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
3dRAG in a Trenchcoat: When Minimal Memory Is Enough for Agentic Systems, and When It Isn'tAgentic memory research has accumulated extraction, supersession, reflection, and graph synthesis on top of retrieval-augmented genera…
↗Not briefed
arxiv
arxiv-ai
3dCorrectCast: Selective Multimodal Context for Scale-Aware Residual Correction in Precipitation NowcastingPrecipitation nowcasting models suffer from errors in intensity, location, and spatial structure. Addressing these errors by retrainin…
↗Not briefed
arxiv
arxiv-ai
3dThe Only Edge Was Memorized in Training: LLM Trading Agents Around Earnings and Bad NewsLarge language model (LLM) trading agents are tested on news, but news reaches stock prices within minutes, and a benchmark that repla…
↗Not briefed
bluesky
bluesky-ai
3dDo it! I’m tired of big tech acting like bullies. #Facebook refuses to remove racist comments on its platform. Don’t even get me started on…Do it! I’m tired of big tech acting like bullies. #Facebook refuses to remove racist comments on its platform. Don’t even get me start…
↗Not briefed
bluesky
bluesky-ai
3dState of CA sued OpenAI… www.cnbc.com/2026/11/16/o...State of CA sued OpenAI… www.cnbc.com/2026/11/16/o...
↗Not briefed
hn
hn-ai
3dTypeScript apps running natively in embedded devices—
↗Not briefed
arxiv
arxiv-ai
3dReconfiguring Procedural Knowledge for Compositional Robot Skill AdaptationGeneral-purpose behavior models increasingly internalize broad procedural knowledge during training, making their adaptation to new go…
↗Not briefed
arxiv
arxiv-ai
3dTowards Optimism-Pessimism Trade-off in Model-based Offline-to-Online Reinforcement LearningModel-based offline-to-online reinforcement learning (RL) enables sample-efficient adaptation by leveraging offline pre-training for o…
↗Not briefed
arxiv
arxiv-ai
3dReadie: Low-Latency Code Sandboxes for AI AgentsFrom quantitative research to personal assistance, AI agents act by iteratively generating and executing code. To decouple the agent h…
↗Not briefed
bluesky
bluesky-ai
3dAnthropic tells investors Claude Mythos Preview got its AI running on AMD GPUs in ~3 days with "basically no human intervention", vs. the u…Anthropic tells investors Claude Mythos Preview got its AI running on AMD GPUs in ~3 days with "basically no human intervention", vs.…
↗Not briefed
bluesky
bluesky-ai
3dI'm not sure DRAM is going to help OpenAI with all those lawsuitsI'm not sure DRAM is going to help OpenAI with all those lawsuits
↗Not briefed
rss
techmeme
3dAnthropic tells investors Claude Mythos Preview got its AI running on AMD GPUs in ~3 days with "basically no human intervention", vs. the usual 6-12 months (Shane Downing/Tom's Hardware)Shane Downing / Tom's Hardware: Anthropic tells investors Claude Mythos Preview got its AI running on AMD GPUs in ~3 days with “basica…
↗Not briefed
bluesky
bluesky-ai
3da reason so many of us have dozens of books that we will not get paid for is because the class was limited to those who had filed copyright…a reason so many of us have dozens of books that we will not get paid for is because the class was limited to those who had filed copy…
↗Not briefed
arxiv
arxiv-ai
3dNot All Evidence Is Admissible: Evidence Governance for Tool-Augmented Medical AgentsTool-augmented medical agents treat whatever their tools return as evidence. We show that this default is costly: medical tools fail s…
↗Not briefed
bluesky
bluesky-ai
3da reminder that the company handling the Anthropic settlement is not Anthropic, and Anthropic doesn't get to keep any money we don't get & …a reminder that the company handling the Anthropic settlement is not Anthropic, and Anthropic doesn't get to keep any money we don't g…
↗Not briefed
hn
hn-ai
3dA game cartridge peripheral for Dreamcast—
↗Not briefed
bluesky
bluesky-ai
3dI don't think that is possible. Anthropic has given the money to a third party at this point. I think it's really just that they did what …I don't think that is possible. Anthropic has given the money to a third party at this point. I think it's really just that they did…
↗Not briefed
bluesky
bluesky-ai
3dshare.google/7p084uqzYT4b...share.google/7p084uqzYT4b...
↗Not briefed
arxiv
arxiv-ai
3dWhat RL on Expert-Built Tasks Teaches Nemotron 3.5 SuperWe post-train NVIDIA's Nemotron 3.5 Super (120B total, 12B active parameters) with reinforcement learning (RL) on five expert-built ta…
↗Not briefed
arxiv
arxiv-ai
3dCentroid Spectral Coupling: Implicit Alignment and Representation Diversity in Self-Supervised LearningSelf-Supervised Learning (SSL) aims to align augmented views while preserving representation diversity and preventing collapse. Existi…
↗Not briefed
hn
hn-ai
3dIPTV USA Subscription – IPTV USA – 50k Channels and 4K Streaming—
↗Not briefed
rss
techmeme
3dFiling: Leopold Aschenbrenner's hedge fund Situational Awareness, after nearly blowing up, loaded back up on AI stocks at the end of Q3; all have since plunged (Sean Conlon/CNBC)Sean Conlon / CNBC: Filing: Leopold Aschenbrenner's hedge fund Situational Awareness, after nearly blowing up, loaded back up on AI st…
↗Not briefed
bluesky
bluesky-ai
3dLLMについては、絵や動画の生成に関してはノー、研究及びプログラム開発には有用、検索には過剰、環境負荷を考えると使用範囲は制限されるべき、世界への影響はevil、使いようによってはとても助かる人がいる、技術としては面白いが人間社会の倫理欠如は嫌い、ビッグテックも全体的に嫌い で…LLMについては、絵や動画の生成に関してはノー、研究及びプログラム開発には有用、検索には過剰、環境負荷を考えると使用範囲は制限されるべき、世界への影響はevil、使いようによってはとても助かる人がいる、技術としては面白いが人間社会の倫理欠如は嫌い、ビッグテックも全体的…
↗Not briefed
arxiv
arxiv-ai
3dEgoHandWAM: Probing the Role of Hand Motion in Egocentric World Action ModelingHow much does hand motion tell us about what a first-person camera will see next? EgoHandWAM forecasts three-dimensional wrist and fin…
↗Not briefed
bluesky
bluesky-ai
3d“DeepSeek is particularly revealing because it drew together three common narratives in AI debates: sovereignty, scale, and openness. Each …“DeepSeek is particularly revealing because it drew together three common narratives in AI debates: sovereignty, scale, and openness.…
↗Not briefed
arxiv
arxiv-ai
3dLong-Context Generation Is a Sampling Problem (When You Turn Up the Heat)Open-weight language models are widely observed to get worse the longer they generate. Often, attempts to fix this involve more traini…
↗Not briefed
arxiv
arxiv-ai
3dWhen does deep reinforcement learning pay off in HVAC supervisory control? Four small Modelica case studies trained directly on FMUsDeep reinforcement learning (DRL) is often proposed for the supervisory control of heating, ventilation and air-conditioning (HVAC) sy…
↗Not briefed
arxiv
arxiv-ai
3dSynthetic Crystal Data Generation for Diffusion Pretraining in Material Property PredictionDiffusion-based pretraining has emerged as a powerful strategy for learning transferable representations in graph neural networks for…
↗Not briefed
arxiv
arxiv-ai
3dRoboTeam: Benchmarking Vision-Based Multi-Agent Control of Many Robot Arms in Real TimeMultimodal agents can now control robot arms by calling tools and writing code. Existing benchmarks fall short in three ways: most tes…
↗Not briefed
bluesky
bluesky-ai
3dまーた怪しいサービスが出てる。 「テキスト楽譜ループ再生」 cocktaillog.com/ja/music/score 以下サイトより引用 "「テキスト楽譜ループ再生」は、テキストで書いた曲から楽譜(五線譜・TAB譜)を生成し、その場で再生できる無料のWebアプリです" …🔧 ChatGPTまーた怪しいサービスが出てる。 「テキスト楽譜ループ再生」 cocktaillog.com/ja/music/score 以下サイトより引用 "「テキスト楽譜ループ再生」は、テキストで書いた曲から楽譜(五線譜・TAB譜)を生成し、その場で再生できる無料のWebアプリ…
↗Not briefed
arxiv
arxiv-ai
3dBig Brains, Tiny Board: Pupavision, an Agentic Physical AI Instrument for Monitoring Drosophila Pupariation and Lifecycle TransitionsMany biology experiments are limited by the availability and ease of use of instruments that monitor and record them. For example, the…
↗Not briefed
bluesky
bluesky-ai
3dOne super cool fun fact about the Anthropic settlement is that they stole most (all?) of my books but I only get compensation for two that …One super cool fun fact about the Anthropic settlement is that they stole most (all?) of my books but I only get compensation for two…
↗Not briefed
rss
techmeme
3dSources: Jensen Huang's daughter remains in a senior Nvidia role; before Jensen Huang resigned, she was his de facto chief of staff with considerable authority (New York Times)New York Times: Sources: Jensen Huang's daughter remains in a senior Nvidia role; before Jensen Huang resigned, she was his de facto c…
↗Not briefed
hn
hn-ai
3dEverything You Wanted to Know About LLM Post-Training [video]—
↗Not briefed
hn
hn-ai
3dShow HN: Pulsar – cross-harness agent to agent communication and messaging—
↗Not briefed
arxiv
arxiv-ai
3dReMoE: An Interpretable Mixture-of-Experts Framework for Human Mobility Recovery Forecasting Across DisastersHuman mobility often declines sharply following disasters and subsequently exhibits complex recovery dynamics. Effective forecasting o…
↗Not briefed
hn
hn-ai
3dShow HN: Find out how cooked your webapp isHey everyone! If your startup has a software component you might want to run a quick scan on GenClass and see what score you get (it’s…
↗Not briefed
bluesky
bluesky-ai
3d“OpenAI, a leading artificial intelligence business, misled and endangered the public and users of its services by knowingly developing adv…“OpenAI, a leading artificial intelligence business, misled and endangered the public and users of its services by knowingly developin…
↗Not briefed
arxiv
arxiv-ai
3dLoCaR: Localized Canvas Retrieval With Frozen Vision-Language ModelWhole image embeddings can under represent small objects and local attributes that distinguish relevant images for localized queries.…
↗Not briefed
bluesky
bluesky-ai
3dA reminder that if the corporations that pirated works to train LLM faced the same penalties that humans do, they would no longer exist.A reminder that if the corporations that pirated works to train LLM faced the same penalties that humans do, they would no longer exis…
↗Not briefed
arxiv
arxiv-ai
3dBandit Online Submodular Minimisation Using Zeroth-Order OraclesWe consider online submodular minimisation under bandit feedback, where at each round a decision maker plays a subset and observes onl…
↗Not briefed
arxiv
arxiv-ai
3dTempQuant: Learning Adaptive Temperatures for Quantized Language ModelsLarge language models are increasingly deployed at low precision, yet quantization can degrade their performance. We study how quantiz…
↗Not briefed
hn
hn-ai
3dShow HN: Toolbelt – Free developer utilities that run in the browser—
↗Not briefed
arxiv
arxiv-ai
3dWhat MoE Routing Controls: Causal Interventions on Expert TrajectoriesIn mixture-of-experts (MoE) language models, a token is routed to a sequence of experts across layers. We investigate how forcing toke…
↗Not briefed
bluesky
bluesky-ai
3dI received a message from Anthropic saying that I will NOT receive a payment in the first tranche despite sending in countless responses to…I received a message from Anthropic saying that I will NOT receive a payment in the first tranche despite sending in countless respons…
↗Not briefed
arxiv
arxiv-ai
3dTeleChat to Xing4: Advancing Frontier Agentic Models on Domestic Chips and FrameworksWe present Xing 4, a series of large language models that extends the TeleChat lineage through coordinated advances in model architect…
↗Not briefed
arxiv
arxiv-ai
3dRegimeCouncil: Paired-Evidence Routing For Memory Conflict Governance In LLM AgentsLanguage agents rely on long-term memory to personalize responses across extended interactions, yet stored facts quietly go stale as t…
↗Not briefed
hn
hn-ai
3dShow HN: Nine is multiplayer sudoku across all Apple devicesI built this universal app as a challenge to myself to learn how to build with the full spectrum of Apple ecosystem capabilities.<p>Ni…
↗Not briefed
hn
hn-ai
3dJitpipe–Sample 25 stdin rows,synthesize regex/Awk filter,stream the rest in Rust—
↗Not briefed
arxiv
arxiv-ai
3dVTT: Reading Vehicle Size and Powertrain Type from Motion Trajectory Across Unseen Car ModelsVehicle size and powertrain information can enrich trajectory data for traffic analysis and simulation. We present VTT, a dataset and…
↗Not briefed