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 |
|---|---|---|---|---|
hn hn-ai | 2mo | Show HN: Fixrs – Refactoring for long absolute paths and clippy:absolute_paths | — | ↗Not briefed |
arxiv arxiv-ai | 2mo | PIVOT: Physics-Grounded Verification for AI-Generated Audio-Video Detection🔧 Veo | As generative models continue to advance, AI-generated content (AIGC) is becoming increasingly realistic, weakening the artifact cues… | ↗Not briefed |
arxiv arxiv-ai | 2mo | Can We Trust the Judges? Validation of Factuality Evaluation Methods via Answer Perturbation | Evaluating the factual correctness of large language models (LLMs) is vital for many applications. But are our evaluation tools themse… | ↗Not briefed |
hn hn-ai | 2mo | Do the AI Studio test NOW to prove Google doesn't delete chats | — | ↗Not briefed |
bluesky bluesky-ai | 2mo | "AI CEOs say they need to slow the pace of development. But will they?" No, they don’t. What they want is to improve their shitty reputati… | "AI CEOs say they need to slow the pace of development. But will they?" No, they don’t. What they want is to improve their shitty rep… | ↗Not briefed |
arxiv arxiv-ai | 2mo | Don't Count the Edits, Judge by the Outcome Alone: Reward-Based Evaluation for Grammatical Error Correction | Grammatical error correction (GEC) evaluation has traditionally relied on reference or edit overlap, which can penalize valid rewrites… | ↗Not briefed |
rss techmeme | 2mo | OpenAI employees need to oppose its support for Leading the Future, which seeks to chill federal regulation, and Congress should hold hearings on AI risks (Matthew Yglesias/Slow Boring) | Matthew Yglesias / Slow Boring: OpenAI employees need to oppose its support for Leading the Future, which seeks to chill federal regul… | ↗Not briefed |
bluesky bluesky-ai | 2mo | Epic Games collaborated with Kingdom Hearts because the main character is named after their favourite AI model. | Epic Games collaborated with Kingdom Hearts because the main character is named after their favourite AI model. | ↗Not briefed |
bluesky bluesky-ai | 2mo | I think it was clear what people would use AI for right from the start. | I think it was clear what people would use AI for right from the start. | ↗Not briefed |
bluesky bluesky-ai | 2mo | AI leaders including Anthropic CEO Dario Amodei and OpenAI CEO Sam Altman are calling attention to the technology’s rapid growth and advoca… | AI leaders including Anthropic CEO Dario Amodei and OpenAI CEO Sam Altman are calling attention to the technology’s rapid growth and a… | ↗Not briefed |
arxiv arxiv-ai | 2mo | Bayesian Optimisation Using Product-of-Experts Gaussian Process Models with Uncertainty Calibration | Bayesian optimisation (BO) typically relies on a single global Gaussian process (GP) model as its surrogate model. However, GP regress… | ↗Not briefed |
hn hn-ai | 2mo | Seabirds produce $470M worth of guano every year (2020) | — | ↗Not briefed |
hn hn-ai | 2mo | METR: The New AI Gatekeepers and the Quiet Path to State Control of AI | — | ↗Not briefed |
bluesky bluesky-ai | 2mo | I'm sharing my baby again because three 🫏 groups are trying to steal her sideways through the Anthropic Settlement. So your support of A… | I'm sharing my baby again because three 🫏 groups are trying to steal her sideways through the Anthropic Settlement. So your support… | ↗Not briefed |
bluesky bluesky-ai | 2mo | A lot of the calls for "frontier throttling" is probably less concerns about existential risk than getting out of the CapEx rat race. Not … | A lot of the calls for "frontier throttling" is probably less concerns about existential risk than getting out of the CapEx rat race.… | ↗Not briefed |
bluesky bluesky-ai | 2mo | Anthropic disclosed last week that it blocked efforts by bad actors to use its AI models for malicious activity, such as cyberattacks, surv… | Anthropic disclosed last week that it blocked efforts by bad actors to use its AI models for malicious activity, such as cyberattacks,… | ↗Not briefed |
bluesky bluesky-ai | 2mo | LMAO, Nvidia, Palantir, and Booz are "limiting" use of LLM models b/c of "fears about data privacy." Everyone's looking for an angle to r… | LMAO, Nvidia, Palantir, and Booz are "limiting" use of LLM models b/c of "fears about data privacy." Everyone's looking for an angle… | ↗Not briefed |
hn hn-ai | 2mo | Show HN: Tools that get cheaper every time you use them (floor: one £/$5) | — | ↗Not briefed |
bluesky bluesky-ai | 2mo | Inteligência Artificial é um campo bem vasto, na faculdade eu tive 2 matérias de IA e uma não tinha nada a ver com a outra. LLMs são UM tip… | Inteligência Artificial é um campo bem vasto, na faculdade eu tive 2 matérias de IA e uma não tinha nada a ver com a outra. LLMs são U… | ↗Not briefed |
bluesky bluesky-ai | 2mo | Künstliche Intelligenz lud Schadsoftware auf eine Plattform mit Millionen Nutzer:innen. Ein Cybersicherheitsforscher aus Österreich verhind… | Künstliche Intelligenz lud Schadsoftware auf eine Plattform mit Millionen Nutzer:innen. Ein Cybersicherheitsforscher aus Österreich ve… | ↗Not briefed |
hn hn-ai | 2mo | Repology.org domain appears to be on registrar hold (resolves to 127.0.0.1) | — | ↗Not briefed |
hn hn-ai | 2mo | Show HN: Salmon – tray and desktop alerts for systemd services and custom checks | — | ↗Not briefed |
arxiv arxiv-ai | 2mo | The Token Before the Value Is the Key: How Hybrid Architectures Organize Induction Circuits | Hybrid language models can improve capability as well as efficiency, raising the question of how architectural complementarity becomes… | ↗Not briefed |
arxiv arxiv-ai | 2mo | Specifying Reward Functions for RL Without Environment Sampling | Enabling human stakeholders to specify reward functions that lead to their desired outcomes is a key challenge in deploying reinforcem… | ↗Not briefed |
bluesky bluesky-ai | 2mo | One thing that irritates me is the idea that "China" moving ahead on AI is somehow supposed to be bad for "us", as if Anthropic and OpenAI … | One thing that irritates me is the idea that "China" moving ahead on AI is somehow supposed to be bad for "us", as if Anthropic and Op… | ↗Not briefed |
arxiv arxiv-ai | 2mo | Strong and Compact Policies for Submodular Markov Decision Processes via LP-Based Submodular Orienteering | Finding policies for Markov Decision Processes (MDPs) is a central problem in areas such as Reinforcement Learning and Operations Rese… | ↗Not briefed |
bluesky bluesky-ai | 2mo | Spiegel kann es gar nicht fassen, dass KI gar nicht so schlimm ist wenn man seine IT nicht vollkommen bescheuert betreibt. Aber die kognit… | Spiegel kann es gar nicht fassen, dass KI gar nicht so schlimm ist wenn man seine IT nicht vollkommen bescheuert betreibt. Aber die k… | ↗Not briefed |
bluesky bluesky-ai | 2mo | Every time I do a thread like this asshole dudes show up to mansplain AI vs LLM vs Machine Learning to me. Bitch. Go write a textbook and a… | Every time I do a thread like this asshole dudes show up to mansplain AI vs LLM vs Machine Learning to me. Bitch. Go write a textbook… | ↗Not briefed |
bluesky bluesky-ai | 2mo | iOS 27 code shows you may be able to replace Siri AI with Claude or ChatGPT [Poll]🔧 ChatGPT | iOS 27 code shows you may be able to replace Siri AI with Claude or ChatGPT [Poll] | ↗Not briefed |
rss techmeme | 2mo | Netflix, Amazon, and YouTube launch the Streaming Access and Choice Alliance, led by trade group TechNet, to advocate for "technology-neutral policies" (Sara Fischer/Axios) | Sara Fischer / Axios: Netflix, Amazon, and YouTube launch the Streaming Access and Choice Alliance, led by trade group TechNet, to adv… | ↗Not briefed |
arxiv arxiv-ai | 2mo | The Misery of Mechanistic Interpretability: A Formal Perspective🔧 Gemma 3 | Mechanistic interpretability has become the dominant lens for understanding frontier language models, as their inner workings are comp… | ↗Not briefed |
bluesky bluesky-ai | 2mo | this is very good (and it was fully insane of the Anthropic guy to say we have >10% chance of extinction in 10 years) but we should make su… | this is very good (and it was fully insane of the Anthropic guy to say we have >10% chance of extinction in 10 years) but we should ma… | ↗Not briefed |
hn hn-ai | 2mo | An AI deal with China could get Donald Trump his Nobel Prize | — | ↗Not briefed |
hn hn-ai | 2mo | Show HN: Pelican-bicycle alternatives (updated for 2026)🔧 OpenRouter | In November and December 2025, inspired by Simon Willison’s pelican-riding-a-bicycle benchmark, I had some then-current LLMs create SV… | ↗Not briefed |
arxiv arxiv-ai | 2mo | Option-Aware Retrieval and Task-Specific VLM Adaptation for Medical VQA | We describe our submission to the MedReason 2026 challenge, covering multiple-choice (MCQ) and open-ended (OE) medical visual question… | ↗Not briefed |
arxiv arxiv-ai | 2mo | To Each Language Its Tokenizer: Modular Tokenizers for Efficient Multilingual LLMs | Multilingual Large Language Models (LLMs) traditionally rely on a single vocabulary shared by all supported languages, which can lead… | ↗Not briefed |
arxiv arxiv-ai | 2mo | Beyond Noise: Understanding and Overcoming Temperature Effects in Analog DNN Inference | The energy efficiency of analog computing makes it one of the most promising candidates for deploying resource-intensive machine learn… | ↗Not briefed |
bluesky bluesky-ai | 2mo | deepseek v4.1 flash is basically legally blind unfortunately and it leads to absolute peak situations like this | deepseek v4.1 flash is basically legally blind unfortunately and it leads to absolute peak situations like this | ↗Not briefed |
bluesky bluesky-ai | 2mo | Nach zahlreichen Vorfällen wollen KI-Chefs das Entwicklungstempo bremsen. Bei einer jüngst bekannt gewordenen Attacke einer Anthropic-KI gr… | Nach zahlreichen Vorfällen wollen KI-Chefs das Entwicklungstempo bremsen. Bei einer jüngst bekannt gewordenen Attacke einer Anthropic-… | ↗Not briefed |
rss techmeme | 2mo | Microsoft publishes a 37-page "humanist AI code of conduct", establishing that "people matter more than AI", rejecting legal personhood for AI, and more (Tom Warren/The Verge) | Tom Warren / The Verge: Microsoft publishes a 37-page “humanist AI code of conduct”, establishing that “people matter more than AI”, r… | ↗Not briefed |
hn hn-ai | 2mo | Show HN: LinXiv v0.5.0: Headless research-DB image with iroh for self-hosting | I am making a lot of progress on the headless research-DB node designed to be somewhat premade for a research group wanting to aggrega… | ↗Not briefed |
arxiv arxiv-ai | 2mo | Automating Attack Graph Construction for Agentic Pentesting. Towards Neuro-Symbolic Vulnerability Hunting | Logic attack graphs grounded in scanner output provide explicit and auditable attack path reasoning LLM-based agents lack. Integrating… | ↗Not briefed |
arxiv arxiv-ai | 2mo | Psychosis involves a deficit of information compression in connected speech | Large language models (LLMs) with human-like performance on linguistic tasks have transformed the study of language in neurodiverse co… | ↗Not briefed |
arxiv arxiv-ai | 2mo | End-to-End Verifiable and Robust Federated Learning | Federated learning enables multiple parties to train a shared model without centralizing raw data with the help of an aggregator, but… | ↗Not briefed |
bluesky bluesky-ai | 2mo | ver um monte de gente séria caindo na pr da anthropic sobre como a máquina de plágio vai destruir o mundo é meio bobo | ver um monte de gente séria caindo na pr da anthropic sobre como a máquina de plágio vai destruir o mundo é meio bobo | ↗Not briefed |
arxiv arxiv-ai | 2mo | Beyond Safe Answers: Segment-Aware Listwise Alignment for Reasoning Safety in Large Reasoning Models | Large Reasoning Models (LRMs) pose a dual-surface safety challenge: both intermediate reasoning traces and final answers can contain h… | ↗Not briefed |
hn hn-ai | 2mo | The Complete Guide to pstack Pt. 2 | — | ↗Not briefed |
bluesky bluesky-ai | 2mo | Also, what does Axios think “political norms” are? | Also, what does Axios think “political norms” are? | ↗Not briefed |
hn hn-ai | 2mo | Show HN: Real-world attack credential lists from our own honeypot network | — | ↗Not briefed |
hn hn-ai | 2mo | Show HN: Self-hosted coffee app for tracking espresso shots, with AI assistance | — | ↗Not briefed |