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
9dWhat Can a VLA Learn from the Agent That Calls It? Task-Level Experience Distillation for Embodied NavigationNavigation agents use a large language model (LLM) planner to call a vision-language-action (VLA) model. Each episode needs the LLM, a…
↗Not briefed
hn
hn-ai
9dShow HN: Rusty Valve – Chemical plant valve benchmark where ACT scores 0/36—
↗Not briefed
arxiv
arxiv-ai
9dOptimal feature selection for firewall log analysis using Machine learning and Hybrid Metaheuristic algorithmsFirewall log classification is important to monitor network traffic. Machine learning classifiers using network-related features have…
↗Not briefed
arxiv
arxiv-ai
9dATune: Unsupervised Photo Retouching Through Image-Editor ControlsPhoto retouching models are trained on pairs of original and retouched photographs, usually edited by experts. Expert edits are costly…
↗Not briefed
bluesky
bluesky-ai
9dPre-LLM but I remain annoyed by a research plan for "the interlocking pillars of democracy". What pillars interlock?Pre-LLM but I remain annoyed by a research plan for "the interlocking pillars of democracy". What pillars interlock?
↗Not briefed
hn
hn-ai
9dShow HN: App to help parents find, plan, and review outings with their kidsBackground: My wife and I have 4 boys 4 and under, so we have to get them out of the house consistently to stay sane. Finding good pla…
↗Not briefed
bluesky
bluesky-ai
9dAlso, pretty sure it's easier to stay blissfully ignorant of the presence of LLM code in a game than it is to ignore poisoning and projecti…Also, pretty sure it's easier to stay blissfully ignorant of the presence of LLM code in a game than it is to ignore poisoning and pro…
↗Not briefed
rss
techmeme
9dSources: the Commerce Department is finalizing a semiconductor control update that would prevent remote access and chip smuggling into China via data centers (Leo Schwartz/The Information)Leo Schwartz / The Information: Sources: the Commerce Department is finalizing a semiconductor control update that would prevent remot…
↗Not briefed
hn
hn-ai
9dWhen LIMIT 10 Means 500—
↗Not briefed
arxiv
arxiv-ai
9dIndic-Task Gains and Regressions after Reinforcement LearningWe run reinforcement learning (RL) with GRPO across four environments, starting from a supervised fine-tuned 30B-A3B mixture-of-expert…
↗Not briefed
arxiv
arxiv-ai
9dWe Are What We Remember: Mental-State-Driven Memory for Long-Term Character DialogueShared history in human dialogue is not merely a repository of facts, but the foundation of evolving relationships and mutual understa…
↗Not briefed
bluesky
bluesky-ai
9dQwen3.8 Max Prime vs Claude Sonnet 5.5: Which LLM API actually protects your SaaS gross margin at 10,000 DAU? Here is the live 2026 token …Qwen3.8 Max Prime vs Claude Sonnet 5.5: Which LLM API actually protects your SaaS gross margin at 10,000 DAU? Here is the live 2026 t…
↗Not briefed
bluesky
bluesky-ai
9dAnd for some reason you're hand-waving away all of these industry insiders, including the CEO of Anthropic, warning about these risks. The …And for some reason you're hand-waving away all of these industry insiders, including the CEO of Anthropic, warning about these risks.…
↗Not briefed
hn
hn-ai
9dShow HN: Eyes-free spoken N-Back where you choose the words (single HTML file)—
↗Not briefed
bluesky
bluesky-ai
9dthe track changes functionality feels like an afterthought, it barely works. the keyboard shortcuts are revolting. it has no native dark mo…the track changes functionality feels like an afterthought, it barely works. the keyboard shortcuts are revolting. it has no native da…
↗Not briefed
rss
techcrunch-ai
9dWho could replace Jensen Huang? A highly unscientific list of possible successorsNvidia has no succession plan and a CEO in hot water, so we’ve drafted a shortlist.
↗Not briefed
bluesky
bluesky-ai
9dMy Chinese handler losing her mind when she realizes everything she's been sleeping with me to get, DeepSeek can now just get on its own, t…My Chinese handler losing her mind when she realizes everything she's been sleeping with me to get, DeepSeek can now just get on its o…
↗Not briefed
arxiv
arxiv-ai
9dGC-MoE on SM120: Overlapping Expert-Parallel Token Dispatch with Warp-Level FP8xFP4 Block-Scaled MMA on Consumer BlackwellExpert-parallel Mixture-of-Experts decoding is latency bound: each layer serializes a token exchange, a grouped GEMM over routed exper…
↗Not briefed
arxiv
arxiv-ai
9dGuided Generation of Linux Kernel Semantic Patches with Small Local Language Models🔧 Gemma 3When a Linux kernel API changes, hundreds of drivers are converted by hand over months, and the earlier conversions are reviewed examp…
↗Not briefed
bluesky
bluesky-ai
9dgood! it is the second-most active model, after the local Qwens (which were running for over a week) and GPT-6 Luna (which is a hell of a w…good! it is the second-most active model, after the local Qwens (which were running for over a week) and GPT-6 Luna (which is a hell o…
↗Not briefed
hn
hn-ai
9dShow HN: Paper Radar – daily ArXiv triage, now runs on Claude Haiku too—
↗Not briefed
arxiv
arxiv-ai
9dOne Fuzzer, Many Reasoners: Adaptive Neuro-Symbolic Fuzzing with Reliability-Calibrated and Context-Gated GuidanceNeural-symbolic AI (NSAI) combines learned perception with programs, constraints, and compositional reasoning, but remains difficult t…
↗Not briefed
bluesky
bluesky-ai
9dIf OpenAI is at the center of an international AI arms race—which is one of the narratives they push—then they either need better opsec or …If OpenAI is at the center of an international AI arms race—which is one of the narratives they push—then they either need better opse…
↗Not briefed
bluesky
bluesky-ai
9dwe already have ~5,400 data centers. ten times more than China. we still have a winner-take-all competition between five frontier labs…we already have ~5,400 data centers. ten times more than China. we still have a winner-take-all competition between five frontier…
↗Not briefed
arxiv
arxiv-ai
9dTurning Multi-Agent Communication into Token-Level Credit with On Policy Self DistillationReinforcement learning improves problem solving in large language model multi-agent systems. Existing approaches overlook communicatio…
↗Not briefed
hn
hn-ai
9dShow HN: Ramasite, a Self-Contained Laravel/Filament/SQLite CMS—
↗Not briefed
bluesky
bluesky-ai
9dIn a way, the complete nothingness of it reflects how AI usage became commonplace, so that's nice. The absolute lack of creativity and regu…In a way, the complete nothingness of it reflects how AI usage became commonplace, so that's nice. The absolute lack of creativity and…
↗Not briefed
arxiv
arxiv-ai
9dAn Extensive Evaluation of Machine Learning and Deep Learning Approaches for Sentiment Analysis: Insights from Amazon Customer ReviewsThe rapid growth of user-generated text on online platforms creates opportunities for understanding customer opinion and supporting da…
↗Not briefed
bluesky
bluesky-ai
9dOf course it goes rogue. Spock: most illogical. Of all people, he [Daystrom, the AI’s creator] should have known how it would perform. C…Of course it goes rogue. Spock: most illogical. Of all people, he [Daystrom, the AI’s creator] should have known how it would perform…
↗Not briefed
bluesky
bluesky-ai
9dDeepSeek V4 Pro 0813 vs Qwen3.8 Max Prime: Which LLM API actually protects your SaaS gross margin at 10,000 DAU? Here is the live 2026 tok…DeepSeek V4 Pro 0813 vs Qwen3.8 Max Prime: Which LLM API actually protects your SaaS gross margin at 10,000 DAU? Here is the live 202…
↗Not briefed
rss
techmeme
9dSources: some Nvidia board members favor naming CFO Colette Kress interim CEO; an administration official says the board briefed the White House on its meeting (Semafor)Semafor: Sources: some Nvidia board members favor naming CFO Colette Kress interim CEO; an administration official says the board brie…
↗Not briefed
bluesky
bluesky-ai
9d"It is worth noting that of the 290 OpenAI theorems which come with Lean code, around 38 of them have the property that the Lean code *does…"It is worth noting that of the 290 OpenAI theorems which come with Lean code, around 38 of them have the property that the Lean code…
↗Not briefed
bluesky
bluesky-ai
9dI am beginning to think it is because OpenAI employs a lot of very stupid people who are convinced they are not very stupidI am beginning to think it is because OpenAI employs a lot of very stupid people who are convinced they are not very stupid
↗Not briefed
rss
techmeme
9dLetter: Sens. Elizabeth Warren and Chris Coons ask the Trump administration to preserve all documents tied to Nvidia's chip exports to China and Huang's affair (Wall Street Journal)Wall Street Journal: Letter: Sens. Elizabeth Warren and Chris Coons ask the Trump administration to preserve all documents tied to Nvi…
↗Not briefed
bluesky
bluesky-ai
9dFor example, a carbon tax would change the calculations that drive decisions like this: finance.yahoo.com/technology/a... - put that money …For example, a carbon tax would change the calculations that drive decisions like this: finance.yahoo.com/technology/a... - put that m…
↗Not briefed
hn
hn-ai
9dCompromised GitHub accounts push CI secret stealer to 61k repositories—
↗Not briefed
hn
hn-ai
9dLessons from 59 Linux Kernel Bug Reports—
↗Not briefed
arxiv
arxiv-ai
9dWhen can an interpretability claim be trusted? Stability and certification of feature dimensionality, ridge leverage and the feature recovery rateInterpretability tools increasingly support safety claims, such as that a sparse dictionary has recovered most of a model's features.…
↗Not briefed
bluesky
bluesky-ai
9dHi! I've been slowly learning the intricacies of a new image model. Here's some of the images I made! #AIart #aiイラスト #AIHi! I've been slowly learning the intricacies of a new image model. Here's some of the images I made! #AIart #aiイラスト #AI
↗Not briefed
arxiv
arxiv-ai
9dFlow-CAD: Flow Matching for Contextual Anomaly Detection and Explanation in Tabular DataContextual anomaly detection in tabular data aims to identify objects whose behavior is unexpected under their contextual conditions.…
↗Not briefed
arxiv
arxiv-ai
9dData-Efficient LLM Safety Alignment Guided by Interpretable Features🔧 ElicitExisting large language models typically rely on massive fine-tuning datasets for safety alignment, which results in higher costs for…
↗Not briefed
bluesky
bluesky-ai
9di do not like the devs for their usage of an LLM. i respect their frustration but not the solution they chose, and believe that it arose si…i do not like the devs for their usage of an LLM. i respect their frustration but not the solution they chose, and believe that it aro…
↗Not briefed
hn
hn-ai
9dSubscription Sharing usage limit on app blocking OpenAI model usage after reset—
↗Not briefed
arxiv
arxiv-ai
9dSame-State Pairs Make One JEPA Proof of the Platonic Representation Hypothesis VacuousIn a Joint-Embedding Predictive Architecture (JEPA), the JEPA proof of the Platonic Representation Hypothesis (PRH) assumes that every…
↗Not briefed
bluesky
bluesky-ai
9dWorker protections from being fired, demoted, or disciplined for refusing to use LLM tools, and protections for those who push back against…Worker protections from being fired, demoted, or disciplined for refusing to use LLM tools, and protections for those who push back ag…
↗Not briefed
hn
hn-ai
9dCan commercial stage-gates keep enterprise AI pilots out of procurement stalls?—
↗Not briefed
hn
hn-ai
9dA Meta Compiler: A framework for building consistent but flexible code—
↗Not briefed
arxiv
arxiv-ai
9dStack more levels: how to get general and human-like Mario playingSuper Mario Bros. is widely used in game AI and procedural content generation (PCG), yet it remains unclear whether a model-free agent…
↗Not briefed
bluesky
bluesky-ai
9dHave brief chat with Claude, encounter the AI sycophancy within 3 turns. Feel smart for 2 minutes, then feel incredibly grossed out. Clos…Have brief chat with Claude, encounter the AI sycophancy within 3 turns. Feel smart for 2 minutes, then feel incredibly grossed out.…
↗Not briefed
bluesky
bluesky-ai
9dWhat a day. Now OpenAI top security officials got punk’d by Russian dudes.What a day. Now OpenAI top security officials got punk’d by Russian dudes.
↗Not briefed