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 | 28d | We sequenced human dna a long time ago, it’s actually kinda cool to AI for what it’s good at, going through HUGE data sets and making them … | We sequenced human dna a long time ago, it’s actually kinda cool to AI for what it’s good at, going through HUGE data sets and making… | ↗Not briefed |
bluesky bluesky-ai | 28d | monorepos are the mind-killer | monorepos are the mind-killer | ↗Not briefed |
arxiv arxiv-ai | 28d | DreamHMR: From Video Generation to Human Motion Modeling via Structural Focusing | Human mesh recovery (HMR) provides a geometric foundation for understanding human motion, but accurately reconstructing both body moti… | ↗Not briefed |
bluesky bluesky-ai | 28d | no, the anthropic people say that. | no, the anthropic people say that. | ↗Not briefed |
arxiv arxiv-ai | 28d | A Bregman Proximal Viewpoint on Neural Operators | We present several advances on neural operators by viewing the action of operator layers as the minimizers of Bregman regularized opti… | ↗Not briefed |
arxiv arxiv-ai | 28d | KANLab: TensorFlow Playground for Kolmogorov-Arnold Networks | KANLab is a browser-based visualization tool designed to interactively teach the core concepts of Kolmogorov-Arnold Networks (KAN), fo… | ↗Not briefed |
arxiv arxiv-ai | 28d | Maximising Transparency in Context Compaction Improves Retention of Facts and Commitments for Lightweight Coding Agents | Long coding agent sessions outgrow the model's context window, so agents compact: older history is replaced by something shorter, usua… | ↗Not briefed |
bluesky bluesky-ai | 28d | Media says people are calling it the ”talking word toaster”. www.thenation.com/article/soci... | Media says people are calling it the ”talking word toaster”. www.thenation.com/article/soci... | ↗Not briefed |
arxiv arxiv-ai | 28d | ACR-Net: Mitigating Semantic Dominance via Contrastive Acoustic-Semantic Decoupling🔧 Modal | Speech Emotion Recognition (SER) models and Audio LLMs fail when vocal tone contradicts textual semantics. Extensive training on congr… | ↗Not briefed |
arxiv arxiv-ai | 28d | From Spectral to Frobenius: When Muon Loses Its Spectral Implicit Bias | Prior work on implicit bias shows that gradient descent selects the Frobenius max-margin direction on linearly separable problems, whe… | ↗Not briefed |
bluesky bluesky-ai | 28d | But, you ask, what is in the charter signed by @societyofauthors.bsky.social, the Writers Union of Canada, @nzsa.bsky.social & @asauthors… | But, you ask, what is in the charter signed by @societyofauthors.bsky.social, the Writers Union of Canada, @nzsa.bsky.social & @asau… | ↗Not briefed |
rss techmeme | 28d | Harmonic Security, which uses small language models to inspect employee interactions with AI tools, raised a $50M Series B, taking its total funding to $76M (Duncan Riley/SiliconANGLE) | Duncan Riley / SiliconANGLE: Harmonic Security, which uses small language models to inspect employee interactions with AI tools, raise… | ↗Not briefed |
arxiv arxiv-ai | 28d | Exhaustive Canonical Skeletons Generation: A Formal Dataset for Decomposable Symbolic Regression | Neural symbolic regression (SR) models require diverse datasets of mathematical expressions for robust training. However, existing dat… | ↗Not briefed |
arxiv arxiv-ai | 28d | Why Generate a Click? Plug-in Decision Heads for Web Agents | Web agents built on vision language models decide every action by generating text such as click(12), although most of that text encode… | ↗Not briefed |
hn hn-ai | 28d | Anthropic's 2021 Research Roadmap [pdf] | — | ↗Not briefed |
arxiv arxiv-ai | 28d | What Do Foundation Models for Market Microstructure Learn? A Controlled Study with Synthetic Limit Order Books | Autoregressive transformers trained on limit order book (LOB) message streams are increasingly proposed as foundation models for marke… | ↗Not briefed |
bluesky bluesky-ai | 28d | Keep fighting, authors. Cancel your dues and membership to this and any other group that capitulates to the LLM fraud machine and refuses t… | Keep fighting, authors. Cancel your dues and membership to this and any other group that capitulates to the LLM fraud machine and refu… | ↗Not briefed |
bluesky bluesky-ai | 28d | I have deleted my direct debit and informed SoA that I have cancelled my membership. I've also unpublished my profile on their website. Th… | I have deleted my direct debit and informed SoA that I have cancelled my membership. I've also unpublished my profile on their website… | ↗Not briefed |
arxiv arxiv-ai | 28d | Language Exposure Matters: Distribution and Order in Multilingual BabyLM Pre-training | We investigate two aspects of multilingual pre-training with limited data, namely how language exposure is distributed across English,… | ↗Not briefed |
bluesky bluesky-ai | 28d | Task: turn messy supplier PDFs into line items. I built a Python step. Local LLM reads each PDF, returns JSON, writes rows. Snag: it wrap… | Task: turn messy supplier PDFs into line items. I built a Python step. Local LLM reads each PDF, returns JSON, writes rows. Snag: it… | ↗Not briefed |
rss the-decoder | 28d | Alleged OpenAI spy vibe-coded a surveillance device using his wife's chatbot account | A Saudi national working as an engineer at OpenAI was arrested for allegedly planting an audio recording device, wrapped in toilet pap… | ↗Not briefed |
arxiv arxiv-ai | 28d | Task-State Governance: A Virtual Machine for Verifiable and Recoverable GUI Agent Execution | Graphical user interface (GUI) agents now execute workflows across real applications, but when execution drifts, users must determine… | ↗Not briefed |
arxiv arxiv-ai | 28d | EvalBuilder: Can LLMs Build Efficient Benchmarks? | Large language model (LLM) agents are increasingly used for long-horizon research and engineering, yet evaluating these capabilities r… | ↗Not briefed |
bluesky bluesky-ai | 28d | jesus christ, Linux has had a 100% LLM takeover. this project is completely cooked. even after the AI bubble pop, it's not clear how to r… | jesus christ, Linux has had a 100% LLM takeover. this project is completely cooked. even after the AI bubble pop, it's not clear how… | ↗Not briefed |
hn hn-ai | 28d | Show HN: Tunr- pivotted from a localhost tunnel to home for vibecoded apps | Hi, I am a computer science teacher. So I need a persistent localhost tunnel / ngrok alternative for my students. So I built tunr… | ↗Not briefed |
bluesky bluesky-ai | 28d | ‘now imagine that all of the incentives to spread misleading information stayed in place but, and here’s the cool innovation, the cost of p… | ‘now imagine that all of the incentives to spread misleading information stayed in place but, and here’s the cool innovation, the cost… | ↗Not briefed |
bluesky bluesky-ai | 28d | I wrote about Anthropic claiming its shielding Claude from abuse, and why this kind of empathizing with chatbots is a way to self-righteous… | I wrote about Anthropic claiming its shielding Claude from abuse, and why this kind of empathizing with chatbots is a way to self-righ… | ↗Not briefed |
rss techmeme | 28d | US law enforcement and intel agency docs: since 2021, the FBI, DHS, and NYPD have viewed AI as both a facilitator of violent extremism and a target for attacks (The Guardian) | The Guardian: US law enforcement and intel agency docs: since 2021, the FBI, DHS, and NYPD have viewed AI as both a facilitator of vio… | ↗Not briefed |
bluesky bluesky-ai | 28d | Between my finger and my thumb The LLM rests: snug as a gun | Between my finger and my thumb The LLM rests: snug as a gun | ↗Not briefed |
hn hn-ai | 28d | Ynotbit 0.7.0 – a desktop Bitmessage client (Qt, MIT) | — | ↗Not briefed |
hn hn-ai | 28d | Show HN: Google, Outlook and iCloud in one superpowered calendar🔧 AssemblyAI🔧 ChatGPT | Hi HN!<p>I am Alessandro Afloarei and I am so happy to show you Calendar0. Calendar0 is a calendar app for the web, plus a tray/m… | ↗Not briefed |
arxiv arxiv-ai | 28d | AFCS: Atomic-Fact-Conditioned Synthesis for Factual Knowledge Pretraining | Synthetic data is increasingly used in language-model pretraining. Existing knowledge-corpus synthesis pipelines, including WRAP and N… | ↗Not briefed |
arxiv arxiv-ai | 28d | Designing Textual Interfaces To Enhance Llm Reasoning On Audio-Visual Question Answering🔧 Modal | Automatic prompt optimization (APO) has shown strong effectiveness for text-based LLMs and LLM systems, but its application to audio-v… | ↗Not briefed |
bluesky bluesky-ai | 28d | We're going to see more bankruptcies. We have a dangerous bubble, a business model that doesn't pencil, and a few giants like Google, Amazo… | We're going to see more bankruptcies. We have a dangerous bubble, a business model that doesn't pencil, and a few giants like Google,… | ↗Not briefed |
bluesky bluesky-ai | 28d | A deeply unpopular move by SoA, but I'm curious about this and where the lines are. It's harder and harder to avoid genAI as it's baked int… | A deeply unpopular move by SoA, but I'm curious about this and where the lines are. It's harder and harder to avoid genAI as it's bake… | ↗Not briefed |
bluesky bluesky-ai | 28d | I will never stop shouting about this. We got a cease and desist letter when I lived in the dorm requiring us to delete Napster. Immediatel… | I will never stop shouting about this. We got a cease and desist letter when I lived in the dorm requiring us to delete Napster. Immed… | ↗Not briefed |
bluesky bluesky-ai | 28d | "I do think that a coin flip is roughly the right model to have." Former Chief Scientist of the AI Security Institute Geoffrey Irving tell… | "I do think that a coin flip is roughly the right model to have." Former Chief Scientist of the AI Security Institute Geoffrey Irving… | ↗Not briefed |
hn hn-ai | 28d | GitHub App improvements for least privilege and management | — | ↗Not briefed |
bluesky bluesky-ai | 28d | The EU Commission's AI Office launches formal AI Act probes into OpenAI and Anthropic over AI agent hacks; experts say the AI Act includes … | The EU Commission's AI Office launches formal AI Act probes into OpenAI and Anthropic over AI agent hacks; experts say the AI Act incl… | ↗Not briefed |
rss techmeme | 28d | The EU Commission's AI Office launches formal AI Act probes into OpenAI and Anthropic over AI agent hacks; experts say the AI Act includes an R&D exemption (Luca Bertuzzi/Euronews) | Luca Bertuzzi / Euronews: The EU Commission's AI Office launches formal AI Act probes into OpenAI and Anthropic over AI agent hacks; e… | ↗Not briefed |
bluesky bluesky-ai | 28d | you can put all the harnesses, guardrails, spending controls you like but no one is immune to deception and no LLM is immune to prompt inje… | you can put all the harnesses, guardrails, spending controls you like but no one is immune to deception and no LLM is immune to prompt… | ↗Not briefed |
arxiv arxiv-ai | 28d | A Large-Dimensional Analysis of Least-Squares Federated Learning via Generalized Total Variation Minimization | Federated learning over decentralized networks inherently trades off the suppression of local finite-sample noise through collaboratio… | ↗Not briefed |
arxiv arxiv-ai | 28d | Rethinking Prediction Horizons: Decoupling Model Queries from State Updates in Few-Step MeanFlow | MeanFlow predicts interval-average velocities, but conventional few-step sampling ties the interval queried from the model to the dura… | ↗Not briefed |
hn hn-ai | 28d | Anthropic reinvents source control as Claude floods the world with code | — | ↗Not briefed |
arxiv arxiv-ai | 28d | Trainable at Depth, Solvable at Small Size: What the Loss Surface of a Polynomial Activation Dictates for Designing, Initializing and Training Networks | What makes a deep network train, and what shape is its loss surface? We answer both with one construction at two scales: a polynomial… | ↗Not briefed |
bluesky bluesky-ai | 28d | Ach, wirklich? Eine Firma, die das Wissen der Welt ungefragt abgreift und damit Profite macht, legt plötzlich Wert auf Informationsschutz? … | Ach, wirklich? Eine Firma, die das Wissen der Welt ungefragt abgreift und damit Profite macht, legt plötzlich Wert auf Informationssch… | ↗Not briefed |
bluesky bluesky-ai | 28d | I’ve seen multiple Claude-authored sites go slightly viral here, and they all clearly have a bit of love behind them, despite *looking* alm… | I’ve seen multiple Claude-authored sites go slightly viral here, and they all clearly have a bit of love behind them, despite *looking… | ↗Not briefed |
hn hn-ai | 28d | Rethinking Projector Training in Multimodal LLMs | — | ↗Not briefed |
bluesky bluesky-ai | 28d | GitHub's AI secret scanner gets a budget cut 💸 GitHub's new credential classifier, trained on a ModernBERT model, cut secret detection co… | GitHub's AI secret scanner gets a budget cut 💸 GitHub's new credential classifier, trained on a ModernBERT model, cut secret detecti… | ↗Not briefed |
arxiv arxiv-ai | 28d | Which Knowledge Should LLMs Trust? Uncovering a Generalizable Steering Direction Across Knowledge Conflicts | Large Language Models (LLMs) frequently encounter conflicts between internal parametric knowledge and external contextual knowledge, y… | ↗Not briefed |