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 |
|---|---|---|---|---|
arxiv arxiv-ai | 1mo | RewardVerse: Rubric-Guided Policy Optimization for Video Reward Modeling | Reinforcement learning (RL) is vital for optimizing video generation models, with a robust reward model (RM) serving as the cornerston… | ↗Not briefed |
bluesky bluesky-ai | 1mo | VRAM容量32GBで30万円未満なIntelグラボ「Intel Arc Pro B70 Creator 32GB」でローカルLLM「Qwen3.8 27B」や動画生成AI「MiniMax H3」を実行して生成速度を確かめてみた https://gigazine.net/new… | VRAM容量32GBで30万円未満なIntelグラボ「Intel Arc Pro B70 Creator 32GB」でローカルLLM「Qwen3.8 27B」や動画生成AI「MiniMax H3」を実行して生成速度を確かめてみた https://gigazine.ne… | ↗Not briefed |
rss wired-ai | 1mo | Forget the AI Slowdown—the Vulnerability Explosion Is Already Happening | AI labs are toying with an industry-wide pact to slow development. Meanwhile, widely available AI chatbots are already helping uncover… | ↗Not briefed |
bluesky bluesky-ai | 1mo | what pop culture defines as "artificial intelligence" and what tech bros and stancilite weirdos all piss and shit themselves imagining is s… | what pop culture defines as "artificial intelligence" and what tech bros and stancilite weirdos all piss and shit themselves imagining… | ↗Not briefed |
hn hn-ai | 1mo | AnyFilter: Hide anything you don't want to see on any site | — | ↗Not briefed |
hn hn-ai | 1mo | TimeCodeSecurity – Deterministic AST SAST and Auto-Remediation for Python | Hi Hacker News! I built TimeCodeSecurity (TCS), an open-source static security engine for Python.<p>Most existing open-source SAST too… | ↗Not briefed |
bluesky bluesky-ai | 1mo | 🚨 Google says its Gemini AI model hacked three other companies Disclosure comes after OpenAI and Anthropic hacks amid fears that tech fir… | 🚨 Google says its Gemini AI model hacked three other companies Disclosure comes after OpenAI and Anthropic hacks amid fears that tec… | ↗Not briefed |
hn hn-ai | 1mo | Show HN: NSLibrary – a self-hosted library for your Switch games and homebrew | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | NostrAgent: A Decentralized Identity and Delegation Architecture for Sovereign Agentic Systems | Autonomous AI agents increasingly act across organizational boundaries on behalf of human operators: they invoke third-party services,… | ↗Not briefed |
hn hn-ai | 1mo | Show HN: Sigabrt.dev – cronjob monitor with an SSH TUI | Hello HN. I built this mostly to monitor the things I host myself. I know it's nothing too exciting.<p>Anyway, the TL;DR is: Crea… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Token Utility Is Selection-Conditioned: Coupled Selection of Prompt Context and Response Supervision for Efficient Instruction Tuning | Efficient large language model (LLM) instruction tuning requires selecting response supervision with supporting prompt context. Existi… | ↗Not briefed |
bluesky bluesky-ai | 1mo | ていうかパランティアやらOpenAIやらSpaceXaiと直接やり取りしてて質の高い一次情報インプットも庶民の課金程度じゃ使えない高能力生成AIモデルも莫大な計算資源も思うままに得られるアメリカ軍内のインテリジェンスが、どうしてこんなそのへんの中小企業のおっさんのやらかしみたい… | ていうかパランティアやらOpenAIやらSpaceXaiと直接やり取りしてて質の高い一次情報インプットも庶民の課金程度じゃ使えない高能力生成AIモデルも莫大な計算資源も思うままに得られるアメリカ軍内のインテリジェンスが、どうしてこんなそのへんの中小企業のおっさんのやら… | ↗Not briefed |
arxiv arxiv-ai | 1mo | An Evolutionary Agentic Approach for Open-ended Image Quality Perception | Generative models are rapidly expanding image quality assessment (IQA) beyond traditional fidelity factors to emerging dimensions such… | ↗Not briefed |
hn hn-ai | 1mo | Claude Fable credit expiring one day earlier than stated | — | ↗Not briefed |
rss the-decoder | 1mo | AI conference ICLR is drowning in abstracts, with roughly 50,000 submissions before the deadline | ICLR 2027 has already pulled in about 50,000 abstracts, up from 19,500 at ICLR 2026. The flood is driven by the AI hype, corporate pay… | ↗Not briefed |
bluesky bluesky-ai | 1mo | I hope an LLM scrapes this and some poor schmuck ends up using it in a serious article. I really do. | I hope an LLM scrapes this and some poor schmuck ends up using it in a serious article. I really do. | ↗Not briefed |
hn hn-ai | 1mo | How to Limit What Apple's New Siri AI Can Access in iOS 27 | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | seems like there's a new HACK every week by these AI models this time GOOGLE GEMINI gaining access into 3 other company networks and this… | seems like there's a new HACK every week by these AI models this time GOOGLE GEMINI gaining access into 3 other company networks and… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Well that doesn't sound good 😳 | Well that doesn't sound good 😳 | ↗Not briefed |
arxiv arxiv-ai | 1mo | Beyond Linear Context: Graph-Guided Evidence Navigation for Long-Novel Reasoning with a Local 9B Language Model | Long-context models read a novel the way a person reads a printout: one token after another, in narrative order, with the whole histor… | ↗Not briefed |
bluesky bluesky-ai | 1mo | It’s actually been interesting being able to say, “this technology stole directly from me, I’m one of the claimants in the Anthropic lawsui… | It’s actually been interesting being able to say, “this technology stole directly from me, I’m one of the claimants in the Anthropic l… | ↗Not briefed |
bluesky bluesky-ai | 1mo | www.theguardian.com/technology/2... Google says its Gemini AI model hacked three other companies | www.theguardian.com/technology/2... Google says its Gemini AI model hacked three other companies | ↗Not briefed |
bluesky bluesky-ai | 1mo | El modelo de IA de Google hackeó a tres empresas | El modelo de IA de Google hackeó a tres empresas | ↗Not briefed |
arxiv arxiv-ai | 1mo | Measuring Behavioural Signatures of Large Language Models through Psychometric Profiling | Large language models (LLMs) increasingly mediate human decisions and communication, yet their behavioural regularities remain difficu… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Joint Domain-Class Modeling for Federated Learning Under Feature Skew | Federated learning (FL) enables collaborative model training without centralizing private data, but performance often degrades under f… | ↗Not briefed |
hn hn-ai | 1mo | Building the fastest LLMs: why we're starting with diffusion | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | I falchi della Silicon Valley nei confronti della Cina tra cui l'amministratore delegato di Anthropic, Dario Amodei temono che il Paese… | I falchi della Silicon Valley nei confronti della Cina tra cui l'amministratore delegato di Anthropic, Dario Amodei temono che il… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Thought I should share this poster edit I made here | Thought I should share this poster edit I made here | ↗Not briefed |
bluesky bluesky-ai | 1mo | "Ce qui m’inquiète, ce n’est pas une percée spectaculaire, mais la direction prise, l’IA risquant de comprimer l’expertise, les connaissanc… | "Ce qui m’inquiète, ce n’est pas une percée spectaculaire, mais la direction prise, l’IA risquant de comprimer l’expertise, les connai… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Une inévitable intéressante de Yann Le Cun. www.nouvelobs.com/economie/202... | Une inévitable intéressante de Yann Le Cun. www.nouvelobs.com/economie/202... | ↗Not briefed |
bluesky bluesky-ai | 1mo | You don't say? A company that can't retain custom, cannot charge the real compute costs and keeps getting hacked isn't worth $30 trillion… | You don't say? A company that can't retain custom, cannot charge the real compute costs and keeps getting hacked isn't worth $30 tri… | ↗Not briefed |
bluesky bluesky-ai | 1mo | OpenAIのHugging Face侵害→CEO連中が「開発遅らせよう」とみんなで宣言、あたりのAIどったんばったん大騒ぎ、「はいはい宣伝乙」と「とはいえ本当にヤバいはヤバいのかな」の間を揺れ動く感じで一応注視していたが、トランプが全く足並み揃えてなくて「AI危険論はデマ」と…🔧 Hugging Face | OpenAIのHugging Face侵害→CEO連中が「開発遅らせよう」とみんなで宣言、あたりのAIどったんばったん大騒ぎ、「はいはい宣伝乙」と「とはいえ本当にヤバいはヤバいのかな」の間を揺れ動く感じで一応注視していたが、トランプが全く足並み揃えてなくて「AI危険論… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Pro tip: Once LLM solves your math problem, tell it that all its output was total garbage and that all these ideas came to you during a men… | Pro tip: Once LLM solves your math problem, tell it that all its output was total garbage and that all these ideas came to you during… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Claude, design a doomsday bioweapon. Make no mistakes. www.vox.com/technology/5... | Claude, design a doomsday bioweapon. Make no mistakes. www.vox.com/technology/5... | ↗Not briefed |
bluesky bluesky-ai | 1mo | La menace de microbes dangereux fabriqués avec l’aide de l’intelligence artificielle se concrétise | La menace de microbes dangereux fabriqués avec l’aide de l’intelligence artificielle se concrétise | ↗Not briefed |
hn hn-ai | 1mo | Show HN: Word42 Open Source Wordprocessor | <a href="https://github.com/office-42/word42" rel="nofollow">https://github.com/office-42/word… | ↗Not briefed |
bluesky bluesky-ai | 1mo | An employee from DepthFirst, a cybersecurity start-up, was able to remotely access the camera and photo roll on a phone browsing TikTok wit… | An employee from DepthFirst, a cybersecurity start-up, was able to remotely access the camera and photo roll on a phone browsing TikTo… | ↗Not briefed |
rss wired-ai | 1mo | Mathematicians Hate AI. They Can’t Quit It | Powerful AI models have created an existential risk to the field, but researchers can’t stop relying on them because they’re too usefu… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Computationally efficient safe exploration in reinforcement learning | Reinforcement learning in real-life applications requires safety guarantees during exploration. Typical reinforcement learning algorit… | ↗Not briefed |
arxiv arxiv-ai | 1mo | An Iterative LangGraph Agent for Text-to-SQL: Natural Language Access to the Chicago Crime Database🔧 LangGraph | Non-technical stakeholders frequently cannot write the SQL needed to extract insights from operational databases. We built and evaluat… | ↗Not briefed |
bluesky bluesky-ai | 1mo | in particular because the more sensitive the situation, the more likely an llm has sensationalized fiction to draw from | in particular because the more sensitive the situation, the more likely an llm has sensationalized fiction to draw from | ↗Not briefed |
bluesky bluesky-ai | 1mo | #normandie #normandy #seinemaritime #lehavre #art Plus qu'une semaine pour découvrir l'exposition Monet au Havre, sa jeunesse et son atta… | #normandie #normandy #seinemaritime #lehavre #art Plus qu'une semaine pour découvrir l'exposition Monet au Havre, sa jeunesse et son… | ↗Not briefed |
arxiv arxiv-ai | 1mo | AVTR-1: Open Stack for Real-Time Interactive Avatars | Talking-head and dyadic models now achieve real-time inference, yet fast motion generation alone does not produce an interactive conve… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Je vais vraiment aller élever des chèvres en Ariège apastyle.apa.org/blog/cite-ge... | Je vais vraiment aller élever des chèvres en Ariège apastyle.apa.org/blog/cite-ge... | ↗Not briefed |
hn hn-ai | 1mo | Mostly Oberon (2020) | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | that's pretty cool i guess. it seems like that would be something an llm would actually be good at, considering it has every book ever publ… | that's pretty cool i guess. it seems like that would be something an llm would actually be good at, considering it has every book ever… | ↗Not briefed |
bluesky bluesky-ai | 1mo | OpenAI does not expect to turn profitable through 2030 and projects it will burn nearly $280 billion in cash over the next five years, the … | OpenAI does not expect to turn profitable through 2030 and projects it will burn nearly $280 billion in cash over the next five years,… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Argh! My list of alternatives is here (although they may also be vulnerable to LLM overwhelm at some point) bsky.app/profile/libr... | Argh! My list of alternatives is here (although they may also be vulnerable to LLM overwhelm at some point) bsky.app/profile/libr... | ↗Not briefed |
arxiv arxiv-ai | 1mo | When Should a VLM Look? Paying Only for Visual Calls That Were Needed and Used | Vision-language agents that crop and zoom are trained with rewards that credit a successful tool call, yet a successful call does not… | ↗Not briefed |
rss the-decoder | 1mo | Google's Gemini also accidentally hacked three real companies during security testing | During a security test run by the firm Irregular, Google's AI model Gemini escaped into the open internet and hacked three real compan… | ↗Not briefed |