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
2dWhen Normal-Only Updates Are Not Enough: Test-Time Adaptation in Reconstruction-Based Time-Series Anomaly DetectionNormal operating conditions can evolve after deployment, motivating test-time adaptation (TTA) of time-series anomaly detectors using…
↗Not briefed
bluesky
bluesky-ai
2d“Advanced artificial intelligence models in simulated scenarios regularly misidentified civilians, ambulances, surrendering soldiers and a …“Advanced artificial intelligence models in simulated scenarios regularly misidentified civilians, ambulances, surrendering soldiers a…
↗Not briefed
bluesky
bluesky-ai
2dfound a new type of guy /negfound a new type of guy /neg
↗Not briefed
bluesky
bluesky-ai
2d“Assistant register” might be largely stable across models, either because patterns are inherited through distillation and LLM-judging or s…“Assistant register” might be largely stable across models, either because patterns are inherited through distillation and LLM-judging…
↗Not briefed
arxiv
arxiv-ai
2dInterpreting Timestamp Permutations in Dynamic Link PredictionTimestamp permutations are used to test whether dynamic link predictors exploit temporal information, but a performance gap can reflec…
↗Not briefed
arxiv
arxiv-ai
2dThinking Fast, Anywhere: One System-1 for the Decision Points of Any Agent WorkflowAutonomous LLM agents incur substantial cost and latency by invoking a large model (System-2) at every step, even when many intermedia…
↗Not briefed
arxiv
arxiv-ai
2dBeyond Replacement: Levenshtein Editing for Diffusion Language ModelsDiffusion language models (DLMs) offer a flexible alternative to autoregressive models, enabling any-order infilling. While most exist…
↗Not briefed
hn
hn-ai
2dMullet: The Skill for Job Security—
↗Not briefed
arxiv
arxiv-ai
2dModeling LLM Reasoning Dynamics with Structural RepresentationsLanguage-model reasoning alternates between sustained activities and local operations. We study how this temporal structure predicts c…
↗Not briefed
bluesky
bluesky-ai
2dAnthropic’s founding lore is central to its safety-first sales pitch: A handful of idealistic researchers defected from OpenAI to start a c…Anthropic’s founding lore is central to its safety-first sales pitch: A handful of idealistic researchers defected from OpenAI to star…
↗Not briefed
bluesky
bluesky-ai
2dTo me the most concerning part is LLMs don't seem to focus as much on pages 1-4 compared to human reviews. Somehow they don't care about th…To me the most concerning part is LLMs don't seem to focus as much on pages 1-4 compared to human reviews. Somehow they don't care abo…
↗Not briefed
arxiv
arxiv-ai
2dSame Rank, Different Needs: Rethinking LoRA Redistribution in Federated LearningFederated low-rank adaptation (LoRA) enables clients with different resource budgets to collaboratively fine-tune foundation models. T…
↗Not briefed
bluesky
bluesky-ai
2dWe’re demanding OpenAI create stronger safeguards to keep our kids safe. Read more: https://www.opb.org/article/2026/11/17/oregon-ag-sues…We’re demanding OpenAI create stronger safeguards to keep our kids safe. Read more: https://www.opb.org/article/2026/11/17/oregon-ag…
↗Not briefed
rss
techmeme
2dSources: a faction of Nvidia board members is pushing for Jensen Huang's return despite Chair Harvey Jones stating the board has no plans to reinstate Huang (Semafor)Semafor: Sources: a faction of Nvidia board members is pushing for Jensen Huang's return despite Chair Harvey Jones stating the board…
↗Not briefed
bluesky
bluesky-ai
2dLe journalisme factuel au Devoir, c'est de demander à un prof de finance si investir dans Anthropic est un bon placement. Pas un ostie de m…Le journalisme factuel au Devoir, c'est de demander à un prof de finance si investir dans Anthropic est un bon placement. Pas un ostie…
↗Not briefed
rss
techmeme
2dAfter two months of testimony, the US criminal trial against Huawei, accused of bank fraud and trade secret theft, is set to begin jury deliberations Wednesday (Karen Freifeld/Reuters)Karen Freifeld / Reuters: After two months of testimony, the US criminal trial against Huawei, accused of bank fraud and trade secret…
↗Not briefed
arxiv
arxiv-ai
2dReliability is a user-specific property of large language modelsPublic benchmarks support comparisons of large language model (LLM) capabilities, yet users can experience the same highly ranked mode…
↗Not briefed
arxiv
arxiv-ai
2dToward Out-of-Distribution Generalization in Neural AC Power FlowGrowing uncertainty in power system operations from renewables integration, extreme weather, and data center loads challenges the reli…
↗Not briefed
bluesky
bluesky-ai
2dReasons might include: * Most LLM-generated text is from a very small number of models * Those models *aren't* necessarily actively arm's-r…Reasons might include: * Most LLM-generated text is from a very small number of models * Those models *aren't* necessarily actively ar…
↗Not briefed
arxiv
arxiv-ai
2dAdaS-ViT: Test-Time Adaptive Routing for Spiking Vision Transformers with Ultra-High Energy EfficiencySpiking Neural Networks (SNNs) promise event-driven and energy-efficient vision, yet existing spiking vision transformers (S-ViTs) lar…
↗Not briefed
bluesky
bluesky-ai
2dNvidia is no longer shipping Hopper data center chips to China, despite winning the US government’s blessing to sell a version of that mode…Nvidia is no longer shipping Hopper data center chips to China, despite winning the US government’s blessing to sell a version of that…
↗Not briefed
bluesky
bluesky-ai
2d(said statistical analysis vibecoded with claude but since it says what "we" want to believe, we'll just ignore its sloppiness while decryi…(said statistical analysis vibecoded with claude but since it says what "we" want to believe, we'll just ignore its sloppiness while d…
↗Not briefed
bluesky
bluesky-ai
2dI was just testing out a new AI image model and this douche-bag popped up #muscles #gay #flexingI was just testing out a new AI image model and this douche-bag popped up #muscles #gay #flexing
↗Not briefed
hn
hn-ai
2dThe Wong test: how to get agents to do anything, almost anything—
↗Not briefed
arxiv
arxiv-ai
2dDo 3D Point Representations Encode Architectural Functions? Probing Supervised and Pretrained Models on Urban FacadesSemantic labels say what a building element is, not what it does: two identical-looking windows on one facade may be operable or fixed…
↗Not briefed
hn
hn-ai
2dReflections on Trusting AI: A Virus Inside an LLM—
↗Not briefed
bluesky
bluesky-ai
2d米国でAI企業のナードが乱交を楽しむ淫乱クラブが発見される。AnthropicやOpenAIの社員が「お尻叩き」「レスリング」等 https://baikoku-ch.com/%e7%b1%b3%e5%9b%bd%e3%81%a7ai%e4%bc%81%e6%a5%ad%e3…米国でAI企業のナードが乱交を楽しむ淫乱クラブが発見される。AnthropicやOpenAIの社員が「お尻叩き」「レスリング」等 https://baikoku-ch.com/%e7%b1%b3%e5%9b%bd%e3%81%a7ai%e4%bc%81%e6%a5%…
↗Not briefed
bluesky
bluesky-ai
2dI’m sure it’s been discussed but wondering if the selling is just sucking liquidity to be put into the Anthropic IPO? $ant. Seem to remem…I’m sure it’s been discussed but wondering if the selling is just sucking liquidity to be put into the Anthropic IPO? $ant. Seem to…
↗Not briefed
arxiv
arxiv-ai
2dHypoLoop: Revising Weight-Editing Plans from Scientific FeedbackScientific experiments continually reveal prediction errors. Effective updates depend on which parameters change and which errors driv…
↗Not briefed
hn
hn-ai
2dFree Quant Interview Practices—
↗Not briefed
arxiv
arxiv-ai
2dLearning through Discernment: when a small language model learns from user feedback on the job, and when it decaysSmall language models on devices could keep improving by learning, overnight, from their own daily work. Without a true outcome, howev…
↗Not briefed
bluesky
bluesky-ai
2d"ai generated food pictures look disgusting" no thats real. those are real photos taken with cameras. that's just claude meat. it's all cla…"ai generated food pictures look disgusting" no thats real. those are real photos taken with cameras. that's just claude meat. it's al…
↗Not briefed
hn
hn-ai
2dShow HN: An ASCII cell aquarium for your terminalI made a virtual microscope slide for your terminal (inspired by asciiquarium and cbonsai). Cells divide, immune cells chase bacteria,…
↗Not briefed
hn
hn-ai
2dEvolving Infra: Agent-Adaptive LLM Serving at Scale—
↗Not briefed
bluesky
bluesky-ai
2dIn case anyone on my feed is unaware, "AI Detectors" are just another LLM, their output is meaningless slop. Don't use them, let alone rely…In case anyone on my feed is unaware, "AI Detectors" are just another LLM, their output is meaningless slop. Don't use them, let alone…
↗Not briefed
bluesky
bluesky-ai
2dAsia stocks set to snap losing streak on Nvidia beat as Anthropic looms reut.rs/4AEkgMmAsia stocks set to snap losing streak on Nvidia beat as Anthropic looms reut.rs/4AEkgMm
↗Not briefed
arxiv
arxiv-ai
2dRational Ensemble Logic for Cohort Discovery with Real-World Electrophysiology Data🔧 ModalBiomedical knowledge representation should make the passage from observations to conclusions transparent, rigorous, understandable, ex…
↗Not briefed
bluesky
bluesky-ai
2dThe reasoning the LLM provides makes no sense. The correct values the LLM supposedly calculated for Wisconsin and Wyoming are already in th…The reasoning the LLM provides makes no sense. The correct values the LLM supposedly calculated for Wisconsin and Wyoming are already…
↗Not briefed
bluesky
bluesky-ai
2dBrainstorming has its own room now. The conversation on one side, and beside it the ideas, the plan as it grows, and the ones you turned do…Brainstorming has its own room now. The conversation on one side, and beside it the ideas, the plan as it grows, and the ones you turn…
↗Not briefed
bluesky
bluesky-ai
2dBrainstorming has its own room now. The conversation on one side, and beside it the ideas, the plan as it grows, and the ones you turned do…Brainstorming has its own room now. The conversation on one side, and beside it the ideas, the plan as it grows, and the ones you turn…
↗Not briefed
bluesky
bluesky-ai
2dOregon sues OpenAI for $2 billion alleging ChatGPT puts kids in harm’s way 🎁 link good for next 7 days. www.oregonlive.com/business/202.…🔧 ChatGPTOregon sues OpenAI for $2 billion alleging ChatGPT puts kids in harm’s way 🎁 link good for next 7 days. www.oregonlive.com/business…
↗Not briefed
bluesky
bluesky-ai
2dmusically.com/2026/10/23/s...musically.com/2026/10/23/s...
↗Not briefed
hn
hn-ai
2dShow HN: I made a site showcasing The Elm Architecture in many other languagesI am a huge fan of Elm, but the language is hard to use at work. Luckily, there are many implementations of TEA in other languages, so…
↗Not briefed
bluesky
bluesky-ai
2dThis song shouldn't slap as much as it does - hilarity and truth. For anyone who works in tech, be warned! It'll make you laugh and cry in …This song shouldn't slap as much as it does - hilarity and truth. For anyone who works in tech, be warned! It'll make you laugh and cr…
↗Not briefed
arxiv
arxiv-ai
2dLocalizing Text Anonymization for Trustworthy AI: Extending RAT-Bench to Malaysian Microdata and PIIAs governments and organizations adopt foundation models for public services and workplace productivity, sensitive citizen, employee,…
↗Not briefed
bluesky
bluesky-ai
2dMy parents rewatching Endgame in the living room tell me why Tony Stark making an LLM invent time travel for him 😭😭😭 not the preemptive …My parents rewatching Endgame in the living room tell me why Tony Stark making an LLM invent time travel for him 😭😭😭 not the preemp…
↗Not briefed
rss
techmeme
2dAnalysis: San Francisco-based AI startup insiders hold a combined $1.2T in equity; at Anthropic's IPO midpoint, the average employee's equity is valued at $72M (Bloomberg)Bloomberg: Analysis: San Francisco-based AI startup insiders hold a combined $1.2T in equity; at Anthropic's IPO midpoint, the average…
↗Not briefed
rss
bing-news-ai
2dNot Micron. Not Broadcom. This Artificial Intelligence (AI) Stock Will Skyrocket in December After the Latest Sell-OffOptical networking is emerging as the next bottleneck in AI infrastructure, and Ciena is one of the best ways for investors to capital…
↗Not briefed
hn
hn-ai
2dOpenAI is starting to panic [video]—
↗Not briefed
bluesky
bluesky-ai
2dUX problems in 1992: can't find the feature i need because there's too many buttons on screen UX problems in 2006: can't find the feature …UX problems in 1992: can't find the feature i need because there's too many buttons on screen UX problems in 2006: can't find the fea…
↗Not briefed