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
hn
hn-ai
2moDart: Denoising Autoregressive Transformer—
↗Not briefed
bluesky
bluesky-ai
2moFrom Stephen Hawking to Jacob Coxon’s viral Anthropic resignation, fears that AI could threaten humanity have shaken the industry without s…From Stephen Hawking to Jacob Coxon’s viral Anthropic resignation, fears that AI could threaten humanity have shaken the industry with…
↗Not briefed
bluesky
bluesky-ai
2moConsistent with Suicide of a Nation in the sense it's hallucinated by an AI model?Consistent with Suicide of a Nation in the sense it's hallucinated by an AI model?
↗Not briefed
bluesky
bluesky-ai
2mo(E.g., Our LLM "blackmailed" people in a fictional scenario. There was never going to be a conclusion of that study where the model was not…(E.g., Our LLM "blackmailed" people in a fictional scenario. There was never going to be a conclusion of that study where the model wa…
↗Not briefed
hn
hn-ai
2moShow HN: MyMarkdown – Make AI Markdown easier to readI made a small free tool for humans (not AI) to read Markdown faster.<p>If you build AI tools, you probably read a lot of Markdown. An…
↗Not briefed
arxiv
arxiv-ai
2moOne Size Does Not Fit All! Dynamic Retriever and Generator Selection for RAGRetrieval-Augmented Generation (RAG) systems typically employ fixed retriever and generator configurations across queries, despite sub…
↗Not briefed
rss
techmeme
2moGoogle launches Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking, its "most advanced live dialogue models yet", to more effectively enable voice agents (Google)Google: Google launches Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking, its “most advanced live dialogue models yet”, to more e…
↗Not briefed
bluesky
bluesky-ai
2moJust putting it out there... Maybe techbro lads haven't suddenly started to care AI might kill us all but rather they're in waaaaaay too de…Just putting it out there... Maybe techbro lads haven't suddenly started to care AI might kill us all but rather they're in waaaaaay t…
↗Not briefed
arxiv
arxiv-ai
2moConfidence Comes from Experience: Experiential Confidence Estimation from Reasoning to AgentsReliable confidence estimation is increasingly central to the trustworthy deployment of language models: a calibrated estimate of the…
↗Not briefed
bluesky
bluesky-ai
2mo@milesklee.bsky.social if i boot up any old llm can i talk to tilly or did they give you a special seed?@milesklee.bsky.social if i boot up any old llm can i talk to tilly or did they give you a special seed?
↗Not briefed
bluesky
bluesky-ai
2moStop calling it an “AI actress.” It’s a LLM GUI prompt recording.Stop calling it an “AI actress.” It’s a LLM GUI prompt recording.
↗Not briefed
bluesky
bluesky-ai
2moO que a China disse sobre a discussão de regulação da IA nos EUA: Uma fala de um ex-funcionário da Anthropic, afirmando que existe um risco…O que a China disse sobre a discussão de regulação da IA nos EUA: Uma fala de um ex-funcionário da Anthropic, afirmando que existe um…
↗Not briefed
bluesky
bluesky-ai
2moThe professors and instructors who turn classrooms back into spaces for reading and discussion, are going to leave the biggest legacies in …The professors and instructors who turn classrooms back into spaces for reading and discussion, are going to leave the biggest legacie…
↗Not briefed
bluesky
bluesky-ai
2mo🤷‍♂️ “Microsoft zakazuje svým AI modelům spouštět kybernetické útoky nebo zvyšovat svůj vlastní přístup.” … tohle už začíná být doslova v…🤷‍♂️ “Microsoft zakazuje svým AI modelům spouštět kybernetické útoky nebo zvyšovat svůj vlastní přístup.” … tohle už začíná být dosl…
↗Not briefed
bluesky
bluesky-ai
2mo**Don't make the AI responsible for enforcing the security boundary that contains the AI.** 😬 AI agents are escaping their intended secur…**Don't make the AI responsible for enforcing the security boundary that contains the AI.** 😬 AI agents are escaping their intended…
↗Not briefed
arxiv
arxiv-ai
2moNeMo Data Designer: An Extensible Framework for Multimodal Synthetic Data Generation🔧 ModalWe present NeMo Data Designer (NDD), an open-source, general-purpose framework for multi-modal synthetic data generation (SDG). Design…
↗Not briefed
hn
hn-ai
2moOpenRouter users spent more on OpenAI models than on Anthropic models last week🔧 OpenRouter—
↗Not briefed
hn
hn-ai
2moDay 13. Three production applications on AWS downWe’re on day 13 of all 3 of our production applications on AWS down. We were going to market and onboarding users.<p>We were resolving…
↗Not briefed
hn
hn-ai
2moFluid Script: Simple .NET server side scripting with browser breakpoints—
↗Not briefed
hn
hn-ai
2moHackers say they breached Italian state email to target Revolut 'crypto whales'—
↗Not briefed
arxiv
arxiv-ai
2moComposite-Gradient Learning for Shared Control Authority Between Deep Reinforcement Learning and Model Predictive ControlIntegrated deep reinforcement learning (DRL) and model predictive control (MPC) methods are increasingly used to control autonomous sy…
↗Not briefed
arxiv
arxiv-ai
2moGVD: Governed Versioning and Deduplication for Document RepositoriesDocument repositories evolve continuously. Guidelines and policies are revised, superseded, and re-uploaded, so the same content recur…
↗Not briefed
bluesky
bluesky-ai
2moSomeone has probably built the LLM chat interface where the tree is allowed to branch, right? There's costs to memory efficiency, but keep…Someone has probably built the LLM chat interface where the tree is allowed to branch, right? There's costs to memory efficiency, but…
↗Not briefed
bluesky
bluesky-ai
2moJust got a phishing email from "OpenAI" like dude, I'd mark you as spam or phishing even if you were legit.Just got a phishing email from "OpenAI" like dude, I'd mark you as spam or phishing even if you were legit.
↗Not briefed
arxiv
arxiv-ai
2moGraphEcho: Structural Redundancy and Evidence Provenance in LLM Graph AgentsA large language model (LLM) agent can follow more graph paths without acquiring more independent evidence. GraphEcho tests whether ag…
↗Not briefed
hn
hn-ai
2moShow HN: A writing coach that's not allowed to write your sentencesHi HN,<p>I&#x27;m the technical advisor and one of the co-founders for Oghmere (<a href="https:&#x2F;&#x2F;oghmere.com" rel="nofollow…
↗Not briefed
bluesky
bluesky-ai
2moOpenAI: AI capabilities are in a ‘different place’ and require urgencyOpenAI: AI capabilities are in a ‘different place’ and require urgency
↗Not briefed
arxiv
arxiv-ai
2moAccelerating Diffusion Sampling via Speculative Draft TreesSpeculative sampling accelerates diffusion model generation by drafting inexpensive candidate states and correcting them under a coupl…
↗Not briefed
bluesky
bluesky-ai
2moI wonder, for example, how universities like my employer are going to square their "Carbon Zero" commitments with mass subsidies for carbon…I wonder, for example, how universities like my employer are going to square their "Carbon Zero" commitments with mass subsidies for c…
↗Not briefed
bluesky
bluesky-ai
2moAI-firmojen pomot väittävät tekoälyjä vaarallisiksi, koska se on helpompi keino valmistautua siihen, että ihmiset huomaavat / ovat huomanne…AI-firmojen pomot väittävät tekoälyjä vaarallisiksi, koska se on helpompi keino valmistautua siihen, että ihmiset huomaavat / ovat huo…
↗Not briefed
bluesky
bluesky-ai
2moSo, there is this idea that an LLM is: - Latently aligned/misaligned - Tests aim to uncover which one is the case But in neither case can…So, there is this idea that an LLM is: - Latently aligned/misaligned - Tests aim to uncover which one is the case But in neither cas…
↗Not briefed
arxiv
arxiv-ai
2moCapMem: A Benchmark for Caption-Based Episodic Memory in Egocentric VideoWearable assistants require episodic memory over egocentric video, yet current vision-language models face bounded frame budgets, grow…
↗Not briefed
arxiv
arxiv-ai
2moThe Missing "I Don't Know": Why Three Reasoning-Reliability Findings Converge on Calibrated AbstentionThree recent results describe what look like unrelated LLM reliability problems. Yin et al. (2026) show reasoning RL collapses tool-re…
↗Not briefed
bluesky
bluesky-ai
2motalkingpointsmemo.com/edblog/ai-fr...talkingpointsmemo.com/edblog/ai-fr...
↗Not briefed
bluesky
bluesky-ai
2moAnd OpenAI must be renamed AIOuvert to comply with French naming rules.And OpenAI must be renamed AIOuvert to comply with French naming rules.
↗Not briefed
arxiv
arxiv-ai
2moDSD: Learning Diverse and Reusable Motor Skills via Diffusion Skill DiscoveryHumans efficiently learn new tasks by reusing a rich repertoire of motor skills across different goals and contexts. A similar strateg…
↗Not briefed
rss
wired-ai
2moAI ‘Actor’ Tilly Norwood Told Me That ‘All Lives Matter’The virtual character, which is promoting its upcoming movie Misaligned, tries to evade politics by repetitively commenting on the clo…
↗Not briefed
hn
hn-ai
2moShow HN: Amika – Multiplayer cloud workstations for coding agents and humans--- title: &quot;Show HN: Amika – Multiplayer cloud workstations for coding agents and humans&quot; ---<p>Hi Hacker News!<p>My cofound…
↗Not briefed
arxiv
arxiv-ai
2moAgentic Societies Need a Social HarnessAn agentic society is a collection of AI agents that coordinate autonomously across trust boundaries, on behalf of different principal…
↗Not briefed
arxiv
arxiv-ai
2moEfficient Robust Learning at the Information-Theoretic LimitIn an important recent work, Blanc (2026) gave an algorithm for robustly learning Boolean concept classes with respect to a fixed dist…
↗Not briefed
hn
hn-ai
2moHugging Face is billing OpenAI $100M for hacking it🔧 Hugging Face—
↗Not briefed
rss
bing-news-ai
2moArtificial intelligence industry leaders are calling for safety guardrailsUNR professor says AI guardrails will build public trust and build a system of accountability ...
↗Not briefed
rss
bing-news-ai
2moArtificial intelligence industry leaders are calling for safety guardrailsUNR professor says AI guardrails will build public trust and build a system of accountability ...
↗Not briefed
arxiv
arxiv-ai
2moScienceBuddy: Recursive-in-Recursive Self-Improvement for Interactive Scientific AgentsWe introduce and release ScienceBuddy, an interactive scientific research workspace that brings continually improving scientific agent…
↗Not briefed
arxiv
arxiv-ai
2moPhysStream: Streaming Physics-Grounded Video Generation with Structured Scene Memory and Fine-Grained Motion ControlInteractive control for video generation is moving from coarse prompts toward fine-grained, physically meaningful manipulation of dyna…
↗Not briefed
bluesky
bluesky-ai
2mo(4/6) (1) requiring large AI firms to report critical safety incidents in a timely manner (2) requiring independent safety reviews before …(4/6) (1) requiring large AI firms to report critical safety incidents in a timely manner (2) requiring independent safety reviews be…
↗Not briefed
arxiv
arxiv-ai
2moWhen Should LLMs Abstain? Chain-of-Self-Questioning for Selective Risk ControlLarge language models can produce fluent answers when their factual support is weak. This paper introduces Chain-of-Self-Questioning (…
↗Not briefed
rss
the-decoder
2moAI labs have a data trust problem that their policies haven't solvedOpenAI and Anthropic tell corporate customers their data won't be used for training. But when Anthropic said it would store usage logs…
↗Not briefed
arxiv
arxiv-ai
2moWhat Breaks Under Pruning in Smart Homes, and When? Evaluating LLM Degradation Across Architectures and Task ComplexityPruning can reduce the deployment cost of large language models (LLMs), but its impact on context-grounded tool calling remains poorly…
↗Not briefed
bluesky
bluesky-ai
2moWell if you want to become less confused about the Anthropic thing definitely don't call the number in that email because I just had the mo…Well if you want to become less confused about the Anthropic thing definitely don't call the number in that email because I just had t…
↗Not briefed