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 |
|---|---|---|---|---|
bluesky bluesky-ai | 29d | Meanwhile, at anthropic | Meanwhile, at anthropic | ↗Not briefed |
rss techmeme | 29d | Anthropic co-founder Jack Clark says "I now think it's more likely we'll see full RSI become possible in 2027", earlier than his previous prediction of 2028 (Jack Clark/Import AI) | Jack Clark / Import AI: Anthropic co-founder Jack Clark says “I now think it's more likely we'll see full RSI become possible in 2027”… | ↗Not briefed |
hn hn-ai | 29d | Show HN: Surprisal – a language identifier in ~250 lines of C. No ML,no training | I made a text/language identifier today, in a single C file, with no ML.<p>It classifies a text by checking which reference text… | ↗Not briefed |
arxiv arxiv-ai | 29d | Rethinking Projector Training in Multimodal LLMs | Multimodal LLMs are typically trained by first aligning a vision encoder with a language model via projector training on large image-c… | ↗Not briefed |
arxiv arxiv-ai | 29d | PhysicsSliders: Learning Physical Controls for Audio Foley Generation | Physical properties provide intuitive controls for sounds made by interacting objects. The challenge is to make a generative model res… | ↗Not briefed |
arxiv arxiv-ai | 29d | ACCRUE: Continual Off-Policy Reinforcement Learning for Robot Foundation Models | Reinforcement learning (RL) post-training improves robot foundation models (RFMs) beyond their demonstrations, but real-world deployme… | ↗Not briefed |
bluesky bluesky-ai | 29d | Even so, I fail to see any qualitative difference between LLM training and, just, a writer reading a lot of books and being influenced by t… | Even so, I fail to see any qualitative difference between LLM training and, just, a writer reading a lot of books and being influenced… | ↗Not briefed |
arxiv arxiv-ai | 29d | Autonomous Cyber Operations Gym for Training and Evaluating AI Models for Defence | This paper proposes an Autonomous Cyber Operations (ACO) gym, a unified framework integrating high-fidelity network emulation, realist… | ↗Not briefed |
bluesky bluesky-ai | 29d | Has Anthropic considered...turning it off? | Has Anthropic considered...turning it off? | ↗Not briefed |
arxiv arxiv-ai | 29d | Fairness under Prediction Error in Predict-then-Optimize Pipelines | Fairness constraints in predict-then-optimize pipelines are necessarily enforced on predicted outcomes $\hat{y}$, since true outcomes… | ↗Not briefed |
hn hn-ai | 29d | Slopcore – A Guide to Claudia | — | ↗Not briefed |
bluesky bluesky-ai | 29d | ME: "So, what can the new Muse Wonder model do with Muse Spark as a collector for Facebook data?" AI: "You can't use Claude for that." ME: … | ME: "So, what can the new Muse Wonder model do with Muse Spark as a collector for Facebook data?" AI: "You can't use Claude for that."… | ↗Not briefed |
hn hn-ai | 29d | Redesigning Source Control at Anthropic | — | ↗Not briefed |
arxiv arxiv-ai | 29d | Decomposing and Predicting Annotation Disagreement in Harmful Content Across Languages | Annotator disagreement is increasingly recognized as a meaningful property of subjective tasks, rather than merely noise to be removed… | ↗Not briefed |
arxiv arxiv-ai | 29d | GeoMER: A Joint Euclidean-Hyperbolic Geometry Framework for Multimodal Emotion Recognition🔧 Modal | Multimodal Emotion Recognition (MER) aims to understand human emotions by integrating information from multiple modalities such as spe… | ↗Not briefed |
arxiv arxiv-ai | 29d | Compare to Correct: Pairwise-Guided Logit Adjustment for LLM Test-Time Scaling | Sampling-based test-time scaling enhances LLM reasoning by aggregating multiple reasoning paths for the same query. Existing methods u… | ↗Not briefed |
arxiv arxiv-ai | 29d | Reading Right, Acting Wrong: Diagnosing Decision Failures of System One Models in Agent Loops | System One models choose an agent's next action by scoring a supplied set of options in a single pass. This makes them fast and cheap,… | ↗Not briefed |
arxiv arxiv-ai | 29d | When More Objectives Make Identification Easier: The Role of Preferences and Feedback | Best-arm identification (BAI) aims to identify the optimal arm using as few samples as possible. While most existing work focuses on a… | ↗Not briefed |
hn hn-ai | 29d | Launch HN: Consus (YC F24) – LLM gateway for companies handling sensitive data🔧 OpenRouter🔧 GPT-5🔧 ChatGPT | Hey Hacker News! I'm Eric, the founder of Consus. We route all AI requests to ITAR restricted and controlled unclassified informa… | ↗Not briefed |
bluesky bluesky-ai | 29d | 1. okay but bender did, it's in the op 2. do you think llm use is equivalent to any of those analogies. do you think they are analogous in … | 1. okay but bender did, it's in the op 2. do you think llm use is equivalent to any of those analogies. do you think they are analogou… | ↗Not briefed |
arxiv arxiv-ai | 29d | The Many Morals of AI: The cautionary bias of North American news narratives in global perspective | The stories told about artificial intelligence (AI) play a central role in shaping public understanding, policy debates, and expectati… | ↗Not briefed |
arxiv arxiv-ai | 29d | LLM Agents for Automated Post-Training via Structured Recipe Optimization🔧 GPT-5 | Automated post-training transforms pre-trained large language models (LLMs) into models capable of handling diverse domains by automat… | ↗Not briefed |
arxiv arxiv-ai | 29d | Order-Consistent Latent Geometry for Reliable Short-Horizon Planning in World Models | Latent planners often rank candidate action sequences by terminal Euclidean distance to a goal latent. Exact agreement with transition… | ↗Not briefed |
arxiv arxiv-ai | 29d | CREATE: Creativity Evaluation of VLMs using the Visual Remote Associates Test | Vision language models (VLMs) are used in creative tasks such as visual storytelling, design ideation, art interpretation, and adverti… | ↗Not briefed |
arxiv arxiv-ai | 29d | Configuring RAG Systems: What Breaks, Where Originates, and How to Fix | Retrieval-Augmented Generation (RAG) has emerged as a dominant paradigm for improving Large Language Model (LLM) by intertwining it wi… | ↗Not briefed |
arxiv arxiv-ai | 29d | Rethinking Quality in Large Language Model Routing: Security-Conditioned Model-and-Budget Selection | Large language model (LLM) routers reduce serving cost by choosing which model to call and how many output tokens to grant it for each… | ↗Not briefed |
hn hn-ai | 29d | Factories are automated. Their back offices are stuck in 1995 | — | ↗Not briefed |
bluesky bluesky-ai | 29d | 🚨NEW🚨 An OpenAI model available to the public was used to generate three zero-days in Chrome, chained to craft an exploit that got the ma… | 🚨NEW🚨 An OpenAI model available to the public was used to generate three zero-days in Chrome, chained to craft an exploit that got t… | ↗Not briefed |
bluesky bluesky-ai | 29d | The developer of chatbot Claude warned that rogue 'AI worms' will soon be able to hijack smart devices such as ovens and TVs. ft.trib.al/DQ… | The developer of chatbot Claude warned that rogue 'AI worms' will soon be able to hijack smart devices such as ovens and TVs. ft.trib.… | ↗Not briefed |
arxiv arxiv-ai | 29d | Linear Last-Iterate Convergence Rates for Optimistic Learning in Harmonic Games | We prove that the last iterate of Optimistic Follow-The-Regularized-Leader (OFTRL) converges at a linear rate to a Nash equilibrium in… | ↗Not briefed |
bluesky bluesky-ai | 29d | We need to win these mid terms and we need to put an immediate moratorium on AI development. We need serious regulations aimed straight at … | We need to win these mid terms and we need to put an immediate moratorium on AI development. We need serious regulations aimed straigh… | ↗Not briefed |
arxiv arxiv-ai | 29d | Last-Iterate Convergence of Clipped SGD with Linear Decay | In this paper, we study the last-iterate convergence of clipped stochastic gradient descent under heavy-tailed noise. We assume that t… | ↗Not briefed |
bluesky bluesky-ai | 29d | I don’t think using an LLM to align a div or look up a restaurant is morally equivalent to consuming CSAM | I don’t think using an LLM to align a div or look up a restaurant is morally equivalent to consuming CSAM | ↗Not briefed |
arxiv arxiv-ai | 29d | Unbalanced W-Flow: Unbalanced Wasserstein Drifting under Finite-Batch Marginal Fluctuations | Teacher-free one-step generative models such as W-Flow learn directly from distributional objectives, with the Sinkhorn divergence pro… | ↗Not briefed |
rss the-decoder | 29d | Character.ai will shut down if no buyer emerges, capping a fall from $2.7 billion Google deal to bankruptcy🔧 Character.AI | Character.ai has filed for bankruptcy and is looking for a buyer. If none steps up, the platform shuts down. Founded by ex-Google rese… | ↗Not briefed |
bluesky bluesky-ai | 29d | Absolutely insane program at Anthropic bribing political scientists and economists to generate RL environments. It's a trap. (Of course, … | Absolutely insane program at Anthropic bribing political scientists and economists to generate RL environments. It's a trap. (Of cou… | ↗Not briefed |
bluesky bluesky-ai | 29d | You don't even need to use an LLM to not be wrong here. You could just trust a better class of person. But it certainly helps. | You don't even need to use an LLM to not be wrong here. You could just trust a better class of person. But it certainly helps. | ↗Not briefed |
arxiv arxiv-ai | 29d | Understanding Circulant Permutation to Extend the CMinHash Estimator | CMinHash promises a new direction for designing MinHash algorithms with reduced variance and storage space via circulant permutations.… | ↗Not briefed |
bluesky bluesky-ai | 29d | Using an LLM isn't risky in this way, and lack of direct experience with LLM's is a reason why you see people make claims about what it can… | Using an LLM isn't risky in this way, and lack of direct experience with LLM's is a reason why you see people make claims about what i… | ↗Not briefed |
hn hn-ai | 29d | Show HN: pg_vectorsearch: Scalable Open-Source Vector Search in PostgreSQL | — | ↗Not briefed |
arxiv arxiv-ai | 29d | Input-Conditioned Search for Efficient Test-Time Scaling of Recursive Reasoners | People often tackle difficult problems by thinking more deeply or exploring more possibilities. Similarly, recursive reasoners, which… | ↗Not briefed |
bluesky bluesky-ai | 29d | I have no doubt in my mind that an LLM could do 30% or more of a middle manager's job, so it's easy to dazzle them with the shiny keys of s… | I have no doubt in my mind that an LLM could do 30% or more of a middle manager's job, so it's easy to dazzle them with the shiny keys… | ↗Not briefed |
bluesky bluesky-ai | 29d | OpenAIに数学者たちが反発。研究成果の公開方法を巡り不信感が高まる│2026.10.12 wired.jp/article/open... "ニューヨーク大学（NYU）の客員数学教授ネストル・ギレンは、「数学者のあいだには、AI企業がまるでマフィアのように振る舞っていると… | OpenAIに数学者たちが反発。研究成果の公開方法を巡り不信感が高まる│2026.10.12 wired.jp/article/open... "ニューヨーク大学（NYU）の客員数学教授ネストル・ギレンは、「数学者のあいだには、AI企業がまるでマフィアのように振る舞… | ↗Not briefed |
bluesky bluesky-ai | 29d | Oh, wait, your data was compiled with Claude AI. Whelp. | Oh, wait, your data was compiled with Claude AI. Whelp. | ↗Not briefed |
arxiv arxiv-ai | 29d | Beyond the Encoder: Object-Centric Representations Emerging from Segmentation Foundation Models | Foundation models are increasingly used as feature extractors for downstream tasks with limited data, typically through representation… | ↗Not briefed |
hn hn-ai | 29d | A New GPU, from Facebook? [video] | — | ↗Not briefed |
arxiv arxiv-ai | 29d | Empathy Is Not the Scapegoat for Declines in Factuality and Reliability | Language models are increasingly expected to support users in stressful or uncertain situations, where warmth and empathy are desirabl… | ↗Not briefed |
hn hn-ai | 29d | Dave Eddy's "Reverie Amidst Ruins" Streams Kitty Protocol Graphics | — | ↗Not briefed |
arxiv arxiv-ai | 29d | On Symmetry for Characterizing Agent Heterogeneity in Multi-Agent Systems | Understanding agent heterogeneity can inform the design of shared and agent-specific policy representations, however existing criteria… | ↗Not briefed |
hn hn-ai | 29d | Researchers Transform Old Blood Cells into Young Brain Cells | — | ↗Not briefed |