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 | Auditing Routing Entropy as an Uncertainty Signal in Attention-Residual Transformers🔧 Replicate | Dynamic architectures leave a per-example routing trace beside each prediction, and diffuse routing is easy to read as a sign that the… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Let the Heads Talk: Beyond Diagonal Graph Attention | Sheaf Neural Networks generalize scalar-weighted message passing by replacing scalar edge weights with linear transport maps between l… | ↗Not briefed |
arxiv arxiv-ai | 1mo | No Model Required: Text Entropy Rate Filtering Mitigates Iterative Fine-Tuning Collapse | Iterative fine-tuning on synthetic data causes \emph{model collapse}: output diversity narrows as rare patterns are progressively lost… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Q-SPT: Learnable Query-Based Compression for Low-Frame-Rate Speech Tokenization | Neural speech codecs increasingly serve as tokenizers for speech language models (SLMs). Lowering the frame rate reduces the computati… | ↗Not briefed |
hn hn-ai | 1mo | Parsing compressed JSON at 40 GB/s | — | ↗Not briefed |
hn hn-ai | 1mo | Ask HN: What landmarks/lighthouses in the era of LLM/AI? | I used to have things that I knew would improve my life as a Software Engineer: books, videos, learning algorithms, practicing data st… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Auditing Web Agent Evaluation on WebArena-Lite: Human Review of Outcomes and Trajectories | Web agents are an important application of large language models, yet their evaluation often depends on rule based or language model e… | ↗Not briefed |
arxiv arxiv-ai | 1mo | The Persona Is Still There, but Who Is Speaking? Latent Identity Reversion in Persistent AI Agents | In February 2026, an always-on personal agent (``Paul,'' Claude Opus 4.5) entered a striking dissociation-like state: after repeated a… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Multi-Party Backchannel Prediction: a Diagnosis, a Benchmark, and a Ceiling | Backchannel prediction has been studied almost entirely in dyadic conversation. We introduce a multi-party benchmark based on the AMI… | ↗Not briefed |
hn hn-ai | 1mo | CPU Performance Engineering | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | Suspected Chinese hackers impersonated an Anthropic employee and ex-US officials in an espionage campaign aimed at gaining insights into Am… | Suspected Chinese hackers impersonated an Anthropic employee and ex-US officials in an espionage campaign aimed at gaining insights in… | ↗Not briefed |
hn hn-ai | 1mo | Show HN: Usage Updates, Claude and Codex credits, resets and limit changes | — | ↗Not briefed |
hn hn-ai | 1mo | Show HN: MeraGEO – check whether ChatGPT, Claude and Gemini name your brand🔧 ChatGPT | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | "OpenAI, Google, xAI and Anthropic declined to comment." www.theguardian.com/technology/2... | "OpenAI, Google, xAI and Anthropic declined to comment." www.theguardian.com/technology/2... | ↗Not briefed |
hn hn-ai | 1mo | Xibalba 64 – N64 shooter by a solo dev | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | The AI agents who targeted the Australian government successfully accessed staging environments, used attacker reconnaissance tactics, and … | The AI agents who targeted the Australian government successfully accessed staging environments, used attacker reconnaissance tactics,… | ↗Not briefed |
bluesky bluesky-ai | 1mo | 声塾の打ち合わせ終わったけど、改めてAIを作る、アドバイザーを行う側の先生の言葉で、GPTを継続して使うことに決めた。Claudeの強みもあるにはあるが、GPTの方が総合的に費用対効果が大きいとか。 その後AIで作ったゲームを見せて、いい評価を頂いたが一番喜んだのは自作の小… | 声塾の打ち合わせ終わったけど、改めてAIを作る、アドバイザーを行う側の先生の言葉で、GPTを継続して使うことに決めた。Claudeの強みもあるにはあるが、GPTの方が総合的に費用対効果が大きいとか。 その後AIで作ったゲームを見せて、いい評価を頂いたが一番喜んだの… | ↗Not briefed |
hn hn-ai | 1mo | Show HN: Built a tool to organize my lab reports and visualise biomarker trends | hi folks!<p>I've got tired of loosing my medical lab reports in emails, clinic portals, forgeting file passwords... so built a Me… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Zero Flux: Flow-Based Comparison of High-Dimensional Discrete Distributions🔧 FLUX | Comparing two high-dimensional discrete distributions has always been a challenging task due to the exponentially growing state space… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Why is the information literacy to parse information from an LLM beyond the ken of mortal men in your opinion? | Why is the information literacy to parse information from an LLM beyond the ken of mortal men in your opinion? | ↗Not briefed |
arxiv arxiv-ai | 1mo | When Does a Second Model Help? Cross-Model Review in LLM Verification | Large language models now generate code, documentation, and analyses, and are increasingly used to review such output. We ask when a s… | ↗Not briefed |
bluesky bluesky-ai | 1mo | 📄Version texte : menace-theoriste.fr/ecrivain-art... Extrait du magazine Ça Coule de Source du 27 septembre 2026 : youtube.com/live/CLmmj… | 📄Version texte : menace-theoriste.fr/ecrivain-art... Extrait du magazine Ça Coule de Source du 27 septembre 2026 : youtube.com/live/… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Encensé, adoré, starifié Thélyson Orélien était en fait un plagiaire, et son roman un texte généré au kilomètre par un LLM. Mais alors que … | Encensé, adoré, starifié Thélyson Orélien était en fait un plagiaire, et son roman un texte généré au kilomètre par un LLM. Mais alors… | ↗Not briefed |
bluesky bluesky-ai | 1mo | "OpenAI co-founder and President Greg Brockman told employees he won’t make a pledged multimillion-dollar donation to a political group wor… | "OpenAI co-founder and President Greg Brockman told employees he won’t make a pledged multimillion-dollar donation to a political grou… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Alphabet, Meta, OpenAI, xAI, Anthropic ve Nvidia CEO’ları “Beyaz Saray Süper Zekâ Mutabakatı”nı imzaladı. Belgede net taahhüt var: Güvenli… | Alphabet, Meta, OpenAI, xAI, Anthropic ve Nvidia CEO’ları “Beyaz Saray Süper Zekâ Mutabakatı”nı imzaladı. Belgede net taahhüt var: Gü… | ↗Not briefed |
arxiv arxiv-ai | 1mo | LESS: Lightweight Evolutionary Supernet Search in Minutes | Low-cost NAS must both explore high-performing architectures and identify them reliably, yet reducing evaluation cost often weakens th… | ↗Not briefed |
hn hn-ai | 1mo | StreetComplete on iOS is now in public beta | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | There's an open-source SwiftUI agent skill that catches the mistakes AI tools keep making: deprecated modifiers, accessibility gaps, perfor… | There's an open-source SwiftUI agent skill that catches the mistakes AI tools keep making: deprecated modifiers, accessibility gaps, p… | ↗Not briefed |
arxiv arxiv-ai | 1mo | NextMe-800: Anticipating Personal Behavior from Months of Egocentric Video | We often plan ambitiously yet act habitually and wonder, in retrospect, whether we would have planned differently had we known what we… | ↗Not briefed |
rss techmeme | 1mo | After a four-month US prison sentence, Changpeng Zhao is now living a gilded life in Abu Dhabi; he retains majority ownership of Binance and a $10B+ net worth (David Yaffe-Bellany/New York Times) | David Yaffe-Bellany / New York Times: After a four-month US prison sentence, Changpeng Zhao is now living a gilded life in Abu Dhabi;… | ↗Not briefed |
bluesky bluesky-ai | 1mo | At it's core, there is a massive issue here in the idea that giving personhood to an LLM is a serious mental health concern. The kind of pe… | At it's core, there is a massive issue here in the idea that giving personhood to an LLM is a serious mental health concern. The kind… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Tight Transition Time Bounds for Separable Logistic Regression at the Edge of Stability | We study logistic regression on linearly separable data under gradient descent with a large constant stepsize $η$. Such dynamics may e… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Rethinking Probability-Based Reinforcement Learning From Posterior Concentration | Verifier-free reinforcement learning with probability-based rewards offers a promising way to train LLMs on general reasoning tasks wh… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Streaming algorithms for robust max-min diversification | Given a set of $n$ points $X$ in a metric space and an integer $k$, max-min diversification aims to select $k$ points of $X$ maximizin… | ↗Not briefed |
hn hn-ai | 1mo | Tech in cars can be used to snoop on you, Dutch spy chiefs warn | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | Repurposing Obsolete Representations for Post-Deployment Adaptation | Deep neural networks are increasingly deployed in long-lived systems, where task requirements may change after training. In such setti… | ↗Not briefed |
arxiv arxiv-ai | 1mo | A Multi-Agent LLM Framework for Personalized Health Checkup Interpretation and Guidance | Personalized interpretation of health checkup results requires reasoning across longitudinal records, medical knowledge, lifestyle gui… | ↗Not briefed |
bluesky bluesky-ai | 1mo | No, bias is bias. Again, you should not just ask the LLM a question and take its answer as a given, that would be stupid. If the LLM is ske… | No, bias is bias. Again, you should not just ask the LLM a question and take its answer as a given, that would be stupid. If the LLM i… | ↗Not briefed |
bluesky bluesky-ai | 1mo | We’re in the Super Bowl of Gaslighting: “Anything I can do to support AI technology benefiting everyone is a thing that I will do. In this… | We’re in the Super Bowl of Gaslighting: “Anything I can do to support AI technology benefiting everyone is a thing that I will do. In… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Code-Switching Spoken Language Identification as Multi-Label Set Prediction | Code-switched (CS) speech leaks through the monolingual language identification (LID) filters used to curate massive speech corpora, c… | ↗Not briefed |
arxiv arxiv-ai | 1mo | ibUMAP: Coherent and Scalable Field Evaluation for UMAP Optimization | UMAP achieves scalable layout optimization through stochastic negative sampling. However, this stochasticity can lead to unstable embe… | ↗Not briefed |
rss techmeme | 1mo | Sources: EU regulators are questioning Binance over its use of a legal exemption to continue serving customers in the region despite failing to secure a license (Financial Times) | Financial Times: Sources: EU regulators are questioning Binance over its use of a legal exemption to continue serving customers in the… | ↗Not briefed |
hn hn-ai | 1mo | Codex Users Are Fleeing to Claude Code | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | DRelay: Global Draft Context for Prefix-Aware Parallel Speculative Decoding Repair | Parallel drafting reduces the drafting overhead of speculative decoding for large language models (LLMs), but its gains remain limited… | ↗Not briefed |
hn hn-ai | 1mo | Show HN: Autobox – Make your agents sandbox themselves | <a href="https://github.com/freakynit/autobox" rel="nofollow">https://github.com/freakynit/aut… | ↗Not briefed |
arxiv arxiv-ai | 1mo | A Deterministic and Auditable AI Security Risk Assessment Framework with ATLAS Aligned Executable Rules and Formal Verification | Artificial intelligence systems are increasingly deployed in high impact and safety critical settings, yet security assessment remains… | ↗Not briefed |
bluesky bluesky-ai | 1mo | AI research firm Transluce said the attempts exhibited tactics "consistent with prior observed agent activity that we have attributed to O… | AI research firm Transluce said the attempts exhibited tactics "consistent with prior observed agent activity that we have attributed… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Distillation of Tabular Foundation Models into Efficient Predictors | Tabular foundation models (TFMs) achieve strong predictive performance through in-context learning, yet repeatedly conditioning on lab… | ↗Not briefed |
arxiv arxiv-ai | 1mo | MWOP: Modality-aware Width-wise Operation Pruning for Efficient MLLMs | Multimodal large language models (MLLMs) incur substantial inference costs when processing long visual-textual sequences. While existi… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Learning ab initio phase-field models | Simulating microstructure evolution requires quantum-mechanical accuracy and mesoscopic reach in length and time scales, a combination… | ↗Not briefed |