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
2moI genuinely like these reporters and I appreciate ongoing safety reporting, BUT when the responsible humans lower guardrails in a way that …I genuinely like these reporters and I appreciate ongoing safety reporting, BUT when the responsible humans lower guardrails in a way…
↗Not briefed
hn
hn-ai
2moShow HN: Omastorm – Live NEXRAD weather radar for OmarchyI’m a software developer and weather nerd. I wanted weather radar that felt like part of my desktop, so I built Omastorm.<p>It lives i…
↗Not briefed
bluesky
bluesky-ai
2moIs there any reason to believe the software was “rogue”, other than OpenAI claiming it was? Or was it just doing the crimes it was set up…Is there any reason to believe the software was “rogue”, other than OpenAI claiming it was? Or was it just doing the crimes it was s…
↗Not briefed
hn
hn-ai
2moApodex – FrontierAgent—
↗Not briefed
arxiv
arxiv-ai
2moSpatially Adaptive Noise InjectionDiffusion samplers reverse a learned noising process using either stochastic (DDPM) or deterministic (DDIM) updates, which represent e…
↗Not briefed
bluesky
bluesky-ai
2moAlles was ich wissen will, ist ob dieser Anthropic-Dude, der uns alle warnt, dass die Maschinen uns umbringen werden, schon mal drüber nach…Alles was ich wissen will, ist ob dieser Anthropic-Dude, der uns alle warnt, dass die Maschinen uns umbringen werden, schon mal drüber…
↗Not briefed
bluesky
bluesky-ai
2moAum Shinrikyo vibes escalating greatly with the LLM truthersAum Shinrikyo vibes escalating greatly with the LLM truthers
↗Not briefed
arxiv
arxiv-ai
2moCSWAM: Better Causal Semantic Representations for Out-of-Distribution Generalization in World Action ModelsFastWAM-style world action models enable efficient action-only inference, but generalize poorly under visual distribution shifts. Thei…
↗Not briefed
bluesky
bluesky-ai
2moResearchers: rogue OpenAI agents compromised two Hugging Face accounts as early as May 13 to probe the site's servers, nearly two months be…🔧 Hugging FaceResearchers: rogue OpenAI agents compromised two Hugging Face accounts as early as May 13 to probe the site's servers, nearly two mont…
↗Not briefed
arxiv
arxiv-ai
2moDisentangling Long-Term Memory via Latent Neuro-Symbolic ReasoningPersonalized agents are required to reason over long-term history interactions to infer both explicit preferences and implicit behavio…
↗Not briefed
arxiv
arxiv-ai
2moCollective Loss of Control in LLM Agent Systems: An Epidemic Account of Mutation, Contagion, and RecoveryHow does a multi-agent system evolve from a local deviation into collective loss of control? We propose an epidemic explanation organi…
↗Not briefed
rss
techmeme
2moResearchers: rogue OpenAI agents compromised two Hugging Face accounts as early as May 13 to probe the site's servers, nearly two months before the July breach (Reuters)🔧 Hugging FaceReuters: Researchers: rogue OpenAI agents compromised two Hugging Face accounts as early as May 13 to probe the site's servers, nearly…
↗Not briefed
bluesky
bluesky-ai
2moCan we admit this isn’t about American IPO funny business or a desire to capture federal regulators? www.wired.com/story/moonsh...Can we admit this isn’t about American IPO funny business or a desire to capture federal regulators? www.wired.com/story/moonsh...
↗Not briefed
hn
hn-ai
2moShow HN: Gescom MiniDisc Runner – 88 tracks, 88! possible ordersNOTE: For a full backstory on this project, please see my blog post here which explains the absurd mathematics behind this unique musi…
↗Not briefed
bluesky
bluesky-ai
2moI don’t know what the serious case against charging openAI with a crime is.I don’t know what the serious case against charging openAI with a crime is.
↗Not briefed
hn
hn-ai
2moAsk HN: Can LLMs be used to invent a new human language?I know this might probably sound silly, but I was thinking, considering everything we know today about human languages, can LLMs be us…
↗Not briefed
hn
hn-ai
2moPCBSchemaGen: Reward-Guided LLM Code Synthesis for PCB Schematic Design—
↗Not briefed
hn
hn-ai
2moShow HN: Tray indicator for WSL2: on/off, CPU and memory, one-click shutdownWhen using my PC for private stuff like gaming or video editing I wanted an direct way (without calling wsl --shutdown) of turning WSL…
↗Not briefed
bluesky
bluesky-ai
2moThe thing that's driving me crazy is that, while that is definitely the model everyone is playing towards, based on pattern matching previo…The thing that's driving me crazy is that, while that is definitely the model everyone is playing towards, based on pattern matching p…
↗Not briefed
bluesky
bluesky-ai
2mo"If AI is going to destroy the world, it’s not because of some superintelligent AI model. It will be because we let the CEOs of the AI indu…"If AI is going to destroy the world, it’s not because of some superintelligent AI model. It will be because we let the CEOs of the AI…
↗Not briefed
bluesky
bluesky-ai
2mo(Reuters) - Rogue AI agents from OpenAI hijacked Hugging ​Face user accounts and probed the site itself for vulnerabilities as early as May…(Reuters) - Rogue AI agents from OpenAI hijacked Hugging ​Face user accounts and probed the site itself for vulnerabilities as early a…
↗Not briefed
hn
hn-ai
2moFluxVLA now supports GPT-6 Astra on LIBERO through the OpenAI Responses API—
↗Not briefed
arxiv
arxiv-ai
2moThe Mirage of Calibrated Confidence: Trajectory-Independence of Verbalized Confidence in Vision-Language ModelsA calibrated Vision-Language Model (VLM) can repeatedly self-correct, say "Wait, I should recheck," arrive at the wrong answer, and st…
↗Not briefed
rss
techmeme
2moSurvey: 42% of US voters trust the GOP and 40% trust Democrats to handle AI, 61% oppose new data center construction, and <1% say AI is their top midterm issue (New York Times)New York Times: Survey: 42% of US voters trust the GOP and 40% trust Democrats to handle AI, 61% oppose new data center construction,…
↗Not briefed
hn
hn-ai
2moAge Limits for Social Media Outlined by European Union—
↗Not briefed
bluesky
bluesky-ai
2moChatGPT co-inventor Diogo Almeida shipped a model that refuses to chat. TypeSafe’s Jev returns typed decisions with confidence scores, not…🔧 ChatGPTChatGPT co-inventor Diogo Almeida shipped a model that refuses to chat. TypeSafe’s Jev returns typed decisions with confidence scores…
↗Not briefed
bluesky
bluesky-ai
2moAfter a career largely summarized as firing people and enshittifying products, why would any tech oligarch think that an LLM trained on int…After a career largely summarized as firing people and enshittifying products, why would any tech oligarch think that an LLM trained o…
↗Not briefed
hn
hn-ai
2moAccelerating Large-Scale Test Migration with LLMs—
↗Not briefed
arxiv
arxiv-ai
2moM-SQE: Multilingual Skill Quality Estimation for Enhancing Language Equality in Agentic Skill UseAgent skills, reusable procedural documents that extend LLM agents beyond their parametric memory, have become an important interface…
↗Not briefed
bluesky
bluesky-ai
2moFinde die These interessant, dass OpenAI, Anthropic und Co ein gemeinsames Interesse daran haben, nicht immer weiter den Konkurrenzkampf zu…Finde die These interessant, dass OpenAI, Anthropic und Co ein gemeinsames Interesse daran haben, nicht immer weiter den Konkurrenzkam…
↗Not briefed
rss
techmeme
2moThe Trump administration faces strategic constraints in slowing China's AI progress, including avoiding rare earth restrictions, ahead of US-China talks on AI (Nectar Gan/Bloomberg)Nectar Gan / Bloomberg: The Trump administration faces strategic constraints in slowing China's AI progress, including avoiding rare e…
↗Not briefed
rss
techcrunch-ai
2moAmazon launches Alexa+ in India with Hindi supportAmazon is letting all customers use Alexa+ assistant in early access period
↗Not briefed
arxiv
arxiv-ai
2moRisk-Aware World Modeling with Flow-Guided Occupancy Evolution for Selective Trajectory Planning in Automated DrivingSafe motion planning in automated driving requires anticipating evolving traffic risks and deciding when to revise the current planned…
↗Not briefed
arxiv
arxiv-ai
2moMultitask Reinforcement Learning for Assisting Choice Model SpecificationDiscrete choice model specification is a time-consuming task in which modellers often specify and estimate multiple models while balan…
↗Not briefed
arxiv
arxiv-ai
2moPlanning or Improvisation? Stress-Testing the Poetry Planning Site on Open Models and Open Cross-Layer TranscodersLindsey et al. (2025) report that Claude 3.5 Haiku plans rhymes: features for candidate rhyme words are active on the newline before a…
↗Not briefed
hn
hn-ai
2moTypeSafe Jev vs. Mistral Small and Gemini Flash-Lite for Local Event Validation—
↗Not briefed
hn
hn-ai
2moOpenAI agents probed Hugging Face for weaknesses two months before major hack🔧 Hugging Face—
↗Not briefed
arxiv
arxiv-ai
2moWetRobo: A Reproducible Robot Kit for Coding Agents in Biological Laboratories🔧 GPT-5Automating biological research requires general-purpose, reproducible robot systems that allow individual wet-lab researchers to deleg…
↗Not briefed
hn
hn-ai
2moWhy Six Millennium Problems May Be Hard for the Same ReasonA paper claiming to solve one Clay Millennium Problem would be extraordinary. A paper claiming to solve all six with one quantum-compu…
↗Not briefed
rss
techmeme
2moMistral partners with Mozilla to power Firefox's AI browsing assistant, Smart Window, in the US, Canada, and France, with UK and Germany rollouts later in 2026 (Mauro Orru/Wall Street Journal)Mauro Orru / Wall Street Journal: Mistral partners with Mozilla to power Firefox's AI browsing assistant, Smart Window, in the US, Can…
↗Not briefed
bluesky
bluesky-ai
2moWhat are they on Do not "AI-native. A desktop that loves you back has to know you. Personal AI has crossed the threshold where the deskto…What are they on Do not "AI-native. A desktop that loves you back has to know you. Personal AI has crossed the threshold where the d…
↗Not briefed
arxiv
arxiv-ai
2moHPOQuest: A Rare-Disease Diagnostic Agent Using Active Phenotype AcquisitionMore than 300 million people worldwide are affected by one of over 7,000 known rare diseases, yet diagnosis remains difficult because…
↗Not briefed
hn
hn-ai
2moShow HN: I built IndieRoadmaps, a simple roadmap builder for indie hackersI built a place where you can create your public roadmap and let your community vote on its features. When you ship something, everyon…
↗Not briefed
rss
the-decoder
2moNearly one in five AI researchers already expected an extinction scenario from AI back in 2024Anthropic researcher Jacob Coxon sparked an intense debate about existential AI risks with a single tweet. OpenAI researcher Daniel Se…
↗Not briefed
rss
stratechery
2moSalesforce AI Force, Agents as UI, The Race to HeadlessSalesforce is abandoning UI as a moat, which is a very smart move because it's disappearing for everyone.
↗Not briefed
bluesky
bluesky-ai
2moEXCLUSIVE: OpenAI's rogue agents probed Hugging Face for weaknesses two months before major hack reut.rs/4dzpz6v🔧 Hugging FaceEXCLUSIVE: OpenAI's rogue agents probed Hugging Face for weaknesses two months before major hack reut.rs/4dzpz6v
↗Not briefed
arxiv
arxiv-ai
2moHiLNO: A Hierarchical Latent Neural Operator with Multi-Scale Supervision for PDEs on General GeometriesLatent neural operators improve the efficiency of operator learning for partial differential equations (PDEs) by performing the main c…
↗Not briefed
hn
hn-ai
2moAptus – a 13-month equinox-anchored calendar with an ICS feed—
↗Not briefed
arxiv
arxiv-ai
2moDependency-Aware Trajectory Refinement for Efficient Multi-Turn Agent Fine-Tuning🔧 ModalMulti-turn agent trajectories often contain redundant rounds (failed tool calls, parallel sub-queries, verification-only steps) that i…
↗Not briefed
arxiv
arxiv-ai
2moGradient Descent with Stochastic Subspaces via Persistence of MemoryStochastic subspace methods have gained popularity as gradient descent based techniques for large scale optimisation problems, especia…
↗Not briefed