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
arxiv
arxiv-ai
7dPairing AIS with RF: Analysis of a Markovian Approach for Maritime Data FusionMonitoring maritime activities is important for safety and security, particularly for detecting illegal activities, but remains challe…
↗Not briefed
hn
hn-ai
7d2x cheaper real-time object tracking engine for video SOTA—
↗Not briefed
arxiv
arxiv-ai
7dFETTUCCINE: Fast and efficient brain-to-text decoding on mobile devicesRecent advances in speech neuroprostheses have enabled new communication avenues for people who have lost the ability to speak due to…
↗Not briefed
arxiv
arxiv-ai
7dTraining Reasoning Models in Simulation to Natively Use Real RobotsRobot-use agents (RUAs) have emerged as a promising approach to general-purpose robot control. However, strong performance has so far…
↗Not briefed
arxiv
arxiv-ai
7dBAL: Bidirectional Autoregression in Latent Space for Learning Human Movement RepresentationsLearning semantically meaningful human movement representations from 3D pose sequences is essential for human behavior modeling. Exist…
↗Not briefed
bluesky
bluesky-ai
7dI'm just gonna go and say that if you decide to give an LLM access to your wallet and the ability to use it, you absolutely deserve everyth…I'm just gonna go and say that if you decide to give an LLM access to your wallet and the ability to use it, you absolutely deserve ev…
↗Not briefed
bluesky
bluesky-ai
7dAt a workshop where someone is talking about how they're building an LLM of Putin and then going to use it to solve foreign policy problemsAt a workshop where someone is talking about how they're building an LLM of Putin and then going to use it to solve foreign policy pro…
↗Not briefed
arxiv
arxiv-ai
7dWeightBench: Evaluating Language Models as Weight-Space ProgrammersRecent work shows that large language models (LLMs) can operate directly on the parameters of other neural networks. They can read wha…
↗Not briefed
hn
hn-ai
7dShow HN: Re-ranking of the 3000 HN Comments on Jensen Huang resigns threadVibe-coded with Gemini Carbon experimenting with techniques from the &#x27;Scaling Discovery Requires Forgetting&#x27; paper: <a href=…
↗Not briefed
bluesky
bluesky-ai
7dat .3 kWh per million tokens, which is the optimistic end of the range from arxiv.org/abs/2509.20241 since we're looking at hyperscalers, w…at .3 kWh per million tokens, which is the optimistic end of the range from arxiv.org/abs/2509.20241 since we're looking at hyperscale…
↗Not briefed
bluesky
bluesky-ai
7dcopy-pasting LLM output at someone should be considered just as disrespectful as copy-pasting a screenshot of a Google search page at someo…copy-pasting LLM output at someone should be considered just as disrespectful as copy-pasting a screenshot of a Google search page at…
↗Not briefed
rss
the-decoder
7dOpenAI rebuilt AI shopping from scratch with ChatGPT Wallet🔧 ChatGPTOpenAI is taking a second shot at AI-powered shopping with ChatGPT Wallet after scrapping its first approach, Instant Checkout. Users…
↗Not briefed
rss
techmeme
7dClaude Duet review: user-friendly, sends a helpful morning brief, and spots scheduling conflicts, but can feel like a neurotic personal assistant that nags (Allison Johnson/The Verge)Allison Johnson / The Verge: Claude Duet review: user-friendly, sends a helpful morning brief, and spots scheduling conflicts, but can…
↗Not briefed
bluesky
bluesky-ai
7da) weil die LLM mit durchschnittlichem Code trainiert wurden b) sie mit statistischen Modellen arbeiten. und die Trainingsdaten (vereinfach…a) weil die LLM mit durchschnittlichem Code trainiert wurden b) sie mit statistischen Modellen arbeiten. und die Trainingsdaten (verei…
↗Not briefed
bluesky
bluesky-ai
7dWriters room isn't even trying hardWriters room isn't even trying hard
↗Not briefed
bluesky
bluesky-ai
7dA neocloud called Nscale, with commitments to lease AI training and inference bunkers to Anthropic, lost $1.7 billion last quarter and need…A neocloud called Nscale, with commitments to lease AI training and inference bunkers to Anthropic, lost $1.7 billion last quarter and…
↗Not briefed
hn
hn-ai
7dThe Sinus Rinse, Optimized—
↗Not briefed
bluesky
bluesky-ai
7ddeepseek models are powered by smuggled GPUsdeepseek models are powered by smuggled GPUs
↗Not briefed
arxiv
arxiv-ai
7dImitation Bootstrapped Multi-Agent Reinforcement LearningDemonstrations are an effective way to accelerate single-agent reinforcement learning (RL), but extending this idea to multi-agent rei…
↗Not briefed
arxiv
arxiv-ai
7dElliptic Sampling: Optimal Interpolant Schedules for Gaussians and BeyondGenerative modeling based on stochastic interpolants requires specifying two key choices: an interpolant schedule bridging the initial…
↗Not briefed
bluesky
bluesky-ai
7dThis is the Silicon Valley model---so for all those in the Canadian government, universities, schools, healthcare, and other public institu…This is the Silicon Valley model---so for all those in the Canadian government, universities, schools, healthcare, and other public in…
↗Not briefed
bluesky
bluesky-ai
7dThat would seem to be not those people’s fault or problem! Gemini also sucks, and cannot do the things that Google explicitly advertises it…That would seem to be not those people’s fault or problem! Gemini also sucks, and cannot do the things that Google explicitly advertis…
↗Not briefed
arxiv
arxiv-ai
7dWhy Finite Intelligence Works in an Infinite World: A Mathematical TheoryAn information system often has to learn before it knows what it will be asked. A finite representation can succeed without recording…
↗Not briefed
hn
hn-ai
7dOpenAI acquires Sonder co-founder's new travel tech company Odessia—
↗Not briefed
arxiv
arxiv-ai
7dModHarness-RL: Recasting Harnesses as Learnable and Evolving Programs for Agentic RLAgentic reinforcement learning (RL) increasingly relies on a harness, the scaffold around the model that decides what it sees of its o…
↗Not briefed
bluesky
bluesky-ai
7d🙌 Community pick unorouter — Storefront for unorouter.com. Next.js 16, React 19, Tailwind v4 frontend for browsing and purchasing AI mode…🙌 Community pick unorouter — Storefront for unorouter.com. Next.js 16, React 19, Tailwind v4 frontend for browsing and purchasing AI…
↗Not briefed
hn
hn-ai
7dShow HN: Port of Context – Token efficient MCP context layerHi HN, I&#x27;m working on a context layer built on OSS and open standards that lets agents search your company knowledge and use MCP…
↗Not briefed
bluesky
bluesky-ai
7dI love how when we get the post mortem of the unprecedented MS shitshow it could read 'LLM DEPENDENCE DID THIS!!!' and the sloppers will st…I love how when we get the post mortem of the unprecedented MS shitshow it could read 'LLM DEPENDENCE DID THIS!!!' and the sloppers wi…
↗Not briefed
arxiv
arxiv-ai
7dAlignment by Socialization: Learning Tradeoffs Under SelectionArtificial intelligence (AI) agents are expected to complete tasks while following rules of conduct, but these objectives can conflict…
↗Not briefed
arxiv
arxiv-ai
7dUnderstanding Contextual Bias in Agricultural CNNs Through Hidden-Neuron Concept InductionConvolutional Neural Networks (CNNs) can classify images using background-dominated patches that contain little or none of the target…
↗Not briefed
bluesky
bluesky-ai
7dGenuinely I don't think execs realize that writing the code is the easy part anyways. You need to understand what it's doing, plan for how …Genuinely I don't think execs realize that writing the code is the easy part anyways. You need to understand what it's doing, plan for…
↗Not briefed
arxiv
arxiv-ai
7dWaferTetris: Efficient Freeform Mapping on Wafer-Scale Accelerators🔧 CerebrasWafer-scale accelerators offer massive compute capacity through hundreds of thousands of processing elements (PEs) interconnected acro…
↗Not briefed
hn
hn-ai
7d6-11X speedup running Qwen3.6 on my RTX 4090, full repo.—
↗Not briefed
rss
techcrunch-ai
7dExport-banned Nvidia chips are behind China’s most powerful AI models—uh oh, Jensen!New research shows the most advanced AI models from China were trained on banned Nvidia chips.
↗Not briefed
bluesky
bluesky-ai
7dinasmuch as it had to do with AI, its because these companies want to believe they can pare everything down to the bone to increase their “…inasmuch as it had to do with AI, its because these companies want to believe they can pare everything down to the bone to increase th…
↗Not briefed
hn
hn-ai
7dShow HN: Arcstone-Airlock-Blueprint – Offline Charon/Aeneas/Coq for Rust—
↗Not briefed
arxiv
arxiv-ai
7dStreamEmbed: Real-Time Indexing from Pixels for Screenshot RetrievalScreen history can serve as searchable visual memory for people and agents, but unlike an offline corpus, it must be encoded and index…
↗Not briefed
bluesky
bluesky-ai
7dThis Week in TUXEDO OS #46-2026 www.tuxedocomputers.com/de/This-Week... #Linux #LinuxNews #LinuxDE #LinuxNewsDE #EULE #EULEde #TUXEDO #TUX…This Week in TUXEDO OS #46-2026 www.tuxedocomputers.com/de/This-Week... #Linux #LinuxNews #LinuxDE #LinuxNewsDE #EULE #EULEde #TUXEDO…
↗Not briefed
bluesky
bluesky-ai
7dThat frustration is amplified considering that the news of NVIDIA's (significant and illegal) sales to Mainland China broke in October of l…That frustration is amplified considering that the news of NVIDIA's (significant and illegal) sales to Mainland China broke in October…
↗Not briefed
arxiv
arxiv-ai
7dCan Mechanistic Interpretability Help Explain LLM Agent Failures? A Case Study on Web Element HallucinationMechanistic Interpretability (MI) has made significant progress in explaining the internal computations of Large Language Models (LLMs…
↗Not briefed
bluesky
bluesky-ai
7dI just burned the one that expires Dec 10. have a couple of more in the bag. Working on some cool new shit. These models (Fable 5.2 and G…I just burned the one that expires Dec 10. have a couple of more in the bag. Working on some cool new shit. These models (Fable 5.2…
↗Not briefed
arxiv
arxiv-ai
7dMeReader: An Offline, Privacy-Preserving, Progress-Aware AI Assistant for Narrative eBook ReadingThis paper introduces MeReader, a privacy-preserving, offline and progress-aware AI assistant designed to function as a contextual com…
↗Not briefed
bluesky
bluesky-ai
7dChatGPT told Bowen Wan to stop her epilepsy medication. And it killed her. ChatGPT devised a deadly withdrawal plan for her & now OpenAI a…🔧 ChatGPTChatGPT told Bowen Wan to stop her epilepsy medication. And it killed her. ChatGPT devised a deadly withdrawal plan for her & now Ope…
↗Not briefed
hn
hn-ai
7dKalshi hiring a 'GTA VI Researcher' to livestream the game and study its economy—
↗Not briefed
rss
gary-marcus
7dThe great Generative AI unravelingThings are starting to fall apart
↗Not briefed
arxiv
arxiv-ai
7dOn numerical issues of divergences between nearly equal distributionsBarron et al. (2026) recently identified numerical issues in the softmax and soft cross-entropy loss. We show that similar issues are…
↗Not briefed
bluesky
bluesky-ai
7dAnthropics Aktien sollen am 19. November in den Handel gehen. Der finale Preis pro Aktie wird einen Tag vorher festgelegt. #AnthropicAnthropics Aktien sollen am 19. November in den Handel gehen. Der finale Preis pro Aktie wird einen Tag vorher festgelegt. #Anthropic
↗Not briefed
bluesky
bluesky-ai
7d#Skills evaluation framework from #Nvidia to check how #skill affects #agent. Quality, security checks, more. #OpenSource Apache 2.0 #AI #L…#Skills evaluation framework from #Nvidia to check how #skill affects #agent. Quality, security checks, more. #OpenSource Apache 2.0 #…
↗Not briefed
bluesky
bluesky-ai
7d“ah ah ah, they denied it’s due to AI/agents” yeah, nobody believes it was due to a fantasy attack by a malicious agent because that only h…“ah ah ah, they denied it’s due to AI/agents” yeah, nobody believes it was due to a fantasy attack by a malicious agent because that o…
↗Not briefed
bluesky
bluesky-ai
7dI started working on @rainshare.dev because I routinely found myself not able to use my full Claude Code limit each month. I pay $200/mo, b…I started working on @rainshare.dev because I routinely found myself not able to use my full Claude Code limit each month. I pay $200/…
↗Not briefed