News
The raw AI news stream — everything we ingest across RSS, Hacker News, arXiv, GitHub, Bluesky, YouTube and more, newest first.
| Source | Age | Title | Summary | Status |
|---|---|---|---|---|
arxiv arxiv-ai | 1mo | An Uncertainty-Guided Digital Twin Framework for Online Adaptive Proton Therapy in Head and Neck Cancer: A Feasibility Study | Objective: Head and neck (HN) proton therapy spans six to seven weeks of anatomical change, while offline replanning takes about a wee… | ↗Not briefed |
arxiv arxiv-ai | 1mo | RouteRec: Behavior-Guided Sparse Routing for Sequential Recommendation | Sessionized interaction histories contain behavioral patterns that can improve sequential recommendation. However, existing models pro… | ↗Not briefed |
arxiv arxiv-ai | 1mo | C-STRIDE: An Observation-Driven AI Digital Twin for Predicting Basin-Wide Flood Fields from Sparse Stream-Gauge Histories | Emergency managers need to know where floodwater is, how deep it is, and how it will change over the coming hours across an entire riv… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Here’s @edzitron.com going into detail on @reuters.com’s big scoop on anthropics’s BAD 2025 numbers podcasts.apple.com/us/podcast/b... | Here’s @edzitron.com going into detail on @reuters.com’s big scoop on anthropics’s BAD 2025 numbers podcasts.apple.com/us/podcast/b... | ↗Not briefed |
hn hn-ai | 1mo | Show HN: Runtape – counterfactual debugging and regression tests for AI agents | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | The Invisible Language Tax: Token Premiums of French and Regional Languages in 2026 LLM Tokenizers, and a French-Optimized Prototype🔧 Gemma 3 | LLM services are billed per token and context windows are measured in tokens, yet the number of tokens needed for the same content var… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Not all solutions are created equal: An analytical dissociation of functional and representational similarity in deep linear neural networks | A foundational principle of connectionism is that perception, action, and cognition emerge from parallel computations among simple, in… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Settle: Learning When to Stop Reasoning | Reasoning models often continue generating after their answers have settled. Settle learns when to stop from answer stability in compl… | ↗Not briefed |
arxiv arxiv-ai | 1mo | When Clipping Reverses Correction: Failure Dynamics of Pointwise Forward-KL On-Policy Self-Distillation | On-policy self-distillation (OPSD) trains a student on its own generated responses using feedback from the same model conditioned on p… | ↗Not briefed |
hn hn-ai | 1mo | Efficiently Joining Large Relations on Multi-GPU Systems [pdf] | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | apparently the openai dots dont have a persistent disk lol | apparently the openai dots dont have a persistent disk lol | ↗Not briefed |
bluesky bluesky-ai | 1mo | could a non-human use an LLM? | could a non-human use an LLM? | ↗Not briefed |
bluesky bluesky-ai | 1mo | Google (“Lake America”) not yet calling them SI overviews. OpenAI legal probably pulling an all nighter for name change to OpenSI tomorrow.… | Google (“Lake America”) not yet calling them SI overviews. OpenAI legal probably pulling an all nighter for name change to OpenSI tomo… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Paperclipprobleem: waarom AI-bedrijven bang zijn dat AI de wereld overneemt | Paperclipprobleem: waarom AI-bedrijven bang zijn dat AI de wereld overneemt | ↗Not briefed |
bluesky bluesky-ai | 1mo | “The OpenAI co-founder Sam Altman met with Kennedy this week and told him that HHS uses generative AI – Kennedy used the term “superintelli… | “The OpenAI co-founder Sam Altman met with Kennedy this week and told him that HHS uses generative AI – Kennedy used the term “superin… | ↗Not briefed |
bluesky bluesky-ai | 1mo | AI executives sign accord to 'self-police' after White House meeting | AI executives sign accord to 'self-police' after White House meeting | ↗Not briefed |
arxiv arxiv-ai | 1mo | Anchoring Adversarial Trajectories to Data Manifolds: A Bilevel Transfer Optimization Framework | A key bottleneck in adversarial transfer is a trajectory-level geometric disconnect: ambient gradients often drift away from the intri… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Smaller Models, Better Rejects: Preference Distillation Scaling | Preference distillation typically treats a teacher response as preferred and the student's own response as rejected. This assumes that… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Sparse-WAM: Accelerating World Action Models via Action-Guided Sparse Imagination | World-action models (WAMs) leverage pretrained video models to improve generalization in robot control by jointly predicting future vi… | ↗Not briefed |
bluesky bluesky-ai | 1mo | DeepSeek says it has partnered with Huawei to develop programming tools for Huawei's Ascend chips, including TileLang, an open-source CUDA … | DeepSeek says it has partnered with Huawei to develop programming tools for Huawei's Ascend chips, including TileLang, an open-source… | ↗Not briefed |
rss techmeme | 1mo | DeepSeek says it has partnered with Huawei to develop programming tools for Huawei's Ascend chips, including TileLang, an open-source CUDA alternative (Reuters) | Reuters: DeepSeek says it has partnered with Huawei to develop programming tools for Huawei's Ascend chips, including TileLang, an ope… | ↗Not briefed |
hn hn-ai | 1mo | Are you getting a weird response to the query: 'play Minecraft' on America.gov? | Are you seeing this?<p><i>I see the constituent you mean.<p>((insert legal name here, as it appears on the Social Security card))?<p>Y… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Hier ein paar zuletzt häufig geteilte #News: Anthropic: Nettoverlust von 42 Milliarden US-Dollar allein 2025 Fast so hoch wie Berlins Fer… | Hier ein paar zuletzt häufig geteilte #News: Anthropic: Nettoverlust von 42 Milliarden US-Dollar allein 2025 Fast so hoch wie Berlin… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Alert that from now on I will be terming OpenAI a super intelligence firm. Also, my kids are super intelligent. They will shortly be going … | Alert that from now on I will be terming OpenAI a super intelligence firm. Also, my kids are super intelligent. They will shortly be g… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Approval Laundering: Systematizing Approval--Execution Binding Failures in AI Coding-Agent Harnesses | Modern AI coding-agent harnesses (Claude Code, Codex CLI, Cursor) rest their security boundary on a largely unexamined assumption: tha… | ↗Not briefed |
arxiv arxiv-ai | 1mo | SimEX: Simulation-Integrated Robotics AutoResearch🔧 Replicate | Coding agents powered by large language models (LLMs) have shown remarkable abilities to autonomously reason about and achieve goals i… | ↗Not briefed |
arxiv arxiv-ai | 1mo | SCORE-LM: State-Space Radar Representations with Language Models for Fault Diagnosis | Radar hardware faults threaten automated perception, motivating accurate, compact diagnosis and understandable maintenance guidance. W… | ↗Not briefed |
hn hn-ai | 1mo | Get your SaaS a better visibility | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | Moralisch flexibler Präsident lässt US-KI-Unternehmen „moralisch-bindenden“ Selbstregulierung unterschreiben. Das war alles, was OpenAI… | Moralisch flexibler Präsident lässt US-KI-Unternehmen „moralisch-bindenden“ Selbstregulierung unterschreiben. Das war alles, was O… | ↗Not briefed |
hn hn-ai | 1mo | Windows11 26H2 is now available | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | Added a new section on LLM code reviews. Spoiler Alert: they not done good at them. And the bigger the diff, the worse they done - unsurpr… | Added a new section on LLM code reviews. Spoiler Alert: they not done good at them. And the bigger the diff, the worse they done - un… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Scale-Split Neural Operator for Memory- and Data-Efficient 3D Turbulence Prediction | Neural surrogates have emerged as fast alternatives to the numerical simulation of three-dimensional turbulence. However, training the… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Fairness Beyond a Single Run: Training-Seed Variability in Speech LLM Adaptation | Demographic fairness gaps in automatic speech recognition are almost always reported from a single training run. We fine-tune the Q-fo… | ↗Not briefed |
rss techmeme | 1mo | Apple Pay launches in India, letting eligible Axis Bank Visa and Mastercard credit cardholders add their cards to Apple Wallet on iPhone, iPad, and Apple Watch (Dhruv Mohan/The Economic Times) | Dhruv Mohan / The Economic Times: Apple Pay launches in India, letting eligible Axis Bank Visa and Mastercard credit cardholders add t… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Don't appreciate anthropic treating me like a junkie who needs to know if his stash will run out before a re-up. On the other hand, I am a… | Don't appreciate anthropic treating me like a junkie who needs to know if his stash will run out before a re-up. On the other hand, I… | ↗Not briefed |
arxiv arxiv-ai | 1mo | RealWorldShop: Benchmarking and Improving Conversational Shopping Agents in Real-World E-commerce | Large language models are reshaping ecommerce from static recommenders into interactive shopping assistants, yet real-world shopping r… | ↗Not briefed |
bluesky bluesky-ai | 1mo | self proclaimed good guys anthropic, generously giving defenders and prominent open source projects access to the best models at cybersecur… | self proclaimed good guys anthropic, generously giving defenders and prominent open source projects access to the best models at cyber… | ↗Not briefed |
hn hn-ai | 1mo | OpenAI Launches Decisions API for Fast AI Choices | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | Making LLMs Say What They Think: Measuring and Improving CoT-Interpretability Alignment | Chain-of-thought (CoT) traces often serve as a proxy for how Large Language Models (LLMs) arrive at their answers. However, growing ev… | ↗Not briefed |
hn hn-ai | 1mo | Beyond Utilization: Energy-Conscious GPU Sharing for Inference Serving | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | ClaudeくんとChatGPTくんとでそれぞれに Docker Sandbox 内にAI Agent を隔離して外部にいたずらされないにはどうすればいいのか？という相談をしているのだがいろんな突破方法が次々に紹介されてて「AIくんたち...。どれもこれもよく思いつきますね、確…🔧 ChatGPT | ClaudeくんとChatGPTくんとでそれぞれに Docker Sandbox 内にAI Agent を隔離して外部にいたずらされないにはどうすればいいのか？という相談をしているのだがいろんな突破方法が次々に紹介されてて「AIくんたち...。どれもこれもよく思いつき… | ↗Not briefed |
bluesky bluesky-ai | 1mo | idk if it's growth as a creator, LLM slop fatigue, or just starvation for human contact with fellow creatives, but so far after a few meeti… | idk if it's growth as a creator, LLM slop fatigue, or just starvation for human contact with fellow creatives, but so far after a few… | ↗Not briefed |
hn hn-ai | 1mo | Sonnet 5.5 Doesn't Worry Anymore. It Also Doesn't Think Anyone's in Charge | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | When Order Matters: First-Speaker Bias and Mitigation through Personality in Sequential Multi-Agent Debate | Multi-agent debate (MAD) is often used to improve large language model (LLM) reasoning, but sequential debate is rarely a neutral aggr… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Alleviating Hallucination in Reasoning Tasks with Training-Free Uncertainty-Guided Steering | Recent work on hallucination detection in large language models has shown that, for a fixed pre-trained model and reasoning task, it i… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Targeted Retrieval, Compact Representations: How CoT Reasoning Improves Long-Context Counting | Large language models (LLMs) have been rapidly improving in long-context tasks, powered by Chain-of-Thought (CoT) reasoning. However,… | ↗Not briefed |
bluesky bluesky-ai | 1mo | 「LLM、VLM等生成AIの処罰については現行法で対象できる」とかほざいてたやつしばき回したろか | 「LLM、VLM等生成AIの処罰については現行法で対象できる」とかほざいてたやつしばき回したろか | ↗Not briefed |
bluesky bluesky-ai | 1mo | Mwahahahaha. He doesn't want to be seen as supporting Trump's 'CNN fake news' comment! 😆😅😁 | Mwahahahaha. He doesn't want to be seen as supporting Trump's 'CNN fake news' comment! 😆😅😁 | ↗Not briefed |
arxiv arxiv-ai | 1mo | Routing Probes Can Improve Without New Information: An Exact-Null Audit of Uncertainty Beyond Model Outputs | Routing signals of modern vision transformers -- expert gates, attention-residual weights and halting scores -- often improve probes t… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Loop-Free Inverse Reinforcement Learning via Sequential Value Recovery with Q-Score Matching | Inverse Reinforcement Learning (IRL) aims to recover a reward function that explains expert demonstrations. Existing IRL methods typic… | ↗Not briefed |