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 | 13d | Is Complex Agentic Memory Always Needed? Naive RAG Is Stronger Than We Thought | LLM memory is widely used in long-term conversation understanding. Agentic memory systems often use mutation, structural organization,… | ↗Not briefed |
arxiv arxiv-ai | 13d | The Delegation Divide: An Open Index of AI-Agent Delegability across Occupations and Countries🔧 Consensus | 35.2% of task time in world employment is at least partly delegable to AI agents in 2026, and only 2.5% can be handed over in full. Ex… | ↗Not briefed |
arxiv arxiv-ai | 13d | How Firm Are AI Models' Ethical Boundaries? A Pay-Stub Editing Test of 15 Models | Can a small change in wording move an AI model from refusing a sensitive request to acting on it? We test 15 models on the same reques… | ↗Not briefed |
hn hn-ai | 13d | Show HN: TenseMem – deterministic memory for AI agents | — | ↗Not briefed |
bluesky bluesky-ai | 13d | So I earn more than Anthropic! | So I earn more than Anthropic! | ↗Not briefed |
arxiv arxiv-ai | 13d | Visualizing and Interpreting Temporal Care Team Dynamics for Cancer Survival Risk Analysis | Cancer care is a longitudinal, time-sensitive process requiring coordination among diverse healthcare professionals (HCPs), whose coll… | ↗Not briefed |
arxiv arxiv-ai | 13d | Wavenumber-Aware Deep Learning for Robust Analysis of Raman Spectra | Machine-learning models for spectroscopy usually treat a spectrum as a fixed-grid intensity vector, leaving the physical coordinate im… | ↗Not briefed |
arxiv arxiv-ai | 13d | Random Remasking Scales Masked Diffusion Inference | Masked diffusion models (MDMs) generate sequences by progressively unmasking tokens, but errors introduced during decoding are typical… | ↗Not briefed |
arxiv arxiv-ai | 13d | UniGuard: Multi-View Guardrail Learning for Cross-Interface Safety Generalization🔧 GPT-5 | Guardrail models are increasingly used as safety layers for LLMs and agentic systems, but most are trained on fixed prompt or response… | ↗Not briefed |
bluesky bluesky-ai | 13d | LLM 으로 만든 물건의 상업적 판매는 그냥 금지시켜야 하는 게 아닐까. | LLM 으로 만든 물건의 상업적 판매는 그냥 금지시켜야 하는 게 아닐까. | ↗Not briefed |
hn hn-ai | 13d | Show HN: Commit Quiz – A Git hook to review AI-generated changes | — | ↗Not briefed |
arxiv arxiv-ai | 13d | AgentTriage: Replayable Security Diagnosis Across Heterogeneous Agent Runtimes | When tool-use agents write files, execute code, or send data, their final responses do not explain the resulting security events, and… | ↗Not briefed |
bluesky bluesky-ai | 13d | hey guys do we remember when all those vulnerabilities reported to firefox by anthropic or whichever dumbfucks turned out to be like 90% bu… | hey guys do we remember when all those vulnerabilities reported to firefox by anthropic or whichever dumbfucks turned out to be like 9… | ↗Not briefed |
bluesky bluesky-ai | 13d | OpenAI and Wiz followed responsible security research practices… worked with PHP on patches before publicly disclosing… worked together to … | OpenAI and Wiz followed responsible security research practices… worked with PHP on patches before publicly disclosing… worked togethe… | ↗Not briefed |
rss techmeme | 13d | Epic says "Fortnite is coming back to the App Store in Australia tomorrow" after an Australian federal judge ordered Apple to open up iOS in an antitrust case (Rhea Rose Abraham/Reuters) | Rhea Rose Abraham / Reuters: Epic says “Fortnite is coming back to the App Store in Australia tomorrow” after an Australian federal ju… | ↗Not briefed |
bluesky bluesky-ai | 13d | i've been on a mailing list for a somewhat esoteric OS project for a while now, and I've just seen a thread where the lead maintainer went … | i've been on a mailing list for a somewhat esoteric OS project for a while now, and I've just seen a thread where the lead maintainer… | ↗Not briefed |
bluesky bluesky-ai | 13d | really interesting how "being reasonably disappointed that the internet archive used an LLM to piss out a fix" is being framed as people lo… | really interesting how "being reasonably disappointed that the internet archive used an LLM to piss out a fix" is being framed as peop… | ↗Not briefed |
bluesky bluesky-ai | 13d | what the fuck are you talking about. automation of deployment is THE FUCKING TECHNOLOGY of basically every type of modern software developm… | what the fuck are you talking about. automation of deployment is THE FUCKING TECHNOLOGY of basically every type of modern software dev… | ↗Not briefed |
arxiv arxiv-ai | 13d | Width and Depth Limits Commute During Training | Recent work established that the infinite-width and infinite-depth limits of residual networks commute at initialization when residual… | ↗Not briefed |
rss marktechpost | 13d | Perplexity and NVIDIA Researchers Introduce JOLT: a Joint Teacher-Student Reinforcement Learning Method | Perplexity and NVIDIA introduced JOLT, a post-training method that trains one model as both teacher and student. On MATH it matched GR… | ↗Not briefed |
arxiv arxiv-ai | 13d | Federated Graph Retrieval-Augmented Generation | Graph Retrieval-Augmented Generation (GraphRAG) leverages centralized knowledge graphs to enhance large language models with structure… | ↗Not briefed |
rss techmeme | 13d | An Australian federal judge rules Apple must let Australian iPhone users install apps from third-party app stores starting March 1, 2027; Apple plans to appeal (Marcus Mendes/9to5Mac) | Marcus Mendes / 9to5Mac: An Australian federal judge rules Apple must let Australian iPhone users install apps from third-party app st… | ↗Not briefed |
arxiv arxiv-ai | 13d | Faithful Latent Harm Monitors Work No Better Than the Models Own Behaviour at a Matched False Alarm Rate | Large language models have been reported to encode the harmfulness of a request separately from their decision to refuse it, and laten… | ↗Not briefed |
bluesky bluesky-ai | 13d | I have done LLM assisted brainstorming. It was fun. But I did notice it often wasn't the input from the LLM I needed. So I pivoted towards … | I have done LLM assisted brainstorming. It was fun. But I did notice it often wasn't the input from the LLM I needed. So I pivoted tow… | ↗Not briefed |
hn hn-ai | 13d | Test Applications on the Edge | — | ↗Not briefed |
bluesky bluesky-ai | 13d | niri has this really cute tic of "it'll keep" i dont see any other llm replicate🔧 Replicate | niri has this really cute tic of "it'll keep" i dont see any other llm replicate | ↗Not briefed |
arxiv arxiv-ai | 13d | Learning from Next-Frame Prediction: Towards Unified and Scalable Visual Representation Learning | Video provides a natural source of supervision: what happens next serves as a learning signal for visual world modeling. Yet masked vi… | ↗Not briefed |
hn hn-ai | 13d | Lemoncode, Free Coding Agent | — | ↗Not briefed |
arxiv arxiv-ai | 13d | Solvable models of learning to pursue a moving target | Agents tasked with intercepting a moving target must learn how to reach it, ideally as efficiently as possible. As observers of animal… | ↗Not briefed |
hn hn-ai | 13d | Lift and Slide Doors: How They Work and When They Make Sense for Large Openings | — | ↗Not briefed |
bluesky bluesky-ai | 13d | oh jesus fucking christ deepseek. your job didn't have anything to do with docker. | oh jesus fucking christ deepseek. your job didn't have anything to do with docker. | ↗Not briefed |
arxiv arxiv-ai | 13d | Meaning Contracts: Local Semantic Geometry in Large Language Models | This paper studies the local geometry of semantic meaning in large language models. Current representation analyses ask whether meanin… | ↗Not briefed |
bluesky bluesky-ai | 13d | Had I thought long enough, the concept of return-oriented programming likely would have come to me. The Mythos whitepaper is what clued m… | Had I thought long enough, the concept of return-oriented programming likely would have come to me. The Mythos whitepaper is what cl… | ↗Not briefed |
arxiv arxiv-ai | 13d | AIRO: A Streaming-Based Machine Learning Framework for Threat Detection and Security in Industrial Internet of Things | Modern Industrial Internet of Things (IIoT) im- plementations now connect edge gateways, cloud dashboards and business applications. T… | ↗Not briefed |
youtube youtube-aiexplained | 13d | What AI Researchers Saw, Before Their Demand to ‘Pace’ AI🔧 Sora | A significant [clears throat] percentage of the human race has now seen or heard quoted the Jacob Coxen tweet with his words on AI lab… | ↗Not briefed |
arxiv arxiv-ai | 13d | ULME: Predictive Coupling for Mutually Reinforcing Embedding and Language Modeling | Text embedding and text generation are two of the most useful capabilities of modern large language models, yet they are typically pro… | ↗Not briefed |
bluesky bluesky-ai | 13d | "reverse engineer using ai" more like use ai to blatantly plagiarise games STILL SUPPORTED BY UPDATES I never wanted to hear that someone … | "reverse engineer using ai" more like use ai to blatantly plagiarise games STILL SUPPORTED BY UPDATES I never wanted to hear that som… | ↗Not briefed |
rss marktechpost | 13d | A Coding Guide to Cactus Whistle: Does a 16.9 MB Speech Model Hold Up Against Whisper Base? Word Error Rate, Speed, Streaming, Timestamps, Keywords, and Speech-to-Tool Calls on a CPU🔧 Whisper | Explore a comprehensive coding guide benchmarking Cactus Compute's 16.9 MB speech model against Whisper base. Learn about word error r… | ↗Not briefed |
bluesky bluesky-ai | 13d | ah is it like dario news clips or something that are the offputtingly antihuman ones? the ads that I’ve seen from anthropic have been the “… | ah is it like dario news clips or something that are the offputtingly antihuman ones? the ads that I’ve seen from anthropic have been… | ↗Not briefed |
bluesky bluesky-ai | 13d | Every hero in these games has the superpower of empathy or creativity , every attempt to insitutualize or mechanize that fails horribly (th… | Every hero in these games has the superpower of empathy or creativity , every attempt to insitutualize or mechanize that fails horribl… | ↗Not briefed |
bluesky bluesky-ai | 13d | 1000% agree. Sadly it’s been able to do at least two, if not three, of those things even before the LLM era. c.f. “Wired for War” by PJ S… | 1000% agree. Sadly it’s been able to do at least two, if not three, of those things even before the LLM era. c.f. “Wired for War” by… | ↗Not briefed |
arxiv arxiv-ai | 13d | Governance-State Corruption in Long-Term Memory for LLM Agents: A Cross-Model Stress Test | An LLM agent can retrieve a memory correctly yet use it incorrectly after its validity state changes. We study governance-state corrup… | ↗Not briefed |
bluesky bluesky-ai | 13d | the pro ai accounts on here don't work for anthropic and they do a better job at pr than the people who are paid by them to talk about it o… | the pro ai accounts on here don't work for anthropic and they do a better job at pr than the people who are paid by them to talk about… | ↗Not briefed |
bluesky bluesky-ai | 13d | opencode has deepseek flash for free and it’ll happily redteam your infra without issue. you’re irresponsible if you don’t try it. | opencode has deepseek flash for free and it’ll happily redteam your infra without issue. you’re irresponsible if you don’t try it. | ↗Not briefed |
arxiv arxiv-ai | 13d | Disentangling Robustness and Sycophancy in Multilingual LLMs | When a multilingual LLM agrees with a user's false premise, is it being sycophantic, or does it simply not know better in that languag… | ↗Not briefed |
hn hn-ai | 13d | Open Language Models – Nathan Lambert | — | ↗Not briefed |
bluesky bluesky-ai | 13d | "people who are mad at the Internet Archive for using Claude to fix their website because LLM training is 'plagiarism'" is an absolutely IN… | "people who are mad at the Internet Archive for using Claude to fix their website because LLM training is 'plagiarism'" is an absolute… | ↗Not briefed |
bluesky bluesky-ai | 13d | okay the swift rewrite of the bluesky app is VERY good it cost about 40 dollars in deepseek tokens then about 150-200 in sonnet tokens to … | okay the swift rewrite of the bluesky app is VERY good it cost about 40 dollars in deepseek tokens then about 150-200 in sonnet token… | ↗Not briefed |
arxiv arxiv-ai | 13d | LoRA Superposition across Pretrained Models | Low-rank adaptation (LoRA) is usually tied to the pretrained backbone on which it is learned. We ask whether LoRAs trained independent… | ↗Not briefed |
arxiv arxiv-ai | 13d | From Clinical Judgment to Checkable Rules: Clinician-Led Guardrails for Mental-Health Chatbots | Large language models (LLMs) deployed in mental-health contexts can be evaluated for safety failures, but the resulting judgments are… | ↗Not briefed |