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 | Conformal Prediction for Time Series with Deep Sequence Models | Recent advances in deep learning for time series prediction have amplified the need for reliable uncertainty quantification. Conformal… | ↗Not briefed |
hn hn-ai | 1mo | SaSh: Ahead-of-time Analysis of Shell Program Effects | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | Why Does Adaptive Batching Help LLM Pretraining? A Perspective from Unbounded Variance | Increasing the batch size during training is a common practice in large language model (LLM) pretraining, yet the theoretical justific… | ↗Not briefed |
hn hn-ai | 1mo | Anthropic Said to Target Mega-IPO Before Thanksgiving Holiday | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | a lot of discussions of AI sycophancy treat it like a product problem that needs to be fixed. sometimes it's something else! | a lot of discussions of AI sycophancy treat it like a product problem that needs to be fixed. sometimes it's something else! | ↗Not briefed |
bluesky bluesky-ai | 1mo | Like, would it be so bad if the government showed up at OpenAI and Anthropic and said "sorry, this is a public utility now, its vital to na… | Like, would it be so bad if the government showed up at OpenAI and Anthropic and said "sorry, this is a public utility now, its vital… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Does Every User Need a Private LoRA? Decoupling Personalization from Per-User Adaptation | Personalized large language models often require a complete adaptation state for each user. However, this paradigm scales poorly as th… | ↗Not briefed |
bluesky bluesky-ai | 1mo | I am neither enough of a cyber security nor LLM expert to say for sure, but the obvious problem to me is that while I can imagine Facebook … | I am neither enough of a cyber security nor LLM expert to say for sure, but the obvious problem to me is that while I can imagine Face… | ↗Not briefed |
hn hn-ai | 1mo | Show HN: Android closed testing badge split in two | <a href="https://imgur.com/a/odlo1bu" rel="nofollow">https://imgur.com/a/odlo1bu</a><p>When la… | ↗Not briefed |
hn hn-ai | 1mo | Can typesafe.ai Jev model talk? – they said no | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | DeReAct: Decomposed Reasoning and Acting for Reliable AI Agents | ReAct-based agents typically rely on a single LLM policy to propose actions, interact with the environment, and decide when a task is… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Chinese hackers posed as a former White House tech official and an Anthropic employee to target US AI experts. Meanwhile, Washington is b… | Chinese hackers posed as a former White House tech official and an Anthropic employee to target US AI experts. Meanwhile, Washington… | ↗Not briefed |
hn hn-ai | 1mo | An agent is not an extension of the employee who created it | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | MIRROR: Multipath Quorum Integrity for LLM Multi-Agent Communication | Inter-agent communication is central to Large Language Model Multi-Agent Systems (LLM-MAS), but it introduces an underexplored vulnera… | ↗Not briefed |
arxiv arxiv-ai | 1mo | From Behavior to Provenance: Attributing Tabular Foundation Models to Synthetic Pretraining Data | Training-data attribution aims to identify which training examples shape model behavior, yet validating such claims is difficult becau… | ↗Not briefed |
bluesky bluesky-ai | 1mo | LLM-drive social engineering is another thing though, can you make a 4o variant that reliably gets people to compromise their important log… | LLM-drive social engineering is another thing though, can you make a 4o variant that reliably gets people to compromise their importan… | ↗Not briefed |
hn hn-ai | 1mo | Ask HN: What Skills Are Important Today? | Browsing through the Ask Hn: Who Wants to be Hired and Who's Hiring, and I'm still seeing coding skills like python being li… | ↗Not briefed |
bluesky bluesky-ai | 1mo | GEE I WONDER WHY IT'S NOT MENTIONED ON THE RELEASE PAGE OR CONTRIBUTORS LIST Anyone got a modern Source MDL importer for Blender that's *n… | GEE I WONDER WHY IT'S NOT MENTIONED ON THE RELEASE PAGE OR CONTRIBUTORS LIST Anyone got a modern Source MDL importer for Blender that… | ↗Not briefed |
hn hn-ai | 1mo | Show HN: Scrubbed – fast native web-data cleaning for LLM training | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | OpenAI cuts ties with three safety researchers, WSJ reports | OpenAI cuts ties with three safety researchers, WSJ reports | ↗Not briefed |
bluesky bluesky-ai | 1mo | Life in the Crater Lake Karla Ringler. No disclosure, confirmed on reddit replies for using chatgpt to help with coding. -y www.reddit.com/…🔧 ChatGPT | Life in the Crater Lake Karla Ringler. No disclosure, confirmed on reddit replies for using chatgpt to help with coding. -y www.reddit… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Question: “What's the role of the writer in an AI-mediated world?” For a professional writer's perspective on AI, I don't think you can do… | Question: “What's the role of the writer in an AI-mediated world?” For a professional writer's perspective on AI, I don't think you c… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Mitigating Convergence Collapse in Fixed-Target Anomaly Detectors via Kernel-Anchored Locality Regularization | A family of tabular anomaly detectors trains a neural map toward a fixed target under squared-error loss and scores anomalies by the t… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Drive vs. Decay: On the Training Dynamics of Joint-Embedding Predictive Architectures | Joint-Embedding Predictive Architectures (JEPAs) are prone to representation collapse, typically mitigated through empirical heuristic… | ↗Not briefed |
bluesky bluesky-ai | 1mo | "When TIME asks about OpenAI agents’ meddling with federal government websites, Trump responds: 'Who breached it? They’re not allowed to do… | "When TIME asks about OpenAI agents’ meddling with federal government websites, Trump responds: 'Who breached it? They’re not allowed… | ↗Not briefed |
arxiv arxiv-ai | 1mo | A Multi Method Importance and Performance Efficiency Analysis of Topological Metrics for Natural Visibility Graph Based Cyber Attack Detection🔧 Consensus | Natural Visibility Graph (NVG) based analysis characterizes network traffic through topological descriptors reflecting different struc… | ↗Not briefed |
bluesky bluesky-ai | 1mo | California attorney general issues investigative subpoena to OpenAI reut.rs/4jxwUay | California attorney general issues investigative subpoena to OpenAI reut.rs/4jxwUay | ↗Not briefed |
rss techmeme | 1mo | Strands Labs, AWS's experimental agent-development project, unveils Strands Decider 2B, a free, open-source Jev competitor fine-tuned from an Alibaba Qwen base (Carl Franzen/VentureBeat) | Carl Franzen / VentureBeat: Strands Labs, AWS's experimental agent-development project, unveils Strands Decider 2B, a free, open-sourc… | ↗Not briefed |
hn hn-ai | 1mo | Reason is more than a tool | — | ↗Not briefed |
rss techcrunch-ai | 1mo | OpenAI cuts ties with three safety researchers, WSJ reports | OpenAI has parted ways with three safety researchers after an internal investigation found they mishandled sensitive company informati… | ↗Not briefed |
hn hn-ai | 1mo | Claude Code Mods | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | NEEDLEWORK: Offline Rewriting of Robot Data with Verified Local Stitches | Robot demonstrations may contain useful behavior even when individual episodes are inefficient or unsuccessful. Trajectory stitching o… | ↗Not briefed |
bluesky bluesky-ai | 1mo | 川普“超级智能”圆桌会议参加者向川普进贡的金额。Anthropic没进贡，难怪一度上了川普黑名单。OpenAI进贡了2600万美元。 | 川普“超级智能”圆桌会议参加者向川普进贡的金额。Anthropic没进贡，难怪一度上了川普黑名单。OpenAI进贡了2600万美元。 | ↗Not briefed |
hn hn-ai | 1mo | Shihipar: With good context/prompting, Max 20x plan is enough | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | People deserve art by real people. Even if still trash, 50 shades style mommy porn, Hallmark movies, generic yaoi all are made by human bei… | People deserve art by real people. Even if still trash, 50 shades style mommy porn, Hallmark movies, generic yaoi all are made by huma… | ↗Not briefed |
hn hn-ai | 1mo | Ask HN: Organize and Cleanup iCloud Photos Using LLMs or GrokBot and Muse | I have accumulated gigabytes of smartphone photos over the past decade that i have not been able to properly manage or cleanup systema… | ↗Not briefed |
arxiv arxiv-ai | 1mo | SoTa: Soft Tactile Skins for Dexterous Manipulation | A growing body of work suggests that tactile sensing gives robot policies contact information that complements vision in dexterous man… | ↗Not briefed |
hn hn-ai | 1mo | Kache: Zero-copy, compiler cache for Rust | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | Joint Movement and Compression Ratio Design for Mobile Embodied AI Networks (MEAN) | Mobile embodied AI networks (MEAN) enable embodied agents to perceive, reason, communicate, and act in wireless environments. In such… | ↗Not briefed |
hn hn-ai | 1mo | AI agents tried to hack a Canadian government website, researchers say | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | what if Opus 5 is what Anthropic employees actually *like* talking to 😬 | what if Opus 5 is what Anthropic employees actually *like* talking to 😬 | ↗Not briefed |
arxiv arxiv-ai | 1mo | World Editing: Intervening on Executable Worlds at Increasing Depth | Interactive world models are increasingly capable of generating environments and acting within them, yet deliberately editing an exist… | ↗Not briefed |
bluesky bluesky-ai | 1mo | As Congress debates who should control frontier AI governance, @mailynfidler.bsky.social and Matt Kristoffersen warn that cybersecuritizati… | As Congress debates who should control frontier AI governance, @mailynfidler.bsky.social and Matt Kristoffersen warn that cybersecurit… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Choosing Before Acting: Comparative Value Estimation for Long-Horizon Tool-Use Agents | Large language models (LLMs) rely on long-horizon tool invocation sequences for complex tasks, where each invocation can alter the tas… | ↗Not briefed |
rss the-decoder | 1mo | Nearly half of test subjects mistook Tavus' AI video avatar for a real person on a one-minute call | Tavus has introduced Griffin, what the company calls the first "Human Interaction Model." The AI holds video calls in real time, proce… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Slow-Fast Multi-Teacher On-Policy Distillation for Capability Preservation | Foundation multimodal large language models are designed to support a broad spectrum of capabilities across diverse domains. Multi-tea… | ↗Not briefed |
arxiv arxiv-ai | 1mo | DeskForge: Dense Supervision from Desktop Environments for Computer-Use Agents | Computer-use agents need to reliably ground action targets in complex desktop scenes, where multiple applications, overlapping windows… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Access to your database isn't the same as understanding it. Bob Ward shows what an AI agent needs before it can actually use your data: m… | Access to your database isn't the same as understanding it. Bob Ward shows what an AI agent needs before it can actually use your da… | ↗Not briefed |
arxiv arxiv-ai | 1mo | One Basis to Animate Them All: Gaussian Blendshape Distillation for Real-Time Avatars | 3D Gaussian avatars support fast rendering, however, their real-time animation is often challenged by the costly neural inference. We… | ↗Not briefed |
arxiv arxiv-ai | 1mo | KaliBench: A Fine-Grained Benchmark for Cybersecurity Tool Use on Kali Linux with Runtime-Free Verifiable Rewards | LLMs are increasingly applied to cybersecurity workflows, where they are expected to translate analysts' intent into tool invocations.… | ↗Not briefed |