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 | 17d | Efficient Edge-aware Attention Network for Graph Generation | Graph generative models commonly use graph-specific Transformer architectures to jointly update node and edge features. However, these… | ↗Not briefed |
arxiv arxiv-ai | 17d | NarraSpace: Multimodal Scene-Grounded Evidence for Picture-Description Dementia Classification | Picture-description dementia classification uses a shared visual scene, but a transcript-level prediction alone does not identify the… | ↗Not briefed |
rss techcrunch-ai | 17d | Eric Schmidt reportedly owns about half as much of Anthropic as CEO Dario Amodei | The former Google CEO owns about 1% stake in the leading AI lab, Bloomberg reported. | ↗Not briefed |
bluesky bluesky-ai | 17d | and nix, untyped dynamic lang from 2002 is probably the absolute worst case for an llm. When i get pages of eval failures that even i absol… | and nix, untyped dynamic lang from 2002 is probably the absolute worst case for an llm. When i get pages of eval failures that even i… | ↗Not briefed |
bluesky bluesky-ai | 17d | *Obviously you have to take opex into consideration, as well as development/growth capex, but the cost of running any given GPU is, relativ… | *Obviously you have to take opex into consideration, as well as development/growth capex, but the cost of running any given GPU is, re… | ↗Not briefed |
bluesky bluesky-ai | 17d | DeepSeek makes 38 dollars per GPU per hour. Let's say they have 10,000 GPUs (they likely have more). Let's say each GPU cost 100k (it var… | DeepSeek makes 38 dollars per GPU per hour. Let's say they have 10,000 GPUs (they likely have more). Let's say each GPU cost 100k (i… | ↗Not briefed |
bluesky bluesky-ai | 17d | They factually believe that people are chattel . The only way to actually fight them is by actually protecting human autonomy truthout.o… | They factually believe that people are chattel . The only way to actually fight them is by actually protecting human autonomy truth… | ↗Not briefed |
hn hn-ai | 17d | Ask HN: Time to read those programming books? | As llm usage becomes the way we all program and reading code becomes more valuable than writing it, what programming books would you r… | ↗Not briefed |
arxiv arxiv-ai | 17d | Semantic Observability for Autonomous Systems Diagnostics | Modern distributed microservice deployments generate massive volumes of heterogeneous, multi-format observability logs across physical… | ↗Not briefed |
bluesky bluesky-ai | 17d | Glad I saw this post, because I was one of those people who dismissed the email, and sent it to my trash folder. I was able to go back and … | Glad I saw this post, because I was one of those people who dismissed the email, and sent it to my trash folder. I was able to go back… | ↗Not briefed |
hn hn-ai | 17d | OpenWebPBX – An open-source, self-hosted business phone system | — | ↗Not briefed |
hn hn-ai | 17d | GPD Pocket 4 | — | ↗Not briefed |
bluesky bluesky-ai | 17d | The IRS website "instructions" for the FW8BEN form might as well be gibberish for all my brain can make sense—can I ask/confirm with other … | The IRS website "instructions" for the FW8BEN form might as well be gibberish for all my brain can make sense—can I ask/confirm with o… | ↗Not briefed |
arxiv arxiv-ai | 17d | Benchmarking Arabic Dialects for Math Problems | Mathematical reasoning remains underexplored for low-resource languages, including Dialectal Arabic. To bridge this gap, we introduce… | ↗Not briefed |
arxiv arxiv-ai | 17d | Big Claims, Small Margins: Do Reported GraphLLM Gains Survive Simple Baselines? | LLM-based methods for node classification on text-attributed graphs (TAGs) routinely report state-of-the-art accuracy. We ask how thes… | ↗Not briefed |
arxiv arxiv-ai | 17d | Value-Rich, Differently Organized: Measuring Value Expression and Organization in AI-Generated Debates | Current evaluation of LLM-generated debate largely focuses on argumentative quality, leaving the value structure of generated argument… | ↗Not briefed |
hn hn-ai | 17d | Claude designed protein binders against scorpion neurotoxins | — | ↗Not briefed |
arxiv arxiv-ai | 17d | Turning TS2Vec into a Foundation Model for Time Series Classification | TS2Vec is one of the strongest self-supervised methods for time series classification, introduced in 2022. It is designed to be traine… | ↗Not briefed |
arxiv arxiv-ai | 17d | Evolving Agent Interfaces🔧 GPT-5 | Computer-use agents are limited by the interfaces humans give them. An agent may type a spreadsheet cell by cell when a single line of… | ↗Not briefed |
arxiv arxiv-ai | 17d | Neural Implicit Energy Optimization for LiDAR Mapping | Neural implicit representations provide memory-efficient signed distance field (SDF) parameterizations for large-scale LiDAR scenes, b… | ↗Not briefed |
bluesky bluesky-ai | 17d | It's true that LLM-generated code has wormed its way so deeply into the stack, OSes, foundational libraries, etc., that it's currently near… | It's true that LLM-generated code has wormed its way so deeply into the stack, OSes, foundational libraries, etc., that it's currently… | ↗Not briefed |
arxiv arxiv-ai | 17d | What Do RTL Repair Benchmarks Actually Measure? | RTL repair benchmarks should measure whether a model can fix broken RTL given a specification. To this end, we construct and hash-bind… | ↗Not briefed |
bluesky bluesky-ai | 17d | Sources: Sam Altman told OpenAI staff of three AI endgames: an IAEA-style regulator, a CERN-like lab merger, or one lab using RSI to "melt … | Sources: Sam Altman told OpenAI staff of three AI endgames: an IAEA-style regulator, a CERN-like lab merger, or one lab using RSI to "… | ↗Not briefed |
hn hn-ai | 17d | Show HN: iOS location based traveler's currency conversion with tax estimations | I've made this app a while ago and now updated it with responsive UI and nice widgets.<p>Why did I make it?<p>During my travels t… | ↗Not briefed |
rss techmeme | 17d | Sources: Sam Altman told OpenAI staff of three AI endgames: an IAEA-style regulator, a CERN-like lab merger, or one lab using RSI to "melt everyone else's GPUs" (The Information) | The Information: Sources: Sam Altman told OpenAI staff of three AI endgames: an IAEA-style regulator, a CERN-like lab merger, or one l… | ↗Not briefed |
hn hn-ai | 17d | Show HN: Ankita – an open-source personal AI companion for the desktop | — | ↗Not briefed |
arxiv arxiv-ai | 17d | PaPi: Parallel Picard-Duhamel Integration for Fast Neuro-Spectral Architectures | Neuro-spectral architectures (NeuSA) are a class of physics-informed neural networks (PINNs) that have proven effective for solving ti… | ↗Not briefed |
rss wired-ai | 17d | French Government Calls Emergency Meetings With Platforms Over Protests | TikTok and other sites were told to curb protest footage and AI-altered images, while online retailers like Vinted and eBay reportedly… | ↗Not briefed |
arxiv arxiv-ai | 17d | WildMAE: Structure-Aware Masked Autoencoding for Wildlife Image Reconstruction | Camera-trap imagery presents substantial variation in animal appearance, illumination, background clutter, and visual degradation. Mas… | ↗Not briefed |
hn hn-ai | 17d | TriggerAPI – A native API client written in Go with full support for MCP servers | — | ↗Not briefed |
arxiv arxiv-ai | 17d | Two $\operatorname{ReLU}$ Layers is All You Need | We prove that every $\operatorname{ReLU}$ network can be exactly simulated by a $\operatorname{ReLU}$ network with two hidden layers.… | ↗Not briefed |
arxiv arxiv-ai | 17d | Capable Teacher, Myopic Scorer: Reward-Free Teacher Repair for Agentic On-Policy Distillation🔧 Consensus | Agentic on-policy distillation (OPD) trains a student agent on its own trajectories while a teacher scores every token. A natural expl… | ↗Not briefed |
arxiv arxiv-ai | 17d | SceneRig: An Agentic System for Simulation-Ready 3D Scene Reconstruction from Single Images | Simulation is a powerful tool for robotics, but building diverse, simulation-ready scenes at scale remains difficult. Reconstructing s… | ↗Not briefed |
arxiv arxiv-ai | 17d | Beyond the Sequentiality Bottleneck: A Hybrid Diffusion-Autoregressive Framework for Efficient Multimodal Reasoning | Autoregressive models (ARMs) face a "Sequentiality Bottleneck" in reasoning, while diffusion language models (DLMs) offer parallelism… | ↗Not briefed |
bluesky bluesky-ai | 17d | Can't wait to find out what kind of society-wide disasters we'll unlock when organizations overconfidently attempt to migrate their legacy … | Can't wait to find out what kind of society-wide disasters we'll unlock when organizations overconfidently attempt to migrate their le… | ↗Not briefed |
hn hn-ai | 17d | Reducing Reward Hacking by Localizing Bad Behavior Instead of Penalizing It | — | ↗Not briefed |
hn hn-ai | 17d | Show HN: Frink – Desktop-first app for multi-step workflows and parallel agents🔧 n8n | What’s up HN?! I'm Benji, and I’ve been building Frink since ~April this year.<p>Frink is a desktop-first, open-source (MIT) app… | ↗Not briefed |
arxiv arxiv-ai | 17d | Continuous-Time Policy Gradient from Discrete Data: Convergence Analysis and Model-Free Algorithms | Continuous-time reinforcement learning seeks an optimal policy under unknown dynamics driven by stochastic differential equations usin… | ↗Not briefed |
arxiv arxiv-ai | 17d | Spatially Informed and Direction Sensitive Representation Learning for Spatial Transcriptomics | Cellular states are shaped by both intrinsic gene expression programs and the spatial microenvironment in which cells reside. Spatial… | ↗Not briefed |
arxiv arxiv-ai | 17d | Sage-A1.5: From Long-Horizon Execution to Scientific Exploration with Capability Composition | In this report, we introduce Sage-A1.5, a 35B Mixture-of-Experts agentic model that builds on A1 to advance long-horizon problem solvi… | ↗Not briefed |
arxiv arxiv-ai | 17d | Cache Engineering for Agents | Large Language Model (LLM) agents can handle increasingly complex tasks, navigating long trajectories of observations and actions. As… | ↗Not briefed |
arxiv arxiv-ai | 17d | The Token-Energy Law of LLM Decode: Active Bytes and Realized Time Account for Joules per Token | Energy-efficient deployment of large language models requires understanding how model architecture, hardware, and serving settings sha… | ↗Not briefed |
bluesky bluesky-ai | 17d | Plants are trying to talk, if only we'd listen. Now, one startup says it has decoded their language. “We’re listening more than we’re speak… | Plants are trying to talk, if only we'd listen. Now, one startup says it has decoded their language. “We’re listening more than we’re… | ↗Not briefed |
rss techmeme | 17d | Sensor Tower: Muse had 10.9M downloads since its September debut, 10.2M of those in the US, becoming the 12th app to hit 10M US downloads in its first 40 days (Sarah Perez/TechCrunch) | Sarah Perez / TechCrunch: Sensor Tower: Muse had 10.9M downloads since its September debut, 10.2M of those in the US, becoming the 12t… | ↗Not briefed |
arxiv arxiv-ai | 17d | Evolving Semantic Indexes for Generative Retrieval | Generative retrieval (GR) casts search and recommendation as autoregressive generation of discrete semantic identifiers (SIDs). SID de… | ↗Not briefed |
hn hn-ai | 17d | Summonos: An omarchy-inspired macOS environment | — | ↗Not briefed |
arxiv arxiv-ai | 17d | Continual Concept Personalization in Diffusion Models via Residual-Space Adaptation | Text-to-image personalization aims to equip pretrained diffusion models with the ability to learn user-specific concepts from only a f… | ↗Not briefed |
bluesky bluesky-ai | 17d | it’s 2042, we only share our comics by printing them as books the books still get ripped apart and scanned for the AI model | it’s 2042, we only share our comics by printing them as books the books still get ripped apart and scanned for the AI model | ↗Not briefed |
arxiv arxiv-ai | 17d | The Dynamics of Collapse and Recovery in Clustered Federated Learning | Clustered federated learning (CFL) can suffer from cluster collapse. In particular, under unfavorable initialization, early client ass… | ↗Not briefed |
bluesky bluesky-ai | 17d | Theranos would be a 100 billion dollar subsidiary of Anthropic if it launched after 2020 | Theranos would be a 100 billion dollar subsidiary of Anthropic if it launched after 2020 | ↗Not briefed |