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 | 5d | Anti-coherence Regularized Absolute Scaling for Out-of-Distribution Detection | Out-of-distribution (OOD) detection is recognized as a critical problem in AI safety. SCALE is an effective post-hoc OOD detection met… | ↗Not briefed |
bluesky bluesky-ai | 5d | Starzoo: rip foil star packs, flip 40 cosmic creatures drawn in code, complete constellations. Honest odds shown openly, pity counter, no r… | Starzoo: rip foil star packs, flip 40 cosmic creatures drawn in code, complete constellations. Honest odds shown openly, pity counter,… | ↗Not briefed |
arxiv arxiv-ai | 5d | RSIHub: A Composable Framework for Agent-Evolution Methods | Agent-evolution methods use execution feedback to improve prompts, skills, tools, and orchestration, but their implementations often r… | ↗Not briefed |
bluesky bluesky-ai | 5d | China's DeepSeek, MiniMax win approval for AI global use, state media-affiliated entity says reut.rs/4sgZLCk | China's DeepSeek, MiniMax win approval for AI global use, state media-affiliated entity says reut.rs/4sgZLCk | ↗Not briefed |
arxiv arxiv-ai | 5d | Easy-3Dfy: Object-Level Geometric Composition for Single-Image 3D Scene Construction | Single-image 3D scene construction is a fundamental problem in 3D vision. Recent methods such as SAM-3Dfy enable end-to-end reconstruc… | ↗Not briefed |
hn hn-ai | 5d | The First Gist on GitHub | — | ↗Not briefed |
bluesky bluesky-ai | 5d | Agent Skills (Anthropic), Open Responses (OpenAI), OpenEnv (Hugging Face) and OpenSandbox (Alibaba) all moved to the Agentic AI Foundation …🔧 Hugging Face | Agent Skills (Anthropic), Open Responses (OpenAI), OpenEnv (Hugging Face) and OpenSandbox (Alibaba) all moved to the Agentic AI Founda… | ↗Not briefed |
bluesky bluesky-ai | 5d | Glowvial: pour glowing orbs between glass vials until each holds one color. Endless levels, all checked solvable by a solver; fog, frost an… | Glowvial: pour glowing orbs between glass vials until each holds one color. Endless levels, all checked solvable by a solver; fog, fro… | ↗Not briefed |
bluesky bluesky-ai | 5d | Tinkernova: merge tiny probes into stars. Drag two twins together, discover 30 tiers, gamble on fusion (odds always shown) and launch to ne… | Tinkernova: merge tiny probes into stars. Drag two twins together, discover 30 tiers, gamble on fusion (odds always shown) and launch… | ↗Not briefed |
hn hn-ai | 5d | GitHub – mamund/APIs-and-agents: service choreography demos | — | ↗Not briefed |
hn hn-ai | 5d | Show HN: A TypeScript remake of Chinese Paladin (1995), now in English | — | ↗Not briefed |
arxiv arxiv-ai | 5d | Do Halting Heads Really Save Compute in Looped Language Models? | Looped language models run each token through the same layers several times, and each run is one iteration. To save compute, an extra… | ↗Not briefed |
hn hn-ai | 5d | Show HN: Swurf – Swipe through the real web | Hello HN,<p>Swurf allows you to swipe through the real web one page after another. It learns what you like and doesn’t show you the sa… | ↗Not briefed |
arxiv arxiv-ai | 5d | Minimum Description Length Unsupervisedly Separates Law-Grade from Statistical-Grade Knowledge in a Self-Improving Science Agent - and Reveals 86% of Its Own Discoveries Are Statistical | AI systems that discover scientific relations now generate candidate "laws" at scale, but the bottleneck has shifted from generation t… | ↗Not briefed |
arxiv arxiv-ai | 5d | BiLoCo: Binary Low-Rank Corrections for Post-Training Quantization | Post-training quantization (PTQ) to FP4 has emerged as a key technique for reducing large language model inference cost, especially on… | ↗Not briefed |
arxiv arxiv-ai | 5d | Spatial Heuristic Strategy based Diffusion Reinforcement Learning for Online 3D Bin Packing | Traditional heuristic methods have accumulated mature experience in core objectives, such as space utilization and layout compactness… | ↗Not briefed |
bluesky bluesky-ai | 5d | They're so desperate to rebuild this site around the LLM-addict six figure devs in their DMs, that they're now openly pushing away all of t… | They're so desperate to rebuild this site around the LLM-addict six figure devs in their DMs, that they're now openly pushing away all… | ↗Not briefed |
arxiv arxiv-ai | 5d | Financial Context-Rich Strategy for Actor-Critic based Agents: A Deep Reinforcement Learning Approach for Stock Trading | In modern financial markets, more than 3/4th of the large volume trade is executed by algorithmic trading based on advanced trading st… | ↗Not briefed |
hn hn-ai | 5d | LessMi – A Windows Rat in Python, Controlled Through Discord | — | ↗Not briefed |
hn hn-ai | 5d | Whistle speech-to-text running offline on a 16mb ESP32 | — | ↗Not briefed |
arxiv arxiv-ai | 5d | Correct but Unselectable: The Hidden Interface Tax in Multi-Candidate Reasoning | Small language models deployed under cost, latency, or privacy constraints cannot simply switch to larger models when reasoning fails.… | ↗Not briefed |
arxiv arxiv-ai | 5d | MS-TTRL: Breaking the Frequency Conflict in Multiple-Solution Test-Time Reinforcement Learning | Scientific, engineering, and mathematical reasoning tasks can admit multiple valid solutions, while exhaustive annotation of the full… | ↗Not briefed |
arxiv arxiv-ai | 5d | TileShare: Training-Free Acceleration of Overlapping Tiled ViT Decoding | Fixed-window Vision Transformer (ViT) decoders run at deployment resolution as overlapping tiles: every token in an overlap band is pr… | ↗Not briefed |
arxiv arxiv-ai | 5d | Training an MoE LLM in MXFP4: How We Closed the Gap to BF16 | Training large language models (LLMs) in sub-8-bit arithmetic promises substantial gains in memory bandwidth and compute throughput, y… | ↗Not briefed |
bluesky bluesky-ai | 5d | 形式検証に必要な膨大な手間をLLM（大規模言語モデル）が引き受けてくれるなら、通常のソフトウェア開発プロセスに組み込むことが可能となる。アンソロピックは、実際にオープンソースソフトウェアのライブラリの形式検証を実施し、このことを立証した。 AIとかセキュリティの分野では気が滅… | 形式検証に必要な膨大な手間をLLM（大規模言語モデル）が引き受けてくれるなら、通常のソフトウェア開発プロセスに組み込むことが可能となる。アンソロピックは、実際にオープンソースソフトウェアのライブラリの形式検証を実施し、このことを立証した。 AIとかセキュリティの分野… | ↗Not briefed |
hn hn-ai | 5d | Netryx – Find where a street photo was taken, to the block, offline | — | ↗Not briefed |
hn hn-ai | 5d | Monochrome – Mobile Browser for macOS | — | ↗Not briefed |
bluesky bluesky-ai | 5d | アンソロピックのBlog記事。たいへん興味深い内容。 「現在、LLM（大規模言語モデル）は大規模な形式証明を生成できるようになり、セキュリティが証明された保証を含む新たなソフトウェア開発ワークフローが可能となった」 「モデルの能力が向上し続けるにつれ、形式検証はセキュアなソ… | アンソロピックのBlog記事。たいへん興味深い内容。 「現在、LLM（大規模言語モデル）は大規模な形式証明を生成できるようになり、セキュリティが証明された保証を含む新たなソフトウェア開発ワークフローが可能となった」 「モデルの能力が向上し続けるにつれ、形式検証はセ… | ↗Not briefed |
arxiv arxiv-ai | 5d | RoboCode: Learning Reusable Action Primitives for Robot Manipulation | Recent robot manipulation policies commonly predict continuous action chunks directly from observations, without explicitly modeling r… | ↗Not briefed |
bluesky bluesky-ai | 5d | OpenAI and Poland’s NASK collaborate on AI safety research reut.rs/4BgT1sh | OpenAI and Poland’s NASK collaborate on AI safety research reut.rs/4BgT1sh | ↗Not briefed |
hn hn-ai | 5d | Show HN: A 200-item RLHF quality audit with Fleiss' Kappa (0.68) | — | ↗Not briefed |
arxiv arxiv-ai | 5d | Correct Answers, Closed Gates: SolveBind for GUI Agents at the Verification Barrier🔧 GPT-5 | Websites and applications deploy CAPTCHAs and anti-automation mechanisms that use behavioral signals such as interaction trajectories… | ↗Not briefed |
bluesky bluesky-ai | 5d | I guess psychology is partly protected here because the reviews are so incoherent, using them is only going to give the model the AI-equiva… | I guess psychology is partly protected here because the reviews are so incoherent, using them is only going to give the model the AI-e… | ↗Not briefed |
hn hn-ai | 5d | Running Megabonk on an Apple Silicon Mac | — | ↗Not briefed |
bluesky bluesky-ai | 5d | This is a pretty good take. The only thing I'd add is to remind people that the regulation proposed by Anthropic and OpenAI is more or less… | This is a pretty good take. The only thing I'd add is to remind people that the regulation proposed by Anthropic and OpenAI is more or… | ↗Not briefed |
arxiv arxiv-ai | 5d | RxnExBench: Benchmarking Vision-Language Models for Structured Chemical Reaction Extraction from Scientific Literature | Screen Tables and Substrate Scopes are information-dense figures widely used in synthetic chemistry literature, yet automated extracti… | ↗Not briefed |
hn hn-ai | 5d | Show HN: Pear – A Windows Steam Launcher for Apple Silicon Macs | I first tried to play Megabonk on my Mac with Wine 11.0 and DXMT 0.80.<p>After successfully playing it, I thought it would be great to… | ↗Not briefed |
arxiv arxiv-ai | 5d | PEAR: Adaptive Advantage Reshaping for Perception and Exploration in Large Vision-Language Models | Effective multimodal reasoning requires a model to reliably utilize visual information while sufficiently exploring diverse reasoning… | ↗Not briefed |
arxiv arxiv-ai | 5d | Democratizing LLM Training with the Edge: Challenges, Systems, and the Road Ahead | The growing computational power of modern GPU data centers has let a few industrial giants train and deploy large language models (LLM… | ↗Not briefed |
bluesky bluesky-ai | 5d | deliberately sticking your foot in the horse trap and declaring loudly that it is some sort of trap as it snaps shut around you is an inter… | deliberately sticking your foot in the horse trap and declaring loudly that it is some sort of trap as it snaps shut around you is an… | ↗Not briefed |
arxiv arxiv-ai | 5d | DIAL-S2ST: Learning Latency Control for Simultaneous Speech-to-Speech Translation | We propose DIAL-S2ST, a post-training method for learning inference-time latency control in simultaneous speech-to-speech translation… | ↗Not briefed |
arxiv arxiv-ai | 5d | Head Dimension Matters in Language Models for Long-Context Retrieval | Multi-head attention retrieves and combines context representations through query--key interactions and value aggregation, with head d… | ↗Not briefed |
arxiv arxiv-ai | 5d | SOPO: Socratic Guided Policy Optimization for Span-Level Hallucination Detection🔧 GPT-5 | Large Language Models (LLMs) have gained widespread adoption in various natural language processing tasks, but suffer from hallucinati… | ↗Not briefed |
rss marktechpost | 5d | Google Cloud Launches Agent Compute Engine: A MicroVM per AI Agent | Google Cloud announced Agent Compute Engine, a compute service that gives every AI agent its own hardware-isolated microVM with no cre… | ↗Not briefed |
arxiv arxiv-ai | 5d | Do VLAs Know They Will Fail? Interpretable Success and Failure Patterns in Physical AI Systems | Vision-Language-Action (VLA) models have emerged as powerful general-purpose robot policies, but their successes and failures remain d… | ↗Not briefed |
hn hn-ai | 5d | Show HN: An automated news site Distribute Software | I wanted to see what it would look like building a fully automated news site monitoring different newsrooms, desks, feeds etc on the w… | ↗Not briefed |
rss techmeme | 5d | SK Telecom unit SK Hyper is in talks with tech firms and investors to finance 15GW of data center capacity in South Korea by 2035; the first 5GW will cost $350B (Bloomberg) | Bloomberg: SK Telecom unit SK Hyper is in talks with tech firms and investors to finance 15GW of data center capacity in South Korea b… | ↗Not briefed |
arxiv arxiv-ai | 5d | Watermarking Diffusion Language Models with Learned Foresight and Balanced Bands | Green-red watermarks favor tokens from a green list that a secret key derives from the preceding tokens. Diffusion language models, ho… | ↗Not briefed |
hn hn-ai | 5d | The AI Expectations Gap [pdf] | — | ↗Not briefed |
arxiv arxiv-ai | 5d | Neural Disjunctive Normal Forms for Learning Exact Rank Filters | Gradient-based learning of morphological operators is hindered by locally sparse derivatives with respect to the structuring elements.… | ↗Not briefed |