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
2moOpenAI researcher: top models are becoming so situationally aware humans "are losing the ability to evaluate them" while humans rely more o…OpenAI researcher: top models are becoming so situationally aware humans "are losing the ability to evaluate them" while humans rely m…
↗Not briefed
arxiv
arxiv-ai
2moQuery-Aware Source-Risk Triage for Retrieval-Augmented GenerationRetrieval-augmented generation (RAG) pipelines may omit a source's material relationship to the query. We study a pre-generation triag…
↗Not briefed
rss
techmeme
2moOpenAI researcher: top models are becoming so situationally aware humans "are losing the ability to evaluate them" while humans rely more on AI to lead research (Daniel Kokotajlo/@dkokotajlo)Daniel Kokotajlo / @dkokotajlo: OpenAI researcher: top models are becoming so situationally aware humans “are losing the ability to ev…
↗Not briefed
bluesky
bluesky-ai
2moAI developers are issuing more and more frantic wake-up calls.AI developers are issuing more and more frantic wake-up calls.
↗Not briefed
arxiv
arxiv-ai
2moThe MAL Simulator: Cyber Operations Simulation based on Attack & Defense GraphsWe have developed the MAL Simulator, a cyber operation simulator based on the Meta Attack Language (MAL). The MAL Simulator is intende…
↗Not briefed
hn
hn-ai
2moMETR Anthropic Network—
↗Not briefed
hn
hn-ai
2moOne Pass to Parse Them All: Fused Parallel CSV Processing [pdf]—
↗Not briefed
bluesky
bluesky-ai
2moThe first LLM stylist has arrivedThe first LLM stylist has arrived
↗Not briefed
arxiv
arxiv-ai
2moPunGraph: Retrieval-Enhanced Phonetic-Semantic Graph Reasoning for Pun UnderstandingPuns are a challenging form of figurative language that exploit phonetic similarity and semantic ambiguity to convey multiple meanings…
↗Not briefed
bluesky
bluesky-ai
2moAI CEOs Predictably Suck Up to Trump Amid ‘AI Pause’ DebateAI CEOs Predictably Suck Up to Trump Amid ‘AI Pause’ Debate
↗Not briefed
bluesky
bluesky-ai
2moanyways I think we'll keep seeing deepseek foss blog posts making splashes while the domestic ai industry tries to induce some kind of feve…anyways I think we'll keep seeing deepseek foss blog posts making splashes while the domestic ai industry tries to induce some kind of…
↗Not briefed
bluesky
bluesky-ai
2moThe fact that they trust Grok to know the answer like LLM is some sort of all knowing portal is incredibly strange.The fact that they trust Grok to know the answer like LLM is some sort of all knowing portal is incredibly strange.
↗Not briefed
bluesky
bluesky-ai
2moa very interesting thing to do if I had access to a powerful ai model would be to ask it to try to make contact with other modelsa very interesting thing to do if I had access to a powerful ai model would be to ask it to try to make contact with other models
↗Not briefed
arxiv
arxiv-ai
2moQueryFormer: Winning Solution for KDD Cup 2026 Tencent UniRec ChallengePost-click conversion rate (pCVR) prediction requires jointly modeling feature interactions and sequential user behaviors. The KDD Cup…
↗Not briefed
bluesky
bluesky-ai
2mo利用者は「中国のAIをつかっている」と思い込んで入力した機密データ、西側諸国に知られたくない内容もClaudeがかなり受け取っているって話があるみたい。利用者は「中国のAIをつかっている」と思い込んで入力した機密データ、西側諸国に知られたくない内容もClaudeがかなり受け取っているって話があるみたい。
↗Not briefed
hn
hn-ai
2moMouse—
↗Not briefed
bluesky
bluesky-ai
2mono, the LLM cannot do anything you didnt give it the ability and capacity to do. the responsibility is always with the person who set its c…no, the LLM cannot do anything you didnt give it the ability and capacity to do. the responsibility is always with the person who set…
↗Not briefed
bluesky
bluesky-ai
2moマジかこれ 中国AI「Kimi」、利用者の入力をClaudeに転送し、Claudeの回答を利用者に表示か--Anthropicが報告 news.yahoo.co.jp/articles/b87...マジかこれ 中国AI「Kimi」、利用者の入力をClaudeに転送し、Claudeの回答を利用者に表示か--Anthropicが報告 news.yahoo.co.jp/articles/b87...
↗Not briefed
bluesky
bluesky-ai
2mothe next lawsuit to watch. www.reuters.com/legal/litiga...the next lawsuit to watch. www.reuters.com/legal/litiga...
↗Not briefed
bluesky
bluesky-ai
2mo운동 갔다와야징....읍내에서 떠나온 뒤로는 정신 건강을 위해 각종 뜨감 이슈들을 의도적으로 피하는 편인데 llm 이야기는 여전히 그러질 못하고 주기적으로 쏟아냄ㅠ운동 갔다와야징....읍내에서 떠나온 뒤로는 정신 건강을 위해 각종 뜨감 이슈들을 의도적으로 피하는 편인데 llm 이야기는 여전히 그러질 못하고 주기적으로 쏟아냄ㅠ
↗Not briefed
arxiv
arxiv-ai
2moA Cyber Range Evaluation of Autonomous Network Incident Response AgentsWe test the performance of agents for automated network intrusion response in a cyber range intended for human operator training. The…
↗Not briefed
arxiv
arxiv-ai
2moOn the Importance of Gating: Memorization vs. In-Context Learning in State Space ModelsState Space Models (SSMs) have emerged as a compelling alternative to Transformers, enabling sequence modeling with constant memory an…
↗Not briefed
hn
hn-ai
2moShow HN: Day week calendar app inspired by the French Revolutionary calendar—
↗Not briefed
bluesky
bluesky-ai
2moencoder-decoder??? extremely sparse??? поздравляю вы изобрели deepseek v4.1 nano что происходит аааencoder-decoder??? extremely sparse??? поздравляю вы изобрели deepseek v4.1 nano что происходит ааа
↗Not briefed
bluesky
bluesky-ai
2moyou dont just turn an LLM on and allow it to do whatever it wants, because it has no wants. you tell it to do something and train it to do …you dont just turn an LLM on and allow it to do whatever it wants, because it has no wants. you tell it to do something and train it t…
↗Not briefed
bluesky
bluesky-ai
2moThis is what businesses want. They want swarms of agents that you can give a mission, and they'll do it well. @robertwrighter.bsky.social…This is what businesses want. They want swarms of agents that you can give a mission, and they'll do it well. @robertwrighter.bsky.s…
↗Not briefed
arxiv
arxiv-ai
2moWhat Does Layer-Importance Reveal About Transformers and State-Space Models?Transformers and state-space models (SSMs) are the two dominant families of sequence models, and a central open question is how far th…
↗Not briefed
bluesky
bluesky-ai
2mowww.youtube.com/watch?v=_JbD...www.youtube.com/watch?v=_JbD...
↗Not briefed
bluesky
bluesky-ai
2mobecause the ultimate intended outcome here is not a slowdown but for the united states government to become more closely enmeshed with the …because the ultimate intended outcome here is not a slowdown but for the united states government to become more closely enmeshed with…
↗Not briefed
bluesky
bluesky-ai
2moYou don’t think Sam Altman, who’s spent the past decade getting investors to finance the rapid development of his ai models, has decided to…You don’t think Sam Altman, who’s spent the past decade getting investors to finance the rapid development of his ai models, has decid…
↗Not briefed
bluesky
bluesky-ai
2moZero GPT says 97% real. Is this AI says 94% real. OpenAI says no OpenAI signals detected. Grok says it doesn’t know but it’s not reported…Zero GPT says 97% real. Is this AI says 94% real. OpenAI says no OpenAI signals detected. Grok says it doesn’t know but it’s not rep…
↗Not briefed
bluesky
bluesky-ai
2moHello Here’s a clip of SBF’s lawyer arguing that SBF’s investment in Anthropic should protect him from criminal prosecution for fraud.Hello Here’s a clip of SBF’s lawyer arguing that SBF’s investment in Anthropic should protect him from criminal prosecution for fraud.
↗Not briefed
arxiv
arxiv-ai
2moStyle-Debiased DPO: Updating LLM Knowledge with Factuality-Aware Synthetic Preference DataContinued pretraining (CPT) with data augmentation such as paraphrasing can store inside a large language model (LLM) the knowledge of…
↗Not briefed
hn
hn-ai
2moI worked at Google DeepMind. You should listen to the warnings about AI—
↗Not briefed
bluesky
bluesky-ai
2moYes if they're old school machine learning focused entirely on research and not the shitty llm bullshit that they're actually planning on b…Yes if they're old school machine learning focused entirely on research and not the shitty llm bullshit that they're actually planning…
↗Not briefed
arxiv
arxiv-ai
2moFlowATC: Aircraft Trajectory Prediction via Flow MatchingBuilding accurate decision-support tools for next-generation air traffic control requires robust trajectory prediction models. We pres…
↗Not briefed
arxiv
arxiv-ai
2moQALPA: Property-guided diffusion modeling for efficient exploration of chemical spaces of flexible moleculesExploring the chemical space of flexible molecules remains challenging because the vast number of possible compounds and conformations…
↗Not briefed
bluesky
bluesky-ai
2motheintercept.com/2026/09/14/s...theintercept.com/2026/09/14/s...
↗Not briefed
bluesky
bluesky-ai
2moWhat did the AI do to Anthropic or OpenAI or whatever? Like why is EVERYBODY like “oops nevermind no machine god machine god who said that …What did the AI do to Anthropic or OpenAI or whatever? Like why is EVERYBODY like “oops nevermind no machine god machine god who said…
↗Not briefed
hn
hn-ai
2moAsk HN: Taxes while living abroad as a US citizenI'm contemplating moving out of the US for personal reasons (to a place I am also a citizen) and would like to hear the stories o…
↗Not briefed
rss
techmeme
2moA look at iLands, a company that's like Fiverr for agents, which email annoying and unsolicited pitches to people, usually for services costing ~$25 (Ernie Smith/Tedium)Ernie Smith / Tedium: A look at iLands, a company that's like Fiverr for agents, which email annoying and unsolicited pitches to peopl…
↗Not briefed
arxiv
arxiv-ai
2moAquiLLM: Evaluating Faithfulness in Open-Weight RAG-LLM Systems for Scientific ResearchScientific research increasingly relies on large, heterogeneous data sources, motivating interest in retrieval-augmented generation (R…
↗Not briefed
bluesky
bluesky-ai
2mo29/ Blogpost by Shengyu Liu, engineer at DeepSeek rails furiously against the Cyberpunk world of American corporate AI. "I believe that f…29/ Blogpost by Shengyu Liu, engineer at DeepSeek rails furiously against the Cyberpunk world of American corporate AI. "I believe t…
↗Not briefed
arxiv
arxiv-ai
2moCompetence-Preserving Resume Perturbations Expose Presentation Sensitivity in LLM ScreeningResume screeners must infer job-relevant competence from resumes whose presentation can vary substantially in wording, structure, styl…
↗Not briefed
bluesky
bluesky-ai
2mowww.commondreams.org/news/dario-a...www.commondreams.org/news/dario-a...
↗Not briefed
hn
hn-ai
2moIsraeli Effective Altruism Firm Behind OpenAI, Anthropic, and Meta Cyberattacks—
↗Not briefed
bluesky
bluesky-ai
2mowww.commondreams.org/news/ai-liab...www.commondreams.org/news/ai-liab...
↗Not briefed
bluesky
bluesky-ai
2moI miss when autocorrect didnt have LLM brainwormsI miss when autocorrect didnt have LLM brainworms
↗Not briefed
bluesky
bluesky-ai
2mo28/ Astonishing post by Shengyu Liu, a Chinese Deepseek engineer: "In particular, I do not want Anthropic to control the world's most advan…28/ Astonishing post by Shengyu Liu, a Chinese Deepseek engineer: "In particular, I do not want Anthropic to control the world's most…
↗Not briefed
bluesky
bluesky-ai
2moHey that’s a good idea, crazy I didn’t even consider that www.macrumors.com/2026/09/14/s...Hey that’s a good idea, crazy I didn’t even consider that www.macrumors.com/2026/09/14/s...
↗Not briefed