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 |
|---|---|---|---|---|
rss techmeme | 3d | macOS 27.3 beta warns if an app would get full device control, nudging users to "Choose Apps" for the app to control instead, brought on by AI apps' popularity (Juli Clover/MacRumors) | Juli Clover / MacRumors: macOS 27.3 beta warns if an app would get full device control, nudging users to “Choose Apps” for the app to… | ↗Not briefed |
arxiv arxiv-ai | 3d | Every Measurement, Every Direction, All at Once: Multimodal Flow Matching for Molecules and Spectra | Generative modeling in science is often data-limited, where experiments are expensive. However, each experiment may yield several coup… | ↗Not briefed |
arxiv arxiv-ai | 4d | When AUC Meets Generative Reconstruction: A Unified Principle for Balanced Visual Representation Learning | Multimodal large language models (MLLMs) require visual encoders that extract fine-grained visual features while constructing semantic… | ↗Not briefed |
hn hn-ai | 4d | Gemini 4 Hydrogen paints and animates cells frame by frame inside Google Sheets | — | ↗Not briefed |
hn hn-ai | 4d | Localix Studio – an open-source Cursor alternative for any model | — | ↗Not briefed |
hn hn-ai | 4d | Tokendoctor – the same Claude Code for up to 38% less | — | ↗Not briefed |
hn hn-ai | 4d | An Illustrated Guide to AI Agents:Concepts and Code for Building Agents with LLM | — | ↗Not briefed |
bluesky bluesky-ai | 4d | My Anthropic claim is similarly banjaxed. They seem to think it constitutes a “disagreement” when I say I’m “Caleb Crain” and a publisher l… | My Anthropic claim is similarly banjaxed. They seem to think it constitutes a “disagreement” when I say I’m “Caleb Crain” and a publis… | ↗Not briefed |
bluesky bluesky-ai | 4d | I don't think I realized that everyone gets only 70% now, and the rest has to wait until either Anthropic goes public or they get a major i… | I don't think I realized that everyone gets only 70% now, and the rest has to wait until either Anthropic goes public or they get a ma… | ↗Not briefed |
arxiv arxiv-ai | 4d | HyCoInfer: Cloud-Edge Collaborative Inference for Large Language Models on Heterogeneous GPU-FPGA Architectures🔧 vLLM | Large language model (LLM) serving relies predominantly on cloud GPUs, yet autoregressive decoding is memory-bound and often leaves th… | ↗Not briefed |
hn hn-ai | 4d | Mem Is Shutting Down | — | ↗Not briefed |
arxiv arxiv-ai | 4d | ATLAS-Finance: Benchmarking AI Agents in Realistic Banking Environments | Professional work involves under-specified assignments, where relevant information is spread across workplace applications and may con… | ↗Not briefed |
hn hn-ai | 4d | Command and Control (C2): The invisible screen behind almost every cyber attack | — | ↗Not briefed |
bluesky bluesky-ai | 4d | bald austin butler: *casually taking off jacket to reveal anthropic t-shirt* bond, let's be honest. you're a relic! this world doesn't run … | bald austin butler: *casually taking off jacket to reveal anthropic t-shirt* bond, let's be honest. you're a relic! this world doesn't… | ↗Not briefed |
hn hn-ai | 4d | Show HN: Astrum – a terminal multiplexer with integrated AI assistant | — | ↗Not briefed |
arxiv arxiv-ai | 4d | When Lower Harm Misleads: Auditing Concept Corrections | An expert who corrects a concept prediction, such as a BI-RADS descriptor in breast ultrasound, expects the diagnosis to improve. Obje… | ↗Not briefed |
arxiv arxiv-ai | 4d | BernBRS: Certified Backward Reachability for Deep Bernstein Neural Networks | Ensuring the safety of complex neural network systems requires mathematically rigorous verification. While backward reachability analy… | ↗Not briefed |
bluesky bluesky-ai | 4d | Google, SpaceX, Open AI, Anthropic - the list goes on | Google, SpaceX, Open AI, Anthropic - the list goes on | ↗Not briefed |
bluesky bluesky-ai | 4d | Here is the screenshot of the relevant info. | Here is the screenshot of the relevant info. | ↗Not briefed |
hn hn-ai | 4d | Claude's Workbench – Claude architects and reviews, local LLMs write the code | — | ↗Not briefed |
bluesky bluesky-ai | 4d | I was assured that the 10 Disagreements are all minor administrative filing issues that should be able to be dealt with easily . . . .in th… | I was assured that the 10 Disagreements are all minor administrative filing issues that should be able to be dealt with easily . . . .… | ↗Not briefed |
arxiv arxiv-ai | 4d | Segmentation-Guided Direct Capacity Estimation of Rooftop Photovoltaics from Aerial Imagery | Distribution planners need installation-level inventories of rooftop photovoltaic (PV) capacity, but interconnection records may be in… | ↗Not briefed |
arxiv arxiv-ai | 4d | Can You Undo It? Reversibility Classes for AI Agent Actions | Tool-using AI agents change files, databases, access rights and public information, yet permission to call a tool does not establish w… | ↗Not briefed |
bluesky bluesky-ai | 4d | irrespective of what any of us believe about LLM consciousness we are doing some extremely fucked up things in the training pipelines that … | irrespective of what any of us believe about LLM consciousness we are doing some extremely fucked up things in the training pipelines… | ↗Not briefed |
bluesky bluesky-ai | 4d | Anthropic Settlement again. FYI, I just saw a note from a fellow author that they called the administrators and one of them walked them th… | Anthropic Settlement again. FYI, I just saw a note from a fellow author that they called the administrators and one of them walked th… | ↗Not briefed |
bluesky bluesky-ai | 4d | 脱獄したPlayStation 4で90億パラメーターのAIモデルを毎秒最大43トークンで動作させる「Garlic LLM」が登場 https://gigazine.net/news/20261117-playstation-4-garlic-llm/ | 脱獄したPlayStation 4で90億パラメーターのAIモデルを毎秒最大43トークンで動作させる「Garlic LLM」が登場 https://gigazine.net/news/20261117-playstation-4-garlic-llm/ | ↗Not briefed |
bluesky bluesky-ai | 4d | Kubernetes is licensed permissively so it's morally fine to use it to run software on a strategic bomber. But if you train an LLM on it, th… | Kubernetes is licensed permissively so it's morally fine to use it to run software on a strategic bomber. But if you train an LLM on i… | ↗Not briefed |
arxiv arxiv-ai | 4d | MedGuardBench: A Patient-Oriented Benchmark for Personalized Guardrail Development | Safety is personal: the same action can be safe for one person but unsafe for another. Yet current guardrails judge content without co… | ↗Not briefed |
bluesky bluesky-ai | 4d | Anthropic’s mega IPO has investors scrambling for an edge, with many turning to strategies used with SpaceX’s debut to get ahead | Anthropic’s mega IPO has investors scrambling for an edge, with many turning to strategies used with SpaceX’s debut to get ahead | ↗Not briefed |
hn hn-ai | 4d | Claude Is Always Right | I think I can say, at least for me, why Claude's writing is so freaking awful, even when it's completely correct.<p>Claude w… | ↗Not briefed |
arxiv arxiv-ai | 4d | Addressing Exogenous Variability in Cooperative Multi-Agent Reinforcement Learning | Cooperative multi-agent reinforcement learning (MARL) often fails under exogenous variability such as shifts in opponents or environme… | ↗Not briefed |
bluesky bluesky-ai | 4d | Every time we hear another JND / Anthropic horror story, @mostlybree.kitrocha.com and I exchange scandalized messages because it honestly f… | Every time we hear another JND / Anthropic horror story, @mostlybree.kitrocha.com and I exchange scandalized messages because it hones… | ↗Not briefed |
arxiv arxiv-ai | 4d | Two Counterexamples for e-means | e-means (Bessa et al., 2026) combines exponential averaging and cluster revival. We give two counterexamples to its published recipe w… | ↗Not briefed |
arxiv arxiv-ai | 4d | FM-Mamba: An Efficient Native 2D Selective State Space Model with Stability-Aware Discretization | When applied to two-dimensional features such as images, Mamba-based methods either flatten the feature map into 1D sequences and fuse… | ↗Not briefed |
bluesky bluesky-ai | 4d | California sued OpenAI, arguing that a hacking incident by one of the AI start-up’s systems in July was foreseeable and alleging the compan… | California sued OpenAI, arguing that a hacking incident by one of the AI start-up’s systems in July was foreseeable and alleging the c… | ↗Not briefed |
hn hn-ai | 4d | Claude Mythos slashed software work to onboard AMD GPUs to around three days | — | ↗Not briefed |
bluesky bluesky-ai | 4d | My anthropic suit claim did not go through due to "disagreements" for 5/6 claims even though I followed every single step along the way. W… | My anthropic suit claim did not go through due to "disagreements" for 5/6 claims even though I followed every single step along the w… | ↗Not briefed |
arxiv arxiv-ai | 4d | Weight-Space Programming Doesn't Need Transformers | Standard neural architectures such as the Transformer were chosen because they train well with gradient descent and run fast on GPUs.… | ↗Not briefed |
arxiv arxiv-ai | 4d | A single inference process underlies the behavior and neural circuits of working memory | Working memory holds past inputs online for future behavior. Despite decades of research, accounts of working memory remain divided ac… | ↗Not briefed |
bluesky bluesky-ai | 4d | Thing that would be fair, but obviously isn't going to happen. If the LLM was trained on stolen material and that LLM is still in use with… | Thing that would be fair, but obviously isn't going to happen. If the LLM was trained on stolen material and that LLM is still in use… | ↗Not briefed |
arxiv arxiv-ai | 4d | RiDEx: Reward-Informed Directional Exploration for Diffusion Reinforcement Learning🔧 Stable Diffusion🔧 Elicit | Reward-based diffusion post-training updates model parameters from task rewards. While existing approaches improve sampling and policy… | ↗Not briefed |
bluesky bluesky-ai | 4d | Seriously the crimes against corporate governance that are absolutely fucking endemic in tech have to be high on the list. SpaceX, Anthropi… | Seriously the crimes against corporate governance that are absolutely fucking endemic in tech have to be high on the list. SpaceX, Ant… | ↗Not briefed |
bluesky bluesky-ai | 4d | California Attorney General Rob Bonta sued OpenAI, alleging the company failed to stop its advanced AI models from going rogue in dozens of… | California Attorney General Rob Bonta sued OpenAI, alleging the company failed to stop its advanced AI models from going rogue in doze… | ↗Not briefed |
hn hn-ai | 4d | Show HN: Fundlark-open source tool help early-career researcher find funding | I just finished my PhD and am trying to get some funding to start off. I found that opportunities are scattered and it is a pain to fi… | ↗Not briefed |
bluesky bluesky-ai | 4d | I will say that the Mythos security freakout from Anthropic, much derided at the time, and the focus on first deploying cyber models with l… | I will say that the Mythos security freakout from Anthropic, much derided at the time, and the focus on first deploying cyber models w… | ↗Not briefed |
bluesky bluesky-ai | 4d | Kalifornien verklagt OpenAI wegen Cyberangriffen #KünstlicheIntelligenz #OpenAI | Kalifornien verklagt OpenAI wegen Cyberangriffen #KünstlicheIntelligenz #OpenAI | ↗Not briefed |
arxiv arxiv-ai | 4d | DSADiff: Data Structure-Aware Diffusion for Crystal Structure Prediction via Mitigating Complexity- and Imbalance-Induced Biases | Crystal Structure Prediction (CSP) is an important step in accelerating the generation and discovery of new materials. In recent years… | ↗Not briefed |
bluesky bluesky-ai | 4d | Bank of America CEO Brian Moynihan said deals are getting postponed into 2027, with investors pointing to Anthropic's expected blockbuster … | Bank of America CEO Brian Moynihan said deals are getting postponed into 2027, with investors pointing to Anthropic's expected blockbu… | ↗Not briefed |
hn hn-ai | 4d | Show HN: Free weekly backups for Supabase Free projects (pg_dump to R2) | — | ↗Not briefed |
hn hn-ai | 4d | My Annoying Coworker Claude | — | ↗Not briefed |