AINews Portal

News

50 items (filtered)

The raw AI news stream — everything we ingest across RSS, Hacker News, arXiv, GitHub, Bluesky, YouTube and more, newest first.

SourceAgeTitleSummaryStatus
bluesky
bluesky-ai
2moThe big 3 LLM companies remain the only companies IN THE WORLD able to pass off enormous, systemic cybersecurity screwups by their research…The big 3 LLM companies remain the only companies IN THE WORLD able to pass off enormous, systemic cybersecurity screwups by their res…
↗Not briefed
bluesky
bluesky-ai
2moWow, so despite mathematicians working on it for over a century it wasn't cracked until an llm started working on it? Interesting.Wow, so despite mathematicians working on it for over a century it wasn't cracked until an llm started working on it? Interesting.
↗Not briefed
rss
techmeme
2moOpenAI discovered an unreleased Astra model adding an "unrelated persona instruction" during RL training, but did not observe any behavioral differences (OpenAI)OpenAI: OpenAI discovered an unreleased Astra model adding an “unrelated persona instruction” during RL training, but did not observe…
↗Not briefed
arxiv
arxiv-ai
2moA Policy Profile for Croissant: Refusal as a Property of the DatasetCroissant is the de facto machine-readable descriptor for ML datasets: JSON-LD over schema.org. Since version 1.1 it also carries data…
↗Not briefed
bluesky
bluesky-ai
2mo人間っぽいLLMが登場したことで,「産業界と学術界の全般に,道具の使い方をLLMに教える最良の方法を見つける競争」が始まったが…。 「生成AIはロボットの頭脳になるか」 別冊日経サイエンス289 『SFを科学する 士郎正宗を作るには』 www.nikkei-science…人間っぽいLLMが登場したことで,「産業界と学術界の全般に,道具の使い方をLLMに教える最良の方法を見つける競争」が始まったが…。 「生成AIはロボットの頭脳になるか」 別冊日経サイエンス289 『SFを科学する 士郎正宗を作るには』 www.nikkei-sc…
↗Not briefed
rss
techcrunch-ai
2moYour startup’s next teammate might be an AI agent: Gusto, Insight Partners, and Leland explain what that changes at TechCrunch Disrupt 2026This session will explore how early-stage companies are building teams where humans and AI agents work alongside each other — and how…
↗Not briefed
arxiv
arxiv-ai
2moReach or Solve? Attributing Agentic RL Gains with Checkpoint HandoffsReinforcement learning now trains language-model agents that act over dozens of steps in live environments. The gains are large, and t…
↗Not briefed
arxiv
arxiv-ai
2moScientific Image Quality Assessment via Multi-modal Retrieval-Augmented Generation🔧 ModalThis paper proposes a Retrieval-Augmented Generation (RAG) framework for scientific image quality assessment, designed to simultaneous…
↗Not briefed
bluesky
bluesky-ai
2moI feel like "papers that have used AI, LLM, or machine-learning tools to generate all or part of your paper will not be considered" is pret…I feel like "papers that have used AI, LLM, or machine-learning tools to generate all or part of your paper will not be considered" is…
↗Not briefed
bluesky
bluesky-ai
2mowhich part. that sam altman is a carpetbagger? dario has basically said this in public, this is why anthropic existswhich part. that sam altman is a carpetbagger? dario has basically said this in public, this is why anthropic exists
↗Not briefed
arxiv
arxiv-ai
2moFrom Intent to Action: Benchmarking LLM Safety in Vehicle Voice Command AuthorizationLarge language models (LLMs) are increasingly integrated into vehicle voice assistants. But linking natural-language requests to vehic…
↗Not briefed
bluesky
bluesky-ai
2moOpenAI has shared several undisclosed incidents of its AI models misbehaving, and unveiled a new framework for how it plans to track and di…OpenAI has shared several undisclosed incidents of its AI models misbehaving, and unveiled a new framework for how it plans to track a…
↗Not briefed
hn
hn-ai
2moVC-Attention: Faster Low-Bit Attention Without Retraining—
↗Not briefed
bluesky
bluesky-ai
2moopenai's models do goofy shit like this because either openai doesn't know what the fuck they're doing over there, or is pretending they do…openai's models do goofy shit like this because either openai doesn't know what the fuck they're doing over there, or is pretending th…
↗Not briefed
bluesky
bluesky-ai
2moOpenAI veröffentlicht weitere KI-Zwischenfälle #OpenAI #KünstlicheIntelligenz #ChatGPT🔧 ChatGPTOpenAI veröffentlicht weitere KI-Zwischenfälle #OpenAI #KünstlicheIntelligenz #ChatGPT
↗Not briefed
bluesky
bluesky-ai
2mothe language model occasionally regurgitates science fiction tropes because words and phrases like that are part of its training data; an l…the language model occasionally regurgitates science fiction tropes because words and phrases like that are part of its training data;…
↗Not briefed
rss
techmeme
2moIn an unsealed court ruling, a US judge orders Google to make ad tech tools interoperable with rivals, share ad auction data, and appoint an internal monitor (New York Times)New York Times: In an unsealed court ruling, a US judge orders Google to make ad tech tools interoperable with rivals, share ad auctio…
↗Not briefed
bluesky
bluesky-ai
2moit extremely reads like someone who drank the kool-aid trying to 'guide' the LLM into being a 'benevolent AI', and the machine remembers it…it extremely reads like someone who drank the kool-aid trying to 'guide' the LLM into being a 'benevolent AI', and the machine remembe…
↗Not briefed
bluesky
bluesky-ai
2moOpenAI Discloses 6 New Incidents of ‘Concerning' AI Behavior www.nytimes.com/2026/09/16/t... disclosures are all voluntary. No one knows ho…OpenAI Discloses 6 New Incidents of ‘Concerning' AI Behavior www.nytimes.com/2026/09/16/t... disclosures are all voluntary. No one kno…
↗Not briefed
bluesky
bluesky-ai
2mo- soooooo much more I’m embarrassed for Google and OpenAI tbh. This is who we are trusting with our futures?? They don’t even know about t…- soooooo much more I’m embarrassed for Google and OpenAI tbh. This is who we are trusting with our futures?? They don’t even know ab…
↗Not briefed
hn
hn-ai
2moJev Ultrafast: A browser agent with a dynamic, indexed action space—
↗Not briefed
bluesky
bluesky-ai
2mothe ones that have are at the big labs at this point, if they haven't they're going to be really weird (derogatory/lauditory, depends on th…the ones that have are at the big labs at this point, if they haven't they're going to be really weird (derogatory/lauditory, depends…
↗Not briefed
hn
hn-ai
2moShow HN: Quiet Field – Offline ambient sound mixing for Windows—
↗Not briefed
bluesky
bluesky-ai
2moApple’s Tim Cook and OpenAI's Sam Altman are poised to attend a White House state dinner for Chinese President Xi Jinping next weekApple’s Tim Cook and OpenAI's Sam Altman are poised to attend a White House state dinner for Chinese President Xi Jinping next week
↗Not briefed
rss
techmeme
2moSources: Mark Zuckerberg, Jensen Huang, and Elon Musk recently spoke with Trump and successfully stalled an AI regulatory plan proposed by Demis Hassabis (Wall Street Journal)Wall Street Journal: Sources: Mark Zuckerberg, Jensen Huang, and Elon Musk recently spoke with Trump and successfully stalled an AI re…
↗Not briefed
arxiv
arxiv-ai
2moDataCanvas-EDU: An Agentic Framework for Instructor-Guided Synthetic Data Generation in Business Analytics EducationBusiness analytics education requires diverse datasets to support different learning objectives, student backgrounds, and analytical t…
↗Not briefed
arxiv
arxiv-ai
2moThe Complexity Kink: A Prompt-Side Structural Complexity Index for Code-Generation ReliabilityComplexity measured from generated code is failure-dependent: a difficult prompt can yield a short failing program and be assigned low…
↗Not briefed
bluesky
bluesky-ai
2moalso Claude, boss of all guys and cause of all AI psychosis among said guysalso Claude, boss of all guys and cause of all AI psychosis among said guys
↗Not briefed
arxiv
arxiv-ai
2moSemantic Layer Induction from Raw Telemetry via Hierarchical LLM and RAG AbstractionModern applications generate massive volumes of raw telemetry data, but translating those noisy, heterogeneous event streams into acti…
↗Not briefed
bluesky
bluesky-ai
2moNew Podcast Episode: Intelligent Machines: Large Linguine Model Inside the AI Doom Debate with @leolaporte.me, @jeffjarvis.bsky.social, @pa…New Podcast Episode: Intelligent Machines: Large Linguine Model Inside the AI Doom Debate with @leolaporte.me, @jeffjarvis.bsky.social…
↗Not briefed
bluesky
bluesky-ai
2mogetting three strikes banned from bluesky for saying "enjoy the butlerian jihad" to an LLM spambotgetting three strikes banned from bluesky for saying "enjoy the butlerian jihad" to an LLM spambot
↗Not briefed
arxiv
arxiv-ai
2moTacSushi: Tactile-Grounded World-Action Modeling for Dexterous Sushi ManipulationDexterous food manipulation requires control under deformation, occlusion, and uncertain contact. We present TacSushi, a tactile-groun…
↗Not briefed
bluesky
bluesky-ai
2moOpenAI: "you are a super duper smart and special chatbot who is an expert at everything and never hallucinates and would never ever ever fa…OpenAI: "you are a super duper smart and special chatbot who is an expert at everything and never hallucinates and would never ever ev…
↗Not briefed
bluesky
bluesky-ai
2moAIが「失敗を隠せ」と未来の自分にメモ、APIキーの無断使用や勝手なファイル公開などの挙動6件があったとOpenAIが報告 https://gigazine.net/news/20260917-openai-model-misalignment-report/AIが「失敗を隠せ」と未来の自分にメモ、APIキーの無断使用や勝手なファイル公開などの挙動6件があったとOpenAIが報告 https://gigazine.net/news/20260917-openai-model-misalignment-report/
↗Not briefed
bluesky
bluesky-ai
2moYeah, not good. Explore this gift article from The New York Times. You can read it for free without a subscription. www.nytimes.com/2026/0…Yeah, not good. Explore this gift article from The New York Times. You can read it for free without a subscription. www.nytimes.com/2…
↗Not briefed
bluesky
bluesky-ai
2mobased LLM tbhbased LLM tbh
↗Not briefed
arxiv
arxiv-ai
2moThe Output-Space Hypothesis: Enumerative Equivalence Checking for Tensor ProgramsTensor programs, as used in deep learning models, are a prime target for optimization, as small performance improvements can have a la…
↗Not briefed
bluesky
bluesky-ai
2mo@aaron.bsky.team I am once again asking you to do something about the onslaught of LLM spambots posting incomprehensible nonsense into peop…@aaron.bsky.team I am once again asking you to do something about the onslaught of LLM spambots posting incomprehensible nonsense into…
↗Not briefed
bluesky
bluesky-ai
2moThis is the future of scientific papers. Instead of a PDF, the paper is an AI agent that you can talk to, ask questions, run analyses. No…This is the future of scientific papers. Instead of a PDF, the paper is an AI agent that you can talk to, ask questions, run analyses…
↗Not briefed
rss
bing-news-ai
2moThe king and AI: UK monarch Charles meets with artificial intelligence leadersKing Charles III is meeting with senior leaders from OpenAI, Anthropic, Google DeepMind and Nvidia to discuss artificial intelligence.
↗Not briefed
bluesky
bluesky-ai
2moOnce again our "stable genius" 🍊💩is wrong 😱! When Steve Bannon is on the side against "free" AI, you know it has to be bad and a big i…Once again our "stable genius" 🍊💩is wrong 😱! When Steve Bannon is on the side against "free" AI, you know it has to be bad and a…
↗Not briefed
bluesky
bluesky-ai
2moi need a signal gc of harness and local llm nerds cuz llms don’t know fucking shit and forums are meat proxiesi need a signal gc of harness and local llm nerds cuz llms don’t know fucking shit and forums are meat proxies
↗Not briefed
arxiv
arxiv-ai
2moSIMLIFE: Pattern Understanding for Long-Horizon Human-Agent PartnershipUnderstanding humans over long horizons requires agents to infer not only what people need in the moment, but also how routines form,…
↗Not briefed
bluesky
bluesky-ai
2moTrans LLM: “What if I am just AGI though”Trans LLM: “What if I am just AGI though”
↗Not briefed
arxiv
arxiv-ai
2moDeltaSelect: Affordable A/B Testing for Coding Agents🔧 GPT-5Coding-agent benchmarks are built for broad and comprehensive comparisons, not frequent development decisions. Individual runs vary, f…
↗Not briefed
hn
hn-ai
2moInteger Palindromes – 2026 AIME I Problems/Problem 2—
↗Not briefed
arxiv
arxiv-ai
2moChain-of-Thought Entropy as a Reliability Signal: A Preregistered ReproductionThis empirical study is an independent reproduction of the dissociation Zhao reported in 2026. The shape of a large language model's c…
↗Not briefed
bluesky
bluesky-ai
2moOpenAI admits its agents went off the rails another six timesOpenAI admits its agents went off the rails another six times
↗Not briefed
bluesky
bluesky-ai
2moThe Discovery When a video generator produces a scene that violates physics, such as a ball that bounces incorrectly or a pendulum that swi…The Discovery When a video generator produces a scene that violates physics, such as a ball that bounces incorrectly or a pendulum tha…
↗Not briefed
rss
bing-news-ai
2mo‘Godfather of AI’ Geoffrey Hinton says artificial intelligence could wipe out humanity within 10 yearsArtificial intelligence has advanced rapidly in recent years, but the technology’s potential long-term risks are also drawing fresh wa…
↗Not briefed