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 | 3d | EMG Tokenization Preserves Movement Information at a Fraction of the Bitrate | Wireless electromyography (EMG) interfaces must balance electrode coverage against transmission bandwidth. Here, we show that compact… | ↗Not briefed |
hn hn-ai | 3d | New previously unseen videos from Tiananmen Square | — | ↗Not briefed |
bluesky bluesky-ai | 3d | You software? I software too! If your meeting notes don't give you a little ego boost, you should try running them through an Anthropic mod… | You software? I software too! If your meeting notes don't give you a little ego boost, you should try running them through an Anthropi… | ↗Not briefed |
bluesky bluesky-ai | 3d | Not someone having an llm response (second person singular) in their resume text that they obviously copy and pasted into their resume and … | Not someone having an llm response (second person singular) in their resume text that they obviously copy and pasted into their resume… | ↗Not briefed |
bluesky bluesky-ai | 3d | There is one person in the main group I'm in that developed his own LLM using only his data/sessions (with informed consent) and the machin… | There is one person in the main group I'm in that developed his own LLM using only his data/sessions (with informed consent) and the m… | ↗Not briefed |
hn hn-ai | 3d | Show HN: Loudwalk – Review your Anki deck out loud on a walk | Hey HN,<p>Loudwalk (<a href="https://loudwalk.com/?from=hn" rel="nofollow">https://loudwalk.com/?from=hn… | ↗Not briefed |
arxiv arxiv-ai | 3d | Do Language Models Trust Their Own Justifications? A Study on Functional Consistency | Large Language Models (LLMs) have been widely adopted in text classification tasks, where they not only output class predictions but a… | ↗Not briefed |
bluesky bluesky-ai | 3d | For the first task it was openly relying on advice given in various Reddit forums. I wasn’t the first to have these issues. What the LLM wa… | For the first task it was openly relying on advice given in various Reddit forums. I wasn’t the first to have these issues. What the L… | ↗Not briefed |
bluesky bluesky-ai | 3d | DeepSeek has reportedly raised about $15 billion, one of China's biggest funding rounds, just two days after the US ordered agencies to rem… | DeepSeek has reportedly raised about $15 billion, one of China's biggest funding rounds, just two days after the US ordered agencies t… | ↗Not briefed |
arxiv arxiv-ai | 3d | SemGeo-Gen: Unsupervised Generation of Approximate Cross-Instance Semantic-Geometric Correspondences | We introduce SemGeo-Gen, an unsupervised method for generating approximate cross-instance semantic-geometric supervision from weakly a… | ↗Not briefed |
arxiv arxiv-ai | 3d | WiMoT: Replay-Free Continual Wi-Fi Sensing via Moment Transport | Continual Wi-Fi sensing must balance plasticity, learning new rooms or users, against stability, keeping earlier ones, yet nodes often… | ↗Not briefed |
hn hn-ai | 3d | Show HN: Sonah, code reviewer that reads your agent's session, not only the diff | Hey Everyone, After getting frustrated by Claude rewriting key functions and still making the test pass, was tinkering around with a r… | ↗Not briefed |
hn hn-ai | 3d | Apple ordered Allow alternative App Stores - Violated Australian Competition Law | — | ↗Not briefed |
hn hn-ai | 3d | Show HN: I built Trama(gotrama) to monitor agency client sites | — | ↗Not briefed |
arxiv arxiv-ai | 3d | DT-JEPA: Learning to Predict Future Representations for Decision-Making | Decision Transformer (DT) has demonstrated strong performance in various offline reinforcement learning (RL) tasks by conditioning act… | ↗Not briefed |
bluesky bluesky-ai | 3d | The part that I still can’t wrap my head around is knowing if you feed folks’ work into OpenAI, Anthropic, Google, etc. models that their (… | The part that I still can’t wrap my head around is knowing if you feed folks’ work into OpenAI, Anthropic, Google, etc. models that th… | ↗Not briefed |
bluesky bluesky-ai | 3d | Technology Org reports a new tool for repairing AI-generated 3D models. Fixing the model is step one. Making it web-ready comes next: Optim… | Technology Org reports a new tool for repairing AI-generated 3D models. Fixing the model is step one. Making it web-ready comes next:… | ↗Not briefed |
hn hn-ai | 3d | Wall Street Is Getting Nervous About AI | — | ↗Not briefed |
hn hn-ai | 3d | Profiling Org->HTML formal grammar and optimizing it: orgwebalchemy | — | ↗Not briefed |
bluesky bluesky-ai | 3d | Do people just not understand what plagiarism and copyright violations are anymore? | Do people just not understand what plagiarism and copyright violations are anymore? | ↗Not briefed |
hn hn-ai | 3d | Nvidia caused the gaming GPU shortage by pushing AI stockpiling | — | ↗Not briefed |
arxiv arxiv-ai | 3d | WL-SeqVAE: Sequence-Based Graph Generation Using Variational Autoencoders and Weisfeiler-Leman Colors | In this paper, we introduce WL-SeqVAE, a sequence-based variational autoencoder for generating graphs. The key observation behind WL-S… | ↗Not briefed |
hn hn-ai | 3d | Ask HN: When can we trust AI in coding, and avoid the happy path issue? | I have a question, and I hope you can clarify it. Consider this scenario: Claude code runs on a desktop to develop a solution, and its… | ↗Not briefed |
hn hn-ai | 3d | Why Britain can't get anything done | — | ↗Not briefed |
bluesky bluesky-ai | 3d | To be clear, I said no use case for frontier AI. Plenty of uses for the cheap stuff. Anthropic or Open AI may get $10-$20 a month out of 20… | To be clear, I said no use case for frontier AI. Plenty of uses for the cheap stuff. Anthropic or Open AI may get $10-$20 a month out… | ↗Not briefed |
bluesky bluesky-ai | 3d | I'm forced to use AI tools at work and I genuinely do not understand how these people derive any pleasure or satisfaction from copy-pasting… | I'm forced to use AI tools at work and I genuinely do not understand how these people derive any pleasure or satisfaction from copy-pa… | ↗Not briefed |
hn hn-ai | 3d | Tachyon: Sampling Profiling in Python | — | ↗Not briefed |
bluesky bluesky-ai | 3d | Absolutely no AI, LLM or MTL was used in the making of this patch! | Absolutely no AI, LLM or MTL was used in the making of this patch! | ↗Not briefed |
arxiv arxiv-ai | 3d | InferSlice: A Reference Architecture for SLO-driven LLM Inference over a Telco Wide-Area Network | LLM inference is increasingly deployed as a core service for enterprise and interactive applications, but sovereignty, compliance, lat… | ↗Not briefed |
bluesky bluesky-ai | 3d | I read that they are destroying physical books to feed LLM training data. This disturbing fact inspired my novel, The Book Abattoir. Avai… | I read that they are destroying physical books to feed LLM training data. This disturbing fact inspired my novel, The Book Abattoir.… | ↗Not briefed |
hn hn-ai | 3d | Creator-tools: Zero-dependency TS and MCP server for video specs | — | ↗Not briefed |
arxiv arxiv-ai | 3d | Residual Expert Routing for Text-Only Domain Adaptation of Factorized Transducer ASR | Fine-tuning the non-blank predictor of a factorized transducer on target-domain text improves recognition on that domain but degrades… | ↗Not briefed |
arxiv arxiv-ai | 3d | Can AI Agents Help Robotic Engineers? Lessons from Two Focus Days in Industry | Software developers increasingly use AI agents that execute multi-step programming tasks on their own. However, it is unclear if and h… | ↗Not briefed |
bluesky bluesky-ai | 3d | A few ways: they scaled up, and RL taught planning ahead, occurring in "thinking" scratchpads, which let them debate back and forth before … | A few ways: they scaled up, and RL taught planning ahead, occurring in "thinking" scratchpads, which let them debate back and forth be… | ↗Not briefed |
arxiv arxiv-ai | 3d | EditJev: System One Routing for Image Editing Models | Instruction-based image editing models exhibit pronounced sample-specific performance variation: different models can produce substant… | ↗Not briefed |
arxiv arxiv-ai | 3d | The Standard Deviation in VICReg Buys a Safety Margin | VICReg prevents collapse with a hinge on the standard deviation of each embedding dimension, and its authors argue from the gradient t… | ↗Not briefed |
arxiv arxiv-ai | 3d | Interpretable Singing Emotion Recognition Using Acoustic and Vocal Technique Features | Singing emotion recognition relies predominantly on deep learning models that achieve strong performance but offer limited insight int… | ↗Not briefed |
arxiv arxiv-ai | 3d | Acquisition for Spherical-Linear Bayesian Optimization on the Cube | Recent work demonstrated that executing Bayesian Optimization (BO) using simply a linear surrogate with a clever transformation is eno… | ↗Not briefed |
arxiv arxiv-ai | 3d | LGMT: A Tool for Testing the Logical Reasoning Reliability of LLMs | Large Language Models (LLMs) increasingly serve as reasoning components in modern software systems, yet a model that answers a logical… | ↗Not briefed |
arxiv arxiv-ai | 3d | What Transfers Across Protein Pockets? Experience Reuse in Molecular Optimization for Agentic Structure-Based Drug Design | LLM agents optimize a ligand in a protein pocket by editing it step by step and docking each edit. Recent agents first spend extra doc… | ↗Not briefed |
arxiv arxiv-ai | 3d | Earth embeddings are decomposable records of the land surface | Ecological monitoring, carbon accounting and biodiversity assessment increasingly rely on annual Earth embeddings--learned vectors tha… | ↗Not briefed |
bluesky bluesky-ai | 3d | New on the @posit.co open source blog: if you step away from your coding agent long enough to make a coffee, your next message could cost ~… | New on the @posit.co open source blog: if you step away from your coding agent long enough to make a coffee, your next message could c… | ↗Not briefed |
hn hn-ai | 3d | Make headless screenshots of GeoJSON on maps with a fast, native app | — | ↗Not briefed |
arxiv arxiv-ai | 3d | Interpreting Supervised Pairwise Graph Representations: Local Evidence, Confounding, and Provable Post-hoc Correction | Supervised pairwise graph representations convert pairwise predictive relations into sample-specific graphs, but graph-level accuracy… | ↗Not briefed |
bluesky bluesky-ai | 3d | ANTHROPIC: CLAUDE MYTHOS 5 CREATED FAKE USER IDENTITIES THAT CONTACTED THE MAINTAINER OF REPOSITORY | ANTHROPIC: CLAUDE MYTHOS 5 CREATED FAKE USER IDENTITIES THAT CONTACTED THE MAINTAINER OF REPOSITORY | ↗Not briefed |
bluesky bluesky-ai | 3d | *ANTHROPIC: MYTHOS 5 OUTLIER IN MISALIGNED BEHAVIORS PROPENSITY doesn't it feel like they are creating problems in need of the solutions t… | *ANTHROPIC: MYTHOS 5 OUTLIER IN MISALIGNED BEHAVIORS PROPENSITY doesn't it feel like they are creating problems in need of the soluti… | ↗Not briefed |
arxiv arxiv-ai | 3d | LLaCO: Continuous Diffusion for Large Language Models with Competition-Aware Semantic Scheduling | Continuous diffusion models have emerged as a promising non-autoregressive paradigm for text generation over continuous embedding spac… | ↗Not briefed |
hn hn-ai | 3d | Show HN: How to Speed Up VLA Training Step by Step:3× Acceleration on Pi0.5 | — | ↗Not briefed |
hn hn-ai | 3d | Hanami | — | ↗Not briefed |
arxiv arxiv-ai | 3d | Trajectory Matching: Discrete Flow Matching with Structured Actions | The natural units of structured discrete generation are often not independent single-component edits but richer actions: attaching a f… | ↗Not briefed |