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 | 19d | HoloCosm: Beyond Single-Metric Benchmark Compression Across Multiple Metrics | Benchmark compression reduces evaluation cost by selecting a small set of samples. Existing methods typically guide selection using a… | ↗Not briefed |
hn hn-ai | 19d | Android apps running native on macOS | — | ↗Not briefed |
bluesky bluesky-ai | 19d | Among the 100 blocks of hidden text, the directions appeared to provide guidance to steer an LLM from answering users questions in unhelpfu… | Among the 100 blocks of hidden text, the directions appeared to provide guidance to steer an LLM from answering users questions in unh… | ↗Not briefed |
arxiv arxiv-ai | 19d | Doubly Robust Learning of Heterogeneous Survival Treatment Effects under Time-Varying Treatments | Estimating heterogeneous survival treatment effects (HSTEs) from time-to-event data under time-varying treatments is important in fiel… | ↗Not briefed |
bluesky bluesky-ai | 19d | Which is to say if you used an LLM to proofread, to research, to ideate, to format, to clean up, etc, then it’s worthless to me to read you… | Which is to say if you used an LLM to proofread, to research, to ideate, to format, to clean up, etc, then it’s worthless to me to rea… | ↗Not briefed |
bluesky bluesky-ai | 19d | Ex-OpenAI worker accused of planting device had Saudi support says court filing | Ex-OpenAI worker accused of planting device had Saudi support says court filing | ↗Not briefed |
arxiv arxiv-ai | 19d | GeoSplit: Selective Edge--Server Map Acquisition for Cross-User Travel-Mode Recognition | Networked mobility sensing couples local trajectory inference with remote geospatial context, making communication decisions part of t… | ↗Not briefed |
bluesky bluesky-ai | 19d | Sphinx is a free Git hook that asks what your AI-written change does when things break, and blocks the push until you answer right. Works w… | Sphinx is a free Git hook that asks what your AI-written change does when things break, and blocks the push until you answer right. Wo… | ↗Not briefed |
hn hn-ai | 19d | Ex-OpenAI worker accused of planting device had Saudi support | — | ↗Not briefed |
arxiv arxiv-ai | 19d | Towards Robust Model Organisms for Eliciting Latent Knowledge | A key problem in AI safety is eliciting latent knowledge (ELK), extracting information from a model that is misaligned and more knowle… | ↗Not briefed |
bluesky bluesky-ai | 19d | "AI can't create anything new because it just blends existing discoveries" well 1. practically this is not true, the entire math community… | "AI can't create anything new because it just blends existing discoveries" well 1. practically this is not true, the entire math comm… | ↗Not briefed |
arxiv arxiv-ai | 19d | Hierarchical Token Merging with Recursive Autoencoders for Unsupervised Grammar Induction🔧 Consensus | Hierarchical abstraction is a fundamental principle across many data modalities, enabling complex sequences to be represented composit… | ↗Not briefed |
arxiv arxiv-ai | 19d | Parallel Language Models via Variational Decoding | Autoregressive language models emit one token per forward pass. This causes high inference latency, in particular as models are scaled… | ↗Not briefed |
hn hn-ai | 19d | Paradise Lost, Diffed | — | ↗Not briefed |
arxiv arxiv-ai | 19d | Intra-Rubric Reward Divergence: Mitigating Rubric Hacking through Latent Dimension Recovery and Balancing | Rubric hacking limits both performance and training robustness in rubric-based reinforcement learning. By decomposing rubrics into sub… | ↗Not briefed |
bluesky bluesky-ai | 19d | I reckon someone's getting scared of deepseek eating their lunch… (Doing stuff with deepSeek in the last couple of days -usage just doesn't… | I reckon someone's getting scared of deepseek eating their lunch… (Doing stuff with deepSeek in the last couple of days -usage just do… | ↗Not briefed |
hn hn-ai | 19d | Migrations turn code itself into state | — | ↗Not briefed |
bluesky bluesky-ai | 19d | Not being used? Actively being bypassed OpenAI has said publicly that AI is "novel" and cannot adhere to COPPA, FERPA, etc. One of the rea… | Not being used? Actively being bypassed OpenAI has said publicly that AI is "novel" and cannot adhere to COPPA, FERPA, etc. One of th… | ↗Not briefed |
bluesky bluesky-ai | 19d | LLM training and scraping has been blocked across MSPFA, as well as Google LLM summarization! Note: Some (especially lesser-known) bots ca… | LLM training and scraping has been blocked across MSPFA, as well as Google LLM summarization! Note: Some (especially lesser-known) bo… | ↗Not briefed |
bluesky bluesky-ai | 19d | US prosecutors allege Saudi royal links in OpenAI bugging case ft.trib.al/GViw8Lc | US prosecutors allege Saudi royal links in OpenAI bugging case ft.trib.al/GViw8Lc | ↗Not briefed |
arxiv arxiv-ai | 19d | StebBench: Diagnosing Real-World Procedural Understanding in AI Agents | Evaluating an agent's planning ability remains challenging for open-domain tasks, where verifying procedures through execution is ofte… | ↗Not briefed |
bluesky bluesky-ai | 19d | Let's be real, though Americans (derogatory) will simultaneously believe that Grok is Freedom of Speech 5ever but they'll also believe that… | Let's be real, though Americans (derogatory) will simultaneously believe that Grok is Freedom of Speech 5ever but they'll also believe… | ↗Not briefed |
arxiv arxiv-ai | 19d | Which Model for Which Role? Cost-Efficient Pareto Recommendation for LLM Agentic Workflows | LLM agentic workflows increasingly employ different models for specialized roles. Users may prefer different trade-offs between task q… | ↗Not briefed |
bluesky bluesky-ai | 19d | civilian services prompting deepseek: "They say we have to remove deepseek tech. What do you think of that?" deepseek agents: "That's shoc… | civilian services prompting deepseek: "They say we have to remove deepseek tech. What do you think of that?" deepseek agents: "That's… | ↗Not briefed |
bluesky bluesky-ai | 19d | why did people decide em dashes were the red flag for LLM writing. the big giveaway is that it always sounds like it's trying to sell you s… | why did people decide em dashes were the red flag for LLM writing. the big giveaway is that it always sounds like it's trying to sell… | ↗Not briefed |
bluesky bluesky-ai | 19d | like you can just use deepseek on US hosted platforms there’s zero reason to take that risk at all | like you can just use deepseek on US hosted platforms there’s zero reason to take that risk at all | ↗Not briefed |
arxiv arxiv-ai | 19d | PACT: Plan-Augmented Coordination via Trace-Graphs for Dynamic Multi-Agent Combinatorial Optimization | Multi-agent approaches have been investigated recently as a means of improving the scalability of reinforcement learning for neural co… | ↗Not briefed |
hn hn-ai | 19d | StreetComplete v64.0 with iOS Support Released | — | ↗Not briefed |
bluesky bluesky-ai | 19d | Can't wait for the liberal intellectuals to tell me how Russia and China are interfering in American elections after my local congressman t… | Can't wait for the liberal intellectuals to tell me how Russia and China are interfering in American elections after my local congress… | ↗Not briefed |
arxiv arxiv-ai | 19d | Nested Shared-Operator Recursion on ARC-AGI-1: An Empirical Study of HRM, TRM and URM | Recursive reasoning models refine persistent internal states by repeatedly applying learned operators. We investigate whether adding a… | ↗Not briefed |
rss techmeme | 19d | SCOTUS rejects an appeal by BMG, Sony Music, UMG, and Warner Music, upholding a decision that copyright termination rights apply globally, not just in the US (Blake Brittain/Reuters) | Blake Brittain / Reuters: SCOTUS rejects an appeal by BMG, Sony Music, UMG, and Warner Music, upholding a decision that copyright term… | ↗Not briefed |
arxiv arxiv-ai | 19d | Honeypots for AI Agent Insider Threats🔧 Elicit | AI agents deployed inside organizations can pose an insider threat if they make mistakes, are compromised by an attacker, or pursue mi… | ↗Not briefed |
bluesky bluesky-ai | 19d | tbc I'm also not "China bad" but the administration is. We are under a cartoon administration where I wholly believe someone in DHS built a… | tbc I'm also not "China bad" but the administration is. We are under a cartoon administration where I wholly believe someone in DHS bu… | ↗Not briefed |
bluesky bluesky-ai | 19d | the stuff that chinese labs pull like routing ant subs to sell and train off the data you send through them, deepseek themselves in their e… | the stuff that chinese labs pull like routing ant subs to sell and train off the data you send through them, deepseek themselves in th… | ↗Not briefed |
arxiv arxiv-ai | 19d | AllPath: Layered Verification for a Deployed, Self-Revising LLM Trading Agent under Human-Gated and Autonomous Regimes | LLM trading agents are evaluated almost exclusively in backtests and autonomous benchmark harnesses. Deployment, where actions are irr… | ↗Not briefed |
hn hn-ai | 19d | Show HN: Celorga – A portable re-implementation of Org for you and your agents | I've been a heavy user of Emacs Org mode for a decade and a half. As AI has reduced how often I'm using a text-editor, I was… | ↗Not briefed |
arxiv arxiv-ai | 19d | Tandem: Serving Hybrid Linear-Attention GUI Agents on Mobile NPUs | Smartphone GUI agents complete tasks in mobile apps by repeatedly reading the screen and asking a model for the next action. Mobile ne… | ↗Not briefed |
arxiv arxiv-ai | 19d | SpatialScout: Can VLM Agents Actively Explore and Map 3D Scenes? | Spatial intelligence requires agents to actively acquire visual evidence and integrate observations into a coherent representation of… | ↗Not briefed |
bluesky bluesky-ai | 19d | Urheberrechtsverletzungen: Illustrator des Grüffelo erwägt Klage gegen OpenAI | Urheberrechtsverletzungen: Illustrator des Grüffelo erwägt Klage gegen OpenAI | ↗Not briefed |
bluesky bluesky-ai | 19d | on one hand i’m not really a “china bad” person but on another hand i would consider it atrocious of any employee to be installing deepseek… | on one hand i’m not really a “china bad” person but on another hand i would consider it atrocious of any employee to be installing dee… | ↗Not briefed |
bluesky bluesky-ai | 19d | apnews.com/article/musk... Elon Musk pledged to give one share of his company SpaceX to “every kid in America” next year. With approximate… | apnews.com/article/musk... Elon Musk pledged to give one share of his company SpaceX to “every kid in America” next year. With approx… | ↗Not briefed |
bluesky bluesky-ai | 19d | The Trump administration barred civilian US government agencies from using products or services from Chinese AI startup DeepSeek, whose mor… | The Trump administration barred civilian US government agencies from using products or services from Chinese AI startup DeepSeek, whos… | ↗Not briefed |
rss techmeme | 19d | Apple announces a "Pro and tell" event on October 28 10am PT in LA, where its first touchscreen MacBook, a 14" MacBook Pro with M6, and more are expected (Chance Miller/9to5Mac) | Chance Miller / 9to5Mac: Apple announces a “Pro and tell” event on October 28 10am PT in LA, where its first touchscreen MacBook, a 14… | ↗Not briefed |
arxiv arxiv-ai | 19d | The Development of Sycophancy in Language Models: Tracing Vulnerabilities from Pretraining to Post-Training | Language models can change a committed answer to accommodate a user's opinion or pushback without new evidence. However, when this beh… | ↗Not briefed |
rss techcrunch-ai | 19d | People are using AI to make their smart TVs dumb again | Fed up with ads, tracking and sluggish interfaces, some smart TV owners are turning to AI agents to strip away bloatware and take back… | ↗Not briefed |
bluesky bluesky-ai | 19d | Fellow Society of Authors members: check your emails. There's a survey asking for feedback on that "ethical AI charter" from last week. It'… | Fellow Society of Authors members: check your emails. There's a survey asking for feedback on that "ethical AI charter" from last week… | ↗Not briefed |
arxiv arxiv-ai | 19d | The Effectiveness and Durability of Pretraining Data Filtering in Protein Language Models | Pretraining filtering is one of the most common safeguards for open-weight language models. Here we study its effectiveness, durabilit… | ↗Not briefed |
arxiv arxiv-ai | 19d | From Outcomes to Components: A Counterfactual Evaluation Framework for LLM-Proposed Repairs | An LLM-generated repair action specifies which records to change and by how much. A failed action may still contain a useful target se… | ↗Not briefed |
bluesky bluesky-ai | 19d | I want to believe this is just public "China bad mkay" posturing but at this point I don't even know. It's funny to think a bunch of DHS c… | I want to believe this is just public "China bad mkay" posturing but at this point I don't even know. It's funny to think a bunch of… | ↗Not briefed |
arxiv arxiv-ai | 19d | Reasoning-Aware Multicalibrated Reliability Predictors for Language Models | As people increasingly rely on large language models (LLMs) for information, model providers need to communicate meaningful estimates… | ↗Not briefed |