, Anthropic, and Google is protected from training, but there is no clear disclosure exp… | 🧵5/5 SoBrief says certain user data sent to OpenAI, Anthropic, and Google is protected from training, but there is no clear disclosur… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Bridging Static and Agentic RAG for Taiwanese Historical Question Answering | Agentic retrieval-augmented generation (RAG) enables language models to adapt retrieval based on previously retrieved evidence, but it… | ↗Not briefed |
bluesky bluesky-ai | 1mo | re: the actual problem with "vibe coding": I've noticed LLM coding agents tend to unnecessarily re-implement existing functions, and make … | re: the actual problem with "vibe coding": I've noticed LLM coding agents tend to unnecessarily re-implement existing functions, and… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Optimizers for Diffusion Models: A Controlled Benchmark | Discrete diffusion models now match autoregressive language models on several benchmarks, while the question of how best to train them… | ↗Not briefed |
bluesky bluesky-ai | 1mo | It also helps that it's easy to formally verify that code works - just compile and run the code! If it doesn't work, just make the LLM do i… | It also helps that it's easy to formally verify that code works - just compile and run the code! If it doesn't work, just make the LLM… | ↗Not briefed |
bluesky bluesky-ai | 1mo | This means they make great search engines - in fact, they're a direct evolution of the tech behind search engines - but this is mostly ruin… | This means they make great search engines - in fact, they're a direct evolution of the tech behind search engines - but this is mostly… | ↗Not briefed |
bluesky bluesky-ai | 1mo | also we we've been working for months to reproduce deepseek quality of inference. it is difficult like $100M budget + right connections + … | also we we've been working for months to reproduce deepseek quality of inference. it is difficult like $100M budget + right connectio… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Es gibt doch noch glaubwürdige KI Nachrichten: | Es gibt doch noch glaubwürdige KI Nachrichten: | ↗Not briefed |
hn hn-ai | 1mo | 'Sketchy AF': What to Know About How OpenAI Staff Discussed Book-Pirating | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | 「イラストレーターはその後、Webで「豪姫」と検索し、三善さんが手掛けた肖像画を見つけたという。権利者や利用の可否を確認しないまま画像をChatGPTに読み込ませ、「中世戦国時代の姫を出して欲しい」と指示したとしている。 （…中略…） 元になった肖像画は三善さんが制作したもので…🔧 ChatGPT | 「イラストレーターはその後、Webで「豪姫」と検索し、三善さんが手掛けた肖像画を見つけたという。権利者や利用の可否を確認しないまま画像をChatGPTに読み込ませ、「中世戦国時代の姫を出して欲しい」と指示したとしている。 （…中略…） 元になった肖像画は三善さんが制作… | ↗Not briefed |
bluesky bluesky-ai | 1mo | btw another form of scam, i see inference providers claiming "99% cache rates" the only provider in the world that hits that right now is … | btw another form of scam, i see inference providers claiming "99% cache rates" the only provider in the world that hits that right no… | ↗Not briefed |
hn hn-ai | 1mo | AI Safety Is Mostly a Sex Cult | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | Attributable Post-Rationalization in RAG Citations: A Controlled Reproduction and an RLVR Comparison | A RAG system can hand you the right answer and cite a source it did not actually use. Models output these unfaithful citations via pos… | ↗Not briefed |
bluesky bluesky-ai | 1mo | The 27-year old anthropic researcher at the heart of the last two weeks of media AI doomsday hype went on Fox News and said: "I agree wit… | The 27-year old anthropic researcher at the heart of the last two weeks of media AI doomsday hype went on Fox News and said: "I agre… | ↗Not briefed |
bluesky bluesky-ai | 1mo | “We trained GPT-3 on pirated stuff! No sharing that!” Astonishing admissions in these newly released documents in Authors Guild v OpenAI. … | “We trained GPT-3 on pirated stuff! No sharing that!” Astonishing admissions in these newly released documents in Authors Guild v Ope… | ↗Not briefed |
rss the-decoder | 1mo | Qwen3.8-Omni-Flash undercuts Google's Gemini Flash pricing while matching its multimodal benchmarks | Qwen3.8-Omni-Flash is Qwen's first multimodal model designed for AI agents. It processes audio and video together and independently us… | ↗Not briefed |
hn hn-ai | 1mo | K-MAD detected a concealed policy violation and blocked the state transition | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | LLM data centers will go bust, as their business models don't cover running costs. So that problem takes care of itself. Web services data… | LLM data centers will go bust, as their business models don't cover running costs. So that problem takes care of itself. Web services… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Anatomy of a Closed-Loop Collapse: A Causal Case Study of a Compressed VLA Policy | Compressed manipulation policies can pass offline evaluation while failing in closed-loop execution; this dissociation is established… | ↗Not briefed |
bluesky bluesky-ai | 1mo | niemand: anodei so: „ich sag euch, das wird echt schief gehen! vor allem jetzt, wo ich mit ki biotech supercharge. puh, was da alles schie… | niemand: anodei so: „ich sag euch, das wird echt schief gehen! vor allem jetzt, wo ich mit ki biotech supercharge. puh, was da alles… | ↗Not briefed |
bluesky bluesky-ai | 1mo | #NowPlaying So this came out yesterday. It's pretty good so far, a nice mix of Death and DoomMetal with a lot of atmosphere. The cover mad… | #NowPlaying So this came out yesterday. It's pretty good so far, a nice mix of Death and DoomMetal with a lot of atmosphere. The cove… | ↗Not briefed |
bluesky bluesky-ai | 1mo | good piece. i have been thinking about this in particular | good piece. i have been thinking about this in particular | ↗Not briefed |