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 | Caught in the Act: Probes Effectively Detect Sabotage and Catch Unverbalized Deception | Recent incidents have highlighted the challenge of monitoring LLM agents and the danger of models deceiving people. We show that white… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Rounding in Preconditioner Space: Redesigning 4-bit AdamW Optimizer-State Quantization | Quantizing AdamW's optimizer states reduces persistent storage, but quantization errors propagate through the moment recurrences and p… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Density Ratio Estimation with Stein Displacement Fields | Density ratios quantify distribution shift from a probability-mass point of view, whereas displacement fields describe, from a dynamic… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Ecology of AI Agents: Collaboration Creates a Population Threshold for Takeoff | AI agents can now conduct real-world cyberattacks, scale up capabilities with the number of agents, and collectively pursue misaligned… | ↗Not briefed |
arxiv arxiv-ai | 1mo | VioLA: Learning Generalist Humanoid Control Policies from Human Data | Teaching a humanoid to follow instructions with its whole body runs into two obstacles. Its action space is large and tightly coupled:… | ↗Not briefed |
hn hn-ai | 1mo | World Labs: Explicit camera control for image-to-video | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | FAITH: Feasibility-Aware Safety-Filtered RL for High-Dimensional Systems | Safe reinforcement learning commonly places safety and task performance in the same policy objective, where they can introduce competi… | ↗Not briefed |
rss theverge-ai | 1mo | SpaceXAI backs Omarchy, the controversial Linux distro, with $1.5 million in compute | If Elon Musk and SpaceXAI were going to back any Linux distro, it seems obvious they'd back Omarchy. Today it was announced that Space… | ↗Not briefed |
hn hn-ai | 1mo | The Goose Programming Language | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | Toward Joint Optimization of Circuit Depth and Training Data Size in Adaptively Grown Quantum Classifiers | Building a quantum model involves a tradeoff: how complex the circuit should be, and how much training data it needs. Caro et al. show… | ↗Not briefed |
bluesky bluesky-ai | 1mo | What this implies to me is that OpenAI has barely grown revenues since summer. | What this implies to me is that OpenAI has barely grown revenues since summer. | ↗Not briefed |
arxiv arxiv-ai | 1mo | FastBench: Can Streaming VLMs Perceive High-Dynamic Real-World Streams? | Streaming Video Large Language Models (VLMs) enable continuous video understanding, yet existing benchmarks focus on low-dynamic scena… | ↗Not briefed |
arxiv arxiv-ai | 1mo | RoboRSI: Stable, efficient, and reusable robot self-evolution in complex real-world environments | A generalist robot should not only perform diverse tasks but also improve through experience, turning what it learns during execution… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Transformer-Based Masked Diffusion Models Provably Learn Distributions with Latent Structure | Masked diffusion models (MDMs) generate discrete data by iteratively unmasking multiple tokens in parallel using a learned mask predic… | ↗Not briefed |
bluesky bluesky-ai | 1mo | you're hallucinating that, sorry. you can get access to models 95% as good as Anthropic / OpenAI who serve *open-weight* models basically … | you're hallucinating that, sorry. you can get access to models 95% as good as Anthropic / OpenAI who serve *open-weight* models basic… | ↗Not briefed |
hn hn-ai | 1mo | Objgit is not a Git forge | — | ↗Not briefed |
hn hn-ai | 1mo | Show HN: Layer – One API for Polymarket and Kalshi | Hi HN, I'm David. I built Layer, an API + SDK or building trading bots/ai agents across prediction markets.<p>The idea came… | ↗Not briefed |
bluesky bluesky-ai | 1mo | The circuitous route to answer the straightforward question of “what is the revenue run rate” is wild. | The circuitous route to answer the straightforward question of “what is the revenue run rate” is wild. | ↗Not briefed |
arxiv arxiv-ai | 1mo | Beyond Spatio-Temporal Priors: A Generalizable Approach for Dense Correspondence Matching | Dense correspondence matching has historically been bounded by simplifying spatio-temporal priors, such as smooth motion and rigid geo… | ↗Not briefed |
arxiv arxiv-ai | 1mo | A Unified Bellman Operator for Safety-Critical Reinforcement Learning | Reinforcement learning in safety-critical domains requires maximizing task performance while strictly adhering to safety constraints.… | ↗Not briefed |
hn hn-ai | 1mo | Show HN: DocFlare AI – Open-source docs chatbot on Cloudflare's free tier | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | This is the most consequential thing that might come from LLM math. Note that the NSA switched to post-quantum crypto years ago and Signa… | This is the most consequential thing that might come from LLM math. Note that the NSA switched to post-quantum crypto years ago and… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Yesterday OpenAI solved all of math and shockingly it turns out.... not so much | Yesterday OpenAI solved all of math and shockingly it turns out.... not so much | ↗Not briefed |
arxiv arxiv-ai | 1mo | WOVEN: Weaving Visual World Modeling into Multimodal LLMs🔧 GPT-5 | Multimodal large language models (MLLMs) struggle with spatial, embodied, physical, and temporal reasoning. We hypothesize that these… | ↗Not briefed |
arxiv arxiv-ai | 1mo | MAMHOI: Factorizing Scene-Aware Human-Object Interaction through Affordances | Generating realistic human-object interactions (HOI) in complex 3D scenes requires two complementary capabilities: reasoning about int… | ↗Not briefed |
bluesky bluesky-ai | 1mo | OpenAI: | OpenAI: | ↗Not briefed |
hn hn-ai | 1mo | USA Today sues OpenAI for copyright infringement over AI training | — | ↗Not briefed |
hn hn-ai | 1mo | Show HN: Viberuler – measure where your AI coding context goes | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | Sure seems like every AI use case is either A) "I want a little slave" B) "I don't want to have human interactions with my child/spouse/fri… | Sure seems like every AI use case is either A) "I want a little slave" B) "I don't want to have human interactions with my child/spous… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Predicting Alignment Generalization with Value Representations | LLM developers post-train their models to exhibit prosocial values and behavioral traits, which are enumerated in an alignment target.… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Searching for "Harmful Refusal": A Psychometric Audit of an AI Safety Benchmark | Safety benchmarks typically report one overall score for a suite of datasets, each of which may target one or more safety-related attr… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Did you read the description of how he did it? He had Claude write all the code and do the analysis. Yes, he asked the questions, and had C… | Did you read the description of how he did it? He had Claude write all the code and do the analysis. Yes, he asked the questions, and… | ↗Not briefed |
arxiv arxiv-ai | 1mo | LeWAM: A JEPA World Action Model with Diffusion-Steering-Based MPC | World action models (WAMs) predict actions and future observations, typically from a reconstruction-based representation that carries… | ↗Not briefed |
hn hn-ai | 1mo | FreeLinX-base 1.3.1 stable: a ~290 MB ISO Linux system with a NetBSD userland | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | This is gonna be a tough one for OpenAI, huh? | This is gonna be a tough one for OpenAI, huh? | ↗Not briefed |
arxiv arxiv-ai | 1mo | ViSkill: Reinforcing VLM Agents with Evolving Visual-Native Skills🔧 skills Skill | Skill-augmented agents improve sample efficiency by distilling successful trajectories into reusable strategies. Yet most existing app… | ↗Not briefed |
arxiv arxiv-ai | 1mo | SpaceCast-Bench: Evaluating Predictive Spatial Reasoning in Vision-Language Models | Existing spatial reasoning benchmarks mainly test spatial perception: reading off relations already visible in the input. Yet real-wor… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Learning Kilometer-Scale Weather Prediction with Global-Regional Alignment | Kilometer-scale regional weather forecasting is essential for local weather warnings and weather-sensitive decisions. Existing data-dr… | ↗Not briefed |
bluesky bluesky-ai | 1mo | A father and his son are in a car crash. The father dies and when the son is rushed to the hospital, the doctor says, “I can’t read a book … | A father and his son are in a car crash. The father dies and when the son is rushed to the hospital, the doctor says, “I can’t read a… | ↗Not briefed |
arxiv arxiv-ai | 1mo | SpaceFlow: Locally Controllable 3D Generation | Current 3D generation methods lack explicit local control: geometric adherence is often defined by a global control strength, and appe… | ↗Not briefed |
bluesky bluesky-ai | 1mo | LLMs get sub-1-bit compression 📉 LittleBit and LittleBit-2 enable LLM compression down to 0.1 bits per weight, with no inference overhead… | LLMs get sub-1-bit compression 📉 LittleBit and LittleBit-2 enable LLM compression down to 0.1 bits per weight, with no inference ove… | ↗Not briefed |
bluesky bluesky-ai | 1mo | What's 20 billion between friends? (I think OpenAI folding will really bring the whole house of cards to fall) | What's 20 billion between friends? (I think OpenAI folding will really bring the whole house of cards to fall) | ↗Not briefed |
arxiv arxiv-ai | 1mo | Prospective Prediction of OOD Degradation from Source-Side Training Dynamics | We study whether persistent out-of-distribution (OOD) degradation can be predicted before it is directly observed using only source-si… | ↗Not briefed |
hn hn-ai | 1mo | Does Claude Feel the Whip? | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | HRIL: Learning Multimodal Synergy via Higher-Order Tensor Modeling🔧 Modal | Self-supervised multimodal representation learning has achieved remarkable success across diverse domains, yet capturing synergistic i… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Colin is right. OpenAI is staffed by a bunch of mathematicians I respect, and yet the company acts in the worst possible way at all times. … | Colin is right. OpenAI is staffed by a bunch of mathematicians I respect, and yet the company acts in the worst possible way at all ti… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Details in the article below | Details in the article below | ↗Not briefed |
bluesky bluesky-ai | 1mo | It's neither. It's Large Language Model. "AI" is being used as a bullshit marketing term (most don't actually pass the Turing test under de… | It's neither. It's Large Language Model. "AI" is being used as a bullshit marketing term (most don't actually pass the Turing test und… | ↗Not briefed |
arxiv arxiv-ai | 1mo | GeoReform: Reflective Formalization Evolution for Multimodal Geometry Problem Solving | Multimodal large language models (MLLMs) often struggle to identify and use geometric relations in diagrams. Recent methods address th… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Long Text to Predictive Features: LLM-Guided Blockwise Feature Engineering via Executable Program Search | Industrial risk-control systems typically rely on structured-data models for efficient prediction, yet substantial valuable informatio… | ↗Not briefed |