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 | Old Ideas, Novel Problems: The Instability of LLM-Based Novelty Evaluation | Automated ideation systems are often evaluated on the novelty of the ideas they produce, and that judgment is increasingly delegated t… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Task-Adaptive Grounded 3D-Programmers Using 2D VLMs | Recent vision-language models (VLMs) exhibit remarkable generalization and reasoning abilities, yet 3D understanding in these models i… | ↗Not briefed |
bluesky bluesky-ai | 1mo | HuggingFace having to use GLM to troubleshoot the OpenAI hack is just the beginning. If HF had faced prosecution for doing so, we might all… | HuggingFace having to use GLM to troubleshoot the OpenAI hack is just the beginning. If HF had faced prosecution for doing so, we migh… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Controllable Multi-label Video Safety Detection via Adaptive Tversky Policy Optimization | The rapid growth of video-based social media has increased users' exposure to harmful content, creating a need for reliable automated… | ↗Not briefed |
hn hn-ai | 1mo | Maven Central Publishing Limits | — | ↗Not briefed |
hn hn-ai | 1mo | Why do OpenAI's GPT-2 weights beat mine? Part five: data quality | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | On Language Drift during RLVR Post-Training | Recent advances in LLM reasoning models---driven primarily by the paradigm of post-training via reinforcement learning with verifiable… | ↗Not briefed |
arxiv arxiv-ai | 1mo | BranchIP: Learning Adaptive Equivariant Computation for Interatomic Potentials | Equivariant machine learning interatomic potentials (MLIPs) have revolutionized atomistic modeling, but accurate treatment of complex… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Atoms to Processes: The Role of Artificial Intelligence and Machine Learning in Chemical Engineering | The rapid maturation of artificial intelligence (AI) and machine learning (ML) has catalyzed a profound shift in how chemical engineer… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Bellman Meets Lyapunov: Unsupervised Reinforcement Learning via Mastering Chaos | Reinforcement learning (RL) is a powerful paradigm for training agents, yet its success rests on domain expertise of human engineers w… | ↗Not briefed |
hn hn-ai | 1mo | Show HN: Opus6.video – Make Claude Opus 5.5 videos without a Claude account | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | Exploring Weaknesses of Generative Image Watermarks against Latent Frequency Masking | Invisible watermarking has become a central tool for tracing AI-generated images, but its robustness against adaptive removal attacks… | ↗Not briefed |
bluesky bluesky-ai | 1mo | my "Anthropic isn't good but OpenAI is the actual biblical devil" thesis continues to collect dubs (and xAI is, as its bot will sometimes … | my "Anthropic isn't good but OpenAI is the actual biblical devil" thesis continues to collect dubs (and xAI is, as its bot will somet… | ↗Not briefed |
rss ai-snake-oil | 1mo | A big-tent or small-tent AI safety movement? | The unstated disagreement that underpins safety debates | ↗Not briefed |
arxiv arxiv-ai | 1mo | Counting Moves, Weighing Voices: Bayesian Dialectical Argumentation for Calibrated Multi-LLM Councils under Persistent Adversaries | A multi-LLM \emph{council} lets several large language models (LLMs) deliberate on a question and return an answer together with a con… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Three AI safety researchers at OpenAI have left the company. OpenAI says it was because they allegedly shared information with a third-par… | Three AI safety researchers at OpenAI have left the company. OpenAI says it was because they allegedly shared information with a thir… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Mem++: Non-Destructive Memory for Long-Term Organizational LLM Agents | Large Language Model (LLM) agents now take part in organizational work, where many authors record decisions across documents over mont… | ↗Not briefed |
bluesky bluesky-ai | 1mo | OpenAI says it "parted ways" with three staffers for violating its policies on "handling sensitive" info; sources: they shared it with an A… | OpenAI says it "parted ways" with three staffers for violating its policies on "handling sensitive" info; sources: they shared it with… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Mingbird: A Local-First Agent Harness Enabling Small Open Models to Complete Real Tasks🔧 Ollama | Small open-weight models (2-9B) run on ordinary laptops, but under cloud-scale agent harnesses they rarely complete real tasks: tool p… | ↗Not briefed |
rss techmeme | 1mo | OpenAI says it "parted ways" with three staffers for violating its policies on "handling sensitive" info; sources: they shared it with an AI safety organization (Wall Street Journal) | Wall Street Journal: OpenAI says it “parted ways” with three staffers for violating its policies on “handling sensitive” info; sources… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Weather-Aware Domain Adaptation for Street-View Weather Recognition | Adverse conditions such as rain, snow, fog, and dust remain challenging for camera-based perception in autonomous driving. We study mu… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Can AI Oversight Be Zero Knowledge? | AI systems increasingly produce outputs from confidential data, such as a fitness-for-duty assessment from medical records or the pred… | ↗Not briefed |
hn hn-ai | 1mo | GPT-6 Astra Reviews 'The Social Roots of Delusions' | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | Yeah lol, like it looks like they had an llm do a bunch of statistical analysis for them (nothing wrong with that) and then interpreted the… | Yeah lol, like it looks like they had an llm do a bunch of statistical analysis for them (nothing wrong with that) and then interprete… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Comparing a gradient boosting algorithm to the GOES FDC for wildfire detection | Wildfires pose severe risks to human life, ecosystems, and property. This study presents a machine learning approach for wildfire dete… | ↗Not briefed |
bluesky bluesky-ai | 1mo | the way to fight with home assistant is to have an LLM do it for you. on the rest… (shakes head sadly) | the way to fight with home assistant is to have an LLM do it for you. on the rest… (shakes head sadly) | ↗Not briefed |
bluesky bluesky-ai | 1mo | if you're wondering how the tech field is going apparently someone built a "torture box" for an LLM where they basically just tell an LLM t… | if you're wondering how the tech field is going apparently someone built a "torture box" for an LLM where they basically just tell an… | ↗Not briefed |
hn hn-ai | 1mo | Show HN: Anywhich, a which that also checks your package managers | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | Universal Byte-Level Encoding: UTF-8/UTF-16 Routing to Reduce Cross-Script Token-Budget Disparities | Byte-level byte-pair encoding (BBPE) tokenizers are attractive for multilingual large language models (LLMs) because they cover all Un… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Universal interpolation for deep residual self-attention networks | Universal approximation is a necessary qualitative property of learning architectures to benefit from scaling laws. While it is generi… | ↗Not briefed |
arxiv arxiv-ai | 1mo | The Curvature of Regret in Contextual Linear Optimization | Decision-focused learning for linear optimization is complicated by the discontinuity of the optimizer, where small cost errors may le… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Handsome Model (AI) | Handsome Model (AI) | ↗Not briefed |
bluesky bluesky-ai | 1mo | Reading some student work that is pretty flawed but I love it because it is idiosyncratic and interesting, and has not been within a millio… | Reading some student work that is pretty flawed but I love it because it is idiosyncratic and interesting, and has not been within a m… | ↗Not briefed |
hn hn-ai | 1mo | Migrating live Stripe subscriptions to a new catalogue without changing prices | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | Sim+Real: Joint Simulation - Experiment Training Improves Balanced Prediction in Physical Systems | Simulation and experimental measurements provide complementary data for learning spatiotemporal physical systems, but standard simulat… | ↗Not briefed |
hn hn-ai | 1mo | OpenAI Parts Ways with Researchers Who Allegedly Shared Confidential Information | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | OpenAI canned three safety researchers for sharing data with "a third-party AI-safety organization"; presumably this was METR? | OpenAI canned three safety researchers for sharing data with "a third-party AI-safety organization"; presumably this was METR? | ↗Not briefed |
hn hn-ai | 1mo | Extract v2.5: Rebuilding extraction agents on LlamaParse | — | ↗Not briefed |
hn hn-ai | 1mo | Decades After Watergate, Justice Department Reviews Whether Nixon Got a Raw Deal | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | FastCI: Efficient GPU-Intensive CI for LLM Training Frameworks | As large language models (LLMs) keep growing in size and complexity, their training frameworks evolve at a rapid pace as well. Therefo… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Counterfactual Auditing of Bias in Open-Source Large Language Models for Clinical Triage | Emergency department (ED) triage is a high-stakes prioritization task in which demographic, socioeconomic, and system-context informat… | ↗Not briefed |
arxiv arxiv-ai | 1mo | SIEVE: Selective attention-value Suppression for Vision-Language Models Unlearning | The ability of vision-language models (VLMs) to associate visual identities with biographical information creates a need for selective… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Olmo-core 3 tackles MoE scaling issues Allen AI released Olmo-core 3, a framework for large MoE model training designed to scale efficient…🔧 Hugging Face | Olmo-core 3 tackles MoE scaling issues Allen AI released Olmo-core 3, a framework for large MoE model training designed to scale effi… | ↗Not briefed |
bluesky bluesky-ai | 1mo | We’re open for general submissions as of today. This opportunity will close when we reach 400 submissions, & I expect we’ll meet that numbe… | We’re open for general submissions as of today. This opportunity will close when we reach 400 submissions, & I expect we’ll meet that… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Training-Free Diffusion Planning with Analytical Local Scores🔧 Modal | Path finding and multi-robot motion planning require trajectories that are smooth, goal-directed, and collision-free in environments w… | ↗Not briefed |
bluesky bluesky-ai | 1mo | There’s been a lot of hand-wringing about model misalignment in the AI industry lately, and Google claims it designed Argon with systems th… | There’s been a lot of hand-wringing about model misalignment in the AI industry lately, and Google claims it designed Argon with syste… | ↗Not briefed |
rss techmeme | 1mo | Memo: Ryan Roslansky, previously LinkedIn CEO who became Office and Teams head in 2025, is leaving after ~18 years at Microsoft, triggering an exec reshuffle (Tom Warren/The Verge) | Tom Warren / The Verge: Memo: Ryan Roslansky, previously LinkedIn CEO who became Office and Teams head in 2025, is leaving after ~18 y… | ↗Not briefed |
hn hn-ai | 1mo | Lightweight PDF parser with layout, tables, formulas and bounding boxes | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | I'm not as negative on the tech as most but god damn are Anthropic and OpenAI terribly run companies being entirely propped up by fantasy m… | I'm not as negative on the tech as most but god damn are Anthropic and OpenAI terribly run companies being entirely propped up by fant… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Do Your Own Research: Learning to Forecast by Learning to Search | Outcome-based reinforcement learning can train language models to forecast real-world events, but prior forecasting work either freeze… | ↗Not briefed |