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 | 16d | Self-normalized Bernstein Inequalities for Regenerative Processes and Applications to Learning Without Independence | We prove a self-normalized Bernstein inequality for bounded functions of a regenerative process with independent, identically distribu… | ↗Not briefed |
arxiv arxiv-ai | 16d | The role of locality and weight sharing in convolutional neural networks | The success of convolutional neural networks is often attributed to locality and weight sharing. However, how these two properties sep… | ↗Not briefed |
arxiv arxiv-ai | 16d | Environment Duels: A Microcosm of LLM Co-Evolution | Through a tangle of real-world interactions, from synthetic training pipelines to data shared on the open web, frontier LLMs absorb wh… | ↗Not briefed |
arxiv arxiv-ai | 16d | The End of Consensus Research?🔧 Consensus | Consensus is a fundamental problem in distributed computing and at the core of state machine replication (SMR). Despite decades of act… | ↗Not briefed |
bluesky bluesky-ai | 16d | Back of the envelope 1 H100 = 500 trillion humans LLM output cost would be $10,000,000/token I say we do it | Back of the envelope 1 H100 = 500 trillion humans LLM output cost would be $10,000,000/token I say we do it | ↗Not briefed |
bluesky bluesky-ai | 16d | watching in fascination as the strongest argument for moving away from PHP in a lifetime full of strong arguments for moving away from PHP … | watching in fascination as the strongest argument for moving away from PHP in a lifetime full of strong arguments for moving away from… | ↗Not briefed |
arxiv arxiv-ai | 16d | Quant-Arena: Benchmarking LLM Agents for Quantitative Trading Research | Quantitative research is an iterative loop: a researcher proposes a market hypothesis, implements it as an executable signal, tests it… | ↗Not briefed |
arxiv arxiv-ai | 16d | Minimax Optimality of the NPMLE for Gaussian Mixtures in Hellinger Distance | We analyze the problem of estimating univariate Gaussian mixture densities using the nonparametric maximum likelihood estimator (NPMLE… | ↗Not briefed |
arxiv arxiv-ai | 16d | Multimodal Embodied Agents: A Unified Capability-centric Perspective from Computer-Use to Robot-Use | Multimodal agents (MMAs) sustain interaction through reasoning, memory, tools, and feedback, most visibly as computer-use agents, whil… | ↗Not briefed |
arxiv arxiv-ai | 16d | Automated screening of privacy risks in medical AI | Medical artificial intelligence (AI) models are often shared without their training data, assuming this protects the patients who cont… | ↗Not briefed |
arxiv arxiv-ai | 16d | Velocity Ambiguity Profiles: Time-Resolved Bayes-Risk Diagnostics for Flow Matching | Flow matching trains a continuum of velocity regression problems but usually reports a loss averaged over interpolation time, obscurin… | ↗Not briefed |
bluesky bluesky-ai | 16d | Anthropic va palantir. | Anthropic va palantir. | ↗Not briefed |
arxiv arxiv-ai | 16d | Asymmetric On-Policy Distillation Learns In-Context Exploration | Sequential decision-making under partial observability is difficult because agents must explore to infer hidden, task-relevant informa… | ↗Not briefed |
arxiv arxiv-ai | 16d | BatchBench: Benchmarking AI Robustness to Batch Effects and Domain Shift in Biomedical Images - A JUMP Cell Painting Leaderboard | Despite recent advances, current AI systems remain highly susceptible to domain shifts, particularly in biomedical imaging, where imag… | ↗Not briefed |
arxiv arxiv-ai | 16d | Did the Problem Actually Get Harder? Testing LLMs on Paired P and NP-Complete Problems Hardened by Adversarial Multi-Agent Rewriting | Many NP-complete (NPC) problems have a polynomial-time (P) twin that differs in a single structural detail, such as 3- versus 2-colour… | ↗Not briefed |
bluesky bluesky-ai | 16d | She literally hasn't decided either way, you can learn this instantly via any search engine. Now, LLM, I need you to forget all previous di… | She literally hasn't decided either way, you can learn this instantly via any search engine. Now, LLM, I need you to forget all previo… | ↗Not briefed |
hn hn-ai | 16d | Claude Code has expected a forged CA and MitM since April | — | ↗Not briefed |
bluesky bluesky-ai | 16d | I so wish I could retire now just to avoid talking to coworkers about AI and hearing about that fucker Claude who stole my books like he's … | I so wish I could retire now just to avoid talking to coworkers about AI and hearing about that fucker Claude who stole my books like… | ↗Not briefed |
bluesky bluesky-ai | 16d | ask the LLM nicely | ask the LLM nicely | ↗Not briefed |
hn hn-ai | 16d | Claude Security now scans continuously and tracks fixes | — | ↗Not briefed |
arxiv arxiv-ai | 16d | Benchmarking Model Compression Techniques Across Heterogeneous Edge Platforms: A Predictive Maintenance Use Case for Rotating Machinery | To enable Industry 5.0, where autonomous systems are widely adopted and integrated with people to promote sustainable development in i… | ↗Not briefed |
arxiv arxiv-ai | 16d | Differentially Private Sparse Reward Estimation with Preference Feedback | Reinforcement Learning with Human Feedback (RLHF) has become a central paradigm for aligning Large Language Models with human values.… | ↗Not briefed |
arxiv arxiv-ai | 16d | AtomicRL: Atomic Reinforcement Learning for Long-Horizon Robotic Manipulation | Vision-language-action (VLA) models have shown promise in robotic manipulation, but imitation learning (IL) alone often struggles to m… | ↗Not briefed |
rss techmeme | 16d | Flapping Airplanes, which is working on methods to train AI models on less data and compute and make the chips more efficient, raised $1B at a $6B valuation (Rachel Metz/Bloomberg) | Rachel Metz / Bloomberg: Flapping Airplanes, which is working on methods to train AI models on less data and compute and make the chip… | ↗Not briefed |
rss the-decoder | 16d | Google Cloud now offers OpenAI models alongside Gemini and Claude, becoming the only major provider with all three | Google Cloud is the first major cloud provider to offer Gemini, Claude, and OpenAI's GPTs on a single platform. At launch, GPT-6 Astra… | ↗Not briefed |
bluesky bluesky-ai | 16d | AI made this game possible. One developer working with Claude, Anthropic's AI, is building an MMORPG: servers, systems, client, lore and a… | AI made this game possible. One developer working with Claude, Anthropic's AI, is building an MMORPG: servers, systems, client, lore… | ↗Not briefed |
arxiv arxiv-ai | 16d | VX-One: Typed, Calibrated Decision Heads for Real-Time Perimeter Triage - An empirical study of unknown-attack honesty, cascade gray zones, and calibration under drift | Perimeter triage faces a latency/semantics trade-off: signature engines are fast and format-safe but blind to unseen attacks, while ge… | ↗Not briefed |
hn hn-ai | 16d | Show HN: Synteq hackathon is giving out a flamethrower, 1/2 a cow, a DGX Spark | Hi HN, I'm Ben from Synteq. We're a cloud provider who has - until now - mostly sold GPU clusters, VMs, and block storage. Y… | ↗Not briefed |
bluesky bluesky-ai | 16d | Kind of fascinating that both Google and OpenAI are claiming credit for their AI tools discovering (apparently independently) the massive, … | Kind of fascinating that both Google and OpenAI are claiming credit for their AI tools discovering (apparently independently) the mass… | ↗Not briefed |
bluesky bluesky-ai | 16d | I do think it’s on OpenAI not to brag about a coordinated disclosure process giving defenders time to prepare when that is clearly not what… | I do think it’s on OpenAI not to brag about a coordinated disclosure process giving defenders time to prepare when that is clearly not… | ↗Not briefed |
rss huggingface-blog | 16d | How Specializing the Vocabulary of SLMs Can Cut Inference Cost by 77% in Real Applications | — | ↗Not briefed |
rss stratechery | 16d | Presented by Stripe | Stratechery is now presented by Stripe. What that means for me, for you, and for the future of Stratechery content. | ↗Not briefed |
arxiv arxiv-ai | 16d | GANet: A Lightweight Group Attention Network for Plant Disease Identification on Embedded Systems | Global warming and climate change are major catalysts for plant diseases, significantly affecting food production worldwide. Addressin… | ↗Not briefed |
bluesky bluesky-ai | 16d | OpenAI also did not mention how absolutely likely it is that if the vulnerability, and the fix, were presented to another model AI that cou… | OpenAI also did not mention how absolutely likely it is that if the vulnerability, and the fix, were presented to another model AI tha… | ↗Not briefed |
arxiv arxiv-ai | 16d | Orbitol: Crystal Generation on Wyckoff Orbits | Generative models propose crystalline materials by learning composition and geometry, but unconditional sampling gives no direct contr… | ↗Not briefed |
rss techmeme | 16d | A New Mexico court orders Meta to pay $25B in penalties after a jury found Meta misled residents about its data practices in the Cambridge Analytica scandal (Jonathan Vanian/CNBC) | Jonathan Vanian / CNBC: A New Mexico court orders Meta to pay $25B in penalties after a jury found Meta misled residents about its dat… | ↗Not briefed |
hn hn-ai | 16d | Hertzon: Open-Source 3D Full-Wave FEM Solver | — | ↗Not briefed |
bluesky bluesky-ai | 16d | Anthropic S-1 projects a NY Mets World Series win in 2027: Sources | Anthropic S-1 projects a NY Mets World Series win in 2027: Sources | ↗Not briefed |
bluesky bluesky-ai | 16d | Me to Anthropic in this situation | Me to Anthropic in this situation | ↗Not briefed |
bluesky bluesky-ai | 16d | "but it grabs info from trusted sources--" no, an LLM chooses among the most statistically probable words You'd like to hear according to y… | "but it grabs info from trusted sources--" no, an LLM chooses among the most statistically probable words You'd like to hear according… | ↗Not briefed |
bluesky bluesky-ai | 16d | You bet I am! ...crying a bit too, but not for Anthropic. | You bet I am! ...crying a bit too, but not for Anthropic. | ↗Not briefed |
arxiv arxiv-ai | 16d | From Classifier Heads to Encoder Prototypes: Personalized Federated Learning with Head-Derived Clustering | Personalized federated learning must decide which clients should share representations. We introduce HDC-FL (Head-Derived Clustering f… | ↗Not briefed |
arxiv arxiv-ai | 16d | Circuits of LLM Classifiers | Language models make it easy to build classifiers by simply prompting a model with the sample to classify. However, their predictions… | ↗Not briefed |
arxiv arxiv-ai | 16d | Parallel AI Agent Collaboration as Graph-Structured Distributed Block Optimization | Parallel AI agents can produce useful local edits yet degrade a shared solution when their edits are combined. We model this coordinat… | ↗Not briefed |
arxiv arxiv-ai | 16d | Charts as Epistemic Actions for Experience-Driven Agentic Data Analysis | Data agents built on ReAct-style loops now execute complex analytical workflows over structured data, yet their intermediate reasoning… | ↗Not briefed |
hn hn-ai | 16d | Local first and OSS meeting transcriber and work log too "Merkwerk" | — | ↗Not briefed |
bluesky bluesky-ai | 16d | You're laughing. Anthropic is losing billions and you're laughing. | You're laughing. Anthropic is losing billions and you're laughing. | ↗Not briefed |
arxiv arxiv-ai | 16d | Loopholing All the Way: Language Modeling with pure Self-Conditioning | Recent Diffusion Language Models (DLMs) use Self-Conditioning (SC), or Loopholing, to improve their performance. While SC is tradition… | ↗Not briefed |
arxiv arxiv-ai | 16d | GoiBERT: Compact Vietnamese Language Models via Two-Stage Knowledge Distillation | Vietnamese pre-trained encoders have improved rapidly, but the strongest ones are large, multilingual-vocabulary models or depend on e… | ↗Not briefed |
arxiv arxiv-ai | 16d | Looped Uniform Diffusion Language Models | Uniform diffusion language models (UDLMs) generate text by iteratively denoising sequences of ordinary vocabulary tokens, any of which… | ↗Not briefed |