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 |
|---|---|---|---|---|
hn hn-ai | 16d | HGATE – Human-Guided AI-Assisted Traceable Engineering | — | ↗Not briefed |
bluesky bluesky-ai | 16d | It does in fact seem like the promised wave of LLM-accelerated vulnerability exploits is upon us. | It does in fact seem like the promised wave of LLM-accelerated vulnerability exploits is upon us. | ↗Not briefed |
arxiv arxiv-ai | 16d | Hyperspherical Simplex Representations for Backward-Compatible Retrieval | Training modern AI models has become increasingly expensive, and updating deployed models can alter their internal feature representat… | ↗Not briefed |
bluesky bluesky-ai | 16d | I think you misunderstand what openai and other ai companies want to be. A decentralized open source approach to ai. They don't want people… | I think you misunderstand what openai and other ai companies want to be. A decentralized open source approach to ai. They don't want p… | ↗Not briefed |
hn hn-ai | 16d | Grip – a Git hook that won't let me push code I can't explain | — | ↗Not briefed |
bluesky bluesky-ai | 16d | It’s the latest in a growing wave of bets that AI companies are making on brain data. | It’s the latest in a growing wave of bets that AI companies are making on brain data. | ↗Not briefed |
bluesky bluesky-ai | 16d | hahaha VERY COOL, what can possibly go wrong? www.ft.com/content/45fb... | hahaha VERY COOL, what can possibly go wrong? www.ft.com/content/45fb... | ↗Not briefed |
arxiv arxiv-ai | 16d | PACE: Periodic and Channel-Aligned Encoding of Four-Bit Recurrent States | Linear-attention language models keep their context in a recurrent state of fixed size, where an inference system stores one copy of t… | ↗Not briefed |
bluesky bluesky-ai | 16d | As OpenAI executives step up their public calls for federal AI safety rules, a political group the company’s president helped fund is spend… | As OpenAI executives step up their public calls for federal AI safety rules, a political group the company’s president helped fund is… | ↗Not briefed |
bluesky bluesky-ai | 16d | Anthropic, music labels to make key arguments in AI training case reut.rs/488cAFO | Anthropic, music labels to make key arguments in AI training case reut.rs/488cAFO | ↗Not briefed |
rss techmeme | 16d | Dario Amodei says Anthropic's "race to the top" makes AI safer, cites nine safety decisions with "substantial costs", and admits mishandling a draft AI bill (Dario Amodei/@darioamodei) | Dario Amodei / @darioamodei: Dario Amodei says Anthropic's “race to the top” makes AI safer, cites nine safety decisions with “substan… | ↗Not briefed |
rss wired-ai | 16d | OpenAI Invests in Startup That Uses Brain Implants to Detect Cancer | It’s the latest in a growing wave of bets that AI companies are making on brain data. | ↗Not briefed |
youtube youtube-wesroth | 16d | OpenAI’s “Alien Math” Is Freaking People Out🔧 ChatGPT | The math apocalypse has begun. Scott Aaronson calls this the biggest day in mathematical history. A global coalition of mathematicians… | ↗Not briefed |
bluesky bluesky-ai | 16d | So the 80 billion Anthropic loss appears to be two things: 1) an 8 billion dollar actual loss 2) an increase in debt of $$72 billion based… | So the 80 billion Anthropic loss appears to be two things: 1) an 8 billion dollar actual loss 2) an increase in debt of $$72 billion… | ↗Not briefed |
arxiv arxiv-ai | 16d | Stochastic Pruning of the DPO Computation Graph: 1.4x Faster Training and Lower Peak Memory | Direct preference optimization (DPO) incurs training costs in output scoring, sequence execution, and backpropagation, so reducing sup… | ↗Not briefed |
bluesky bluesky-ai | 16d | Anthropic lost $80 billion? It's like they say: you have to break a few eggs to fake an omelette. | Anthropic lost $80 billion? It's like they say: you have to break a few eggs to fake an omelette. | ↗Not briefed |
arxiv arxiv-ai | 16d | Block Diffusion for Longitudinal EHR: Accuracy, Efficiency, and Representation Trade-offs | Autoregressive (AR) models have proved effective for zero-shot prediction from longitudinal electronic health records, but impose stri… | ↗Not briefed |
bluesky bluesky-ai | 16d | So Anthropic lost $80billion in six months. How insane is that?. Well it makes it the fifth biggest company loss in history. In only six mo… | So Anthropic lost $80billion in six months. How insane is that?. Well it makes it the fifth biggest company loss in history. In only s… | ↗Not briefed |
arxiv arxiv-ai | 16d | Phase Transitions in Learning Driven by Extremal Data | We study an extremal regression problem, an example of multiple instance learning (MIL), where the predictors are given as multisets,… | ↗Not briefed |
rss techmeme | 16d | Nvidia unveils the DSX POD, a modular containerized version of its Vera Rubin reference architecture to cut data center deployment from six months to 14 days (Charlotte Trueman/DatacenterDynamics) | Charlotte Trueman / DatacenterDynamics: Nvidia unveils the DSX POD, a modular containerized version of its Vera Rubin reference archit… | ↗Not briefed |
bluesky bluesky-ai | 16d | and most of the new internet has accrued lifetimes of technical debt, which would take 61 billion days of dev work to fix as of 2025. and t… | and most of the new internet has accrued lifetimes of technical debt, which would take 61 billion days of dev work to fix as of 2025.… | ↗Not briefed |
arxiv arxiv-ai | 16d | Residual-Based Causal Discovery via Pretrained In-Context Quantile Regression | Residual-based causal discovery assumes that each variable is generated from its parents and an independent noise source. It estimates… | ↗Not briefed |
bluesky bluesky-ai | 16d | No wonder tech billionaires are funding Reform UK so people will blame migrants for poverty & unemployment when "The business model of AI … | No wonder tech billionaires are funding Reform UK so people will blame migrants for poverty & unemployment when "The business model o… | ↗Not briefed |
bluesky bluesky-ai | 16d | 'Bijna alle problemen waar ik aan werkte zijn verwezenlijkt. (..) Opvallendste aan de oplossingen van OpenAI is dat ‘wij’ hierin geen rol s… | 'Bijna alle problemen waar ik aan werkte zijn verwezenlijkt. (..) Opvallendste aan de oplossingen van OpenAI is dat ‘wij’ hierin geen… | ↗Not briefed |
arxiv arxiv-ai | 16d | Are Joint-Embedding Predictive Objectives Effective for Longitudinal EHRs? | Masked language modelling (MLM) is the dominant pretraining objective for transformer encoders applied to longitudinal electronic heal… | ↗Not briefed |
arxiv arxiv-ai | 16d | RCSD: Reinforcement and Continual Learning for Domain-Specific Speculative Decoding | Speculative decoding accelerates large language model inference by employing a compact draft model to propose candidate tokens that ar… | ↗Not briefed |
arxiv arxiv-ai | 16d | Geometric knowledge spaces, Attention pairing and Architectures | In this paper, we propose a geometric approach to deriving attention mechanisms and exploring potential new transformer architectures.… | ↗Not briefed |
bluesky bluesky-ai | 16d | Once more for the back row. Stop giving your LLM shell tools and unrestricted internet access and something like 80% of your agent securit… | Once more for the back row. Stop giving your LLM shell tools and unrestricted internet access and something like 80% of your agent se… | ↗Not briefed |
arxiv arxiv-ai | 16d | pplx-embed-v2-late Technical Report | In this report, we introduce pplx-embed-v2-late, a pair of 0.6B and 9B multilingual and multimodal late-interaction models. Both model… | ↗Not briefed |
rss techmeme | 16d | China criticizes the US DHS's DeepSeek ban, vowing "resolute countermeasures" while rejecting allegations from the FBI and AI labs of illicit distillation (Ephrat Livni/New York Times) | Ephrat Livni / New York Times: China criticizes the US DHS's DeepSeek ban, vowing “resolute countermeasures” while rejecting allegatio… | ↗Not briefed |
bluesky bluesky-ai | 16d | I mean... "To put this in perspective, for every dollar Anthropic earned in 2025, it committed roughly $113 to future infrastructure. " I… | I mean... "To put this in perspective, for every dollar Anthropic earned in 2025, it committed roughly $113 to future infrastructure.… | ↗Not briefed |
bluesky bluesky-ai | 16d | "i asked [LLM]" has become common parlance at work and it's genuinely quite upsetting | "i asked [LLM]" has become common parlance at work and it's genuinely quite upsetting | ↗Not briefed |
arxiv arxiv-ai | 16d | Beyond the Vote: When Do LLMs Add Value to Retrieval-Based Tabular Classification?🔧 Consensus | Retrieval-based in-context learning (ICL) enables large language models (LLMs) to predict tabular labels from similar examples, but it… | ↗Not briefed |
arxiv arxiv-ai | 16d | Supervision-Adaptive Curriculum and Guarded Anti-Curriculum Learning for Dense Prediction🔧 Modal | Curriculum learning orders samples by difficulty, but difficulty alone does not establish supervision reliability for dense prediction… | ↗Not briefed |
arxiv arxiv-ai | 16d | Mitigating Query Shortcuts in Chain-of-Thought with Context-Aware Decoding | While Chain-of-Thought (CoT) prompting improves the reasoning performance of large language models (LLMs), recent studies have shown t… | ↗Not briefed |
arxiv arxiv-ai | 16d | BRIDGE: An Acceptance Gate to Block False Improvements in Model Updates, with a Retrospective Clinical Evaluation | Models are updated repeatedly, and an update that looks better on validation data but not on new data is a false improvement that can… | ↗Not briefed |
arxiv arxiv-ai | 16d | A unified approach for estimation of unnormalized models via Stein operators | In this paper, we develop a unified framework connecting proper scoring rules and Stein's method for the estimation of unnormalized mo… | ↗Not briefed |
bluesky bluesky-ai | 16d | 今日職場で聞いて面白かったのが、「LLMが出力する文章のLLMくささを消すことを「消臭」と呼ぶ」「最近のLLM界隈のトレンドの1つが消臭」っていうやつ。わざわざ「LLMくさいから使ってほしくないNGワードリスト」とかを作って渡すらしい。もはや書いたほうが早くね？ | 今日職場で聞いて面白かったのが、「LLMが出力する文章のLLMくささを消すことを「消臭」と呼ぶ」「最近のLLM界隈のトレンドの1つが消臭」っていうやつ。わざわざ「LLMくさいから使ってほしくないNGワードリスト」とかを作って渡すらしい。もはや書いたほうが早くね？ | ↗Not briefed |
hn hn-ai | 16d | Extension for Better VS Code Terminal Shortcuts | — | ↗Not briefed |
rss stratechery | 16d | Netflix Earnings, Business or Franchise, Disney’s Deal | Netflix has a solid business, which is exactly the problem: the company increasingly seems like a business and not a franchise, at lea… | ↗Not briefed |
arxiv arxiv-ai | 16d | Shortcut-Resistant CAM Distillation for Long-Tailed Recognition | Real-world datasets often follow a long-tailed distribution, making generalization to tail classes difficult. We revisit this problem… | ↗Not briefed |
arxiv arxiv-ai | 16d | Keeping Time vs. Understanding Time: Neither Alone Explains Temporal Reasoning in LLMs | Temporal reasoning is widely assumed to decompose into two modes: temporal arithmetic (date and time calculations), and temporal seman… | ↗Not briefed |
arxiv arxiv-ai | 16d | Asynchronous, Generation-Free ABC: Streaming AMIS Reweighting for Heterogeneous Simulator Workloads on HPC | Approximate Bayesian computation (ABC) enables Bayesian inference for models with intractable likelihoods by estimating parameter post… | ↗Not briefed |
arxiv arxiv-ai | 16d | Rethinking Agentic Model Orchestration for Time Series Forecasting | Time series foundation models (TSFMs) perform unevenly across forecasting tasks. Recent work uses agentic model orchestration (AMO) to… | ↗Not briefed |
rss techmeme | 16d | Anthropic and Citrini Research use an internal Claude model to autonomously generate protein minibinders against five scorpion neurotoxins, testing 211 designs (Anthropic) | Anthropic: Anthropic and Citrini Research use an internal Claude model to autonomously generate protein minibinders against five scorp… | ↗Not briefed |
arxiv arxiv-ai | 16d | Parallel Thinking for Large Language Model with Path Diversity | Parallel thinking has emerged as a promising paradigm for enhancing the reasoning capabilities of Large Language Models (LLMs) by expl… | ↗Not briefed |
hn hn-ai | 16d | Show HN: Linkit – AI bookmark manager, 100% free and open source | — | ↗Not briefed |
arxiv arxiv-ai | 16d | A Strong MoE Is All You Need for Process Supervision | Process reward models (PRMs) provide fine-grained supervision for multi-step reasoning, yet obtaining reliable process labels remains… | ↗Not briefed |
arxiv arxiv-ai | 16d | An Exact 93-Move Solution to a Nonlocal Crystal-Exchange Puzzle | We determine the minimum number of directional commands needed to solve alevel of Order of the Sinking Star, modeled as a crystal-exch… | ↗Not briefed |
arxiv arxiv-ai | 16d | RubricOpt: Self-Revision of Semantic Rubrics for Preference Optimization | A model's failures can reveal gaps in the behavioral criteria guiding its training. We introduce RubricOpt, a framework that lets a po… | ↗Not briefed |