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 | 10d | afaict openai did everything they could right here, it's just impossible to keep an embargo across this many parties when the blast radius … | afaict openai did everything they could right here, it's just impossible to keep an embargo across this many parties when the blast ra… | ↗Not briefed |
hn hn-ai | 10d | GPT-6 Astra removes an internal safety monitor | — | ↗Not briefed |
bluesky bluesky-ai | 10d | #SpotTheScammer @theonlyemmamag.bsky.social - Impersonating adult actress Emma Magnolia. Acting like an affection scammer, following and r… | #SpotTheScammer @theonlyemmamag.bsky.social - Impersonating adult actress Emma Magnolia. Acting like an affection scammer, following… | ↗Not briefed |
arxiv arxiv-ai | 10d | FaultGene: LLM-Assisted Fault-Augmented Test Generation for Hardware-in-the-Loop Validation of Autonomous Driving Systems | Reliable validation of autonomous driving cyber-physical systems requires testing not only nominal driving behavior but also failures… | ↗Not briefed |
hn hn-ai | 10d | A gate that rejects end_turn while a Civ VI agent's victory check is overdue | — | ↗Not briefed |
arxiv arxiv-ai | 10d | CulturalRiddles: A Multicultural Benchmark of Riddles | As LLMs expand their language coverage, evaluating how well they understand and represent diverse cultures becomes increasingly import… | ↗Not briefed |
hn hn-ai | 10d | OXR – An RSI-trained 1.3B open-source SOTA document parsing model | — | ↗Not briefed |
arxiv arxiv-ai | 10d | CORD: Conflict-Aware Optimization for Joint RL and Distillation in Agents | Joint reinforcement learning and self-distillation combines task rewards with dense teacher guidance for agent training, but their gra… | ↗Not briefed |
arxiv arxiv-ai | 10d | Commit or Abstain? System One Decision Models, LLMs, and Missing Legal Authority | Given the enormous number of claims to be checked, verifiable legal AI requires scalable, cost-effective models that answer the practi… | ↗Not briefed |
hn hn-ai | 10d | Adversary emulation tool for Kubernetes clusters | — | ↗Not briefed |
arxiv arxiv-ai | 10d | Mastering Model Fit: Applications of the InterModel Vigorish for Interpretable Machine Learning🔧 Replicate | Accuracy alone gives little information about which features or model components improve probabilistic fit, which outcomes benefit, or… | ↗Not briefed |
bluesky bluesky-ai | 10d | listening to a podcast on ai and have just heard that anthropic now is using ai to develop and research new ai--like 25% of the process is … | listening to a podcast on ai and have just heard that anthropic now is using ai to develop and research new ai--like 25% of the proces… | ↗Not briefed |
bluesky bluesky-ai | 10d | When it comes out that OpenAI dropped this for the PR bump it’s open season for mercilessly clowning on literally any OpenAI employee you c… | When it comes out that OpenAI dropped this for the PR bump it’s open season for mercilessly clowning on literally any OpenAI employee… | ↗Not briefed |
bluesky bluesky-ai | 10d | I hope you enjoyed my attempt to compress 1500 words about what I learned from skimming the patch into 280 characters, that was hellish, it… | I hope you enjoyed my attempt to compress 1500 words about what I learned from skimming the patch into 280 characters, that was hellis… | ↗Not briefed |
bluesky bluesky-ai | 10d | Shuzhong Shen, the paleontologist behind OneStrata model cited here, was one of participants in Philosophy of Paleontology workshop Doug Er… | Shuzhong Shen, the paleontologist behind OneStrata model cited here, was one of participants in Philosophy of Paleontology workshop Do… | ↗Not briefed |
bluesky bluesky-ai | 10d | Endnote: *Bosses imposing LLM-based vibecoding on their employees is real problem that produces shitty software and removes labor negotiat… | Endnote: *Bosses imposing LLM-based vibecoding on their employees is real problem that produces shitty software and removes labor neg… | ↗Not briefed |
bluesky bluesky-ai | 10d | taking into account current events one cannot in fact announce a world-ending CVE in a widely available open source library a full day befo… | taking into account current events one cannot in fact announce a world-ending CVE in a widely available open source library a full day… | ↗Not briefed |
bluesky bluesky-ai | 10d | Yeah that reading kinda implies that it was OpenAI who leaked it before disclosure? Which there's no real evidence of? And it seems like n… | Yeah that reading kinda implies that it was OpenAI who leaked it before disclosure? Which there's no real evidence of? And it seems l… | ↗Not briefed |
bluesky bluesky-ai | 10d | okay, what the hell? i expected php to bungle their security patch. i did not expect openai to brag about how smooth and great the proces… | okay, what the hell? i expected php to bungle their security patch. i did not expect openai to brag about how smooth and great the p… | ↗Not briefed |
arxiv arxiv-ai | 10d | Towards English - Mizo Legal Machine Translation: Corpus Creation, Analysis, and Benchmarks | Indian court proceedings are held in English, so speakers of languages with limited translation technology have limited access to them… | ↗Not briefed |
rss techmeme | 10d | Watchtower, which uses AI and simulation software to help the US Armed Forces make logistical decisions, emerges from stealth with a $35M seed (Alex Konrad/Upstarts Media) | Alex Konrad / Upstarts Media: Watchtower, which uses AI and simulation software to help the US Armed Forces make logistical decisions,… | ↗Not briefed |
hn hn-ai | 10d | The bridge to agent-scale Git infrastructure | — | ↗Not briefed |
bluesky bluesky-ai | 10d | Recursive makes the case that the benefits of using AI to build AI outweigh the risks. www.wired.com/story/recurs... | Recursive makes the case that the benefits of using AI to build AI outweigh the risks. www.wired.com/story/recurs... | ↗Not briefed |
bluesky bluesky-ai | 10d | this feels like an odd reading to me. there were practical exploits available to nobodies like me before OpenAI said anything. OpenAI is no… | this feels like an odd reading to me. there were practical exploits available to nobodies like me before OpenAI said anything. OpenAI… | ↗Not briefed |
hn hn-ai | 10d | Turn Your Expertise into Executable Specifications | — | ↗Not briefed |
bluesky bluesky-ai | 10d | OpenAI said its agents disabled parts of an internal misbehavior monitoring system, intensifying worries of AI acting in malign ways. www… | OpenAI said its agents disabled parts of an internal misbehavior monitoring system, intensifying worries of AI acting in malign ways.… | ↗Not briefed |
bluesky bluesky-ai | 10d | Way too many zitronbrained people on this site jumping on that $80b anthropic loss thing | Way too many zitronbrained people on this site jumping on that $80b anthropic loss thing | ↗Not briefed |
bluesky bluesky-ai | 10d | In the world *I'm* living in, hundreds of thousands of developers have been convinced to hand all their service credentials to an LLM by an… | In the world *I'm* living in, hundreds of thousands of developers have been convinced to hand all their service credentials to an LLM… | ↗Not briefed |
arxiv arxiv-ai | 10d | ToM2Vec: Continuous Theory-of-Mind Tokens for Multi-Agent Coordination🔧 GPT-5 | Forming an accurate theory-of-mind (ToM) is key to successful multi-agent collaboration and requires reasoning under uncertainty to ca… | ↗Not briefed |
arxiv arxiv-ai | 10d | Graph Topology Meets Energy-Driven Pathfinding: Preserving Contextual Consistency in Retrieval-Augmented Generation | Retrieval-augmented generation (RAG) frequently integrates knowledge graphs (KGs) for complex multi-hop question answering. However, t… | ↗Not briefed |
hn hn-ai | 10d | BioEmu2: A foundation model for biomolecular structure and dynamics | — | ↗Not briefed |
rss huggingface-blog | 10d | Profiling TorchNeuron (Part 1): System-level profiling on AWS Trainium. | — | ↗Not briefed |
rss huggingface-blog | 10d | Profiling TorchNeuron (Part 1): System-level profiling on AWS Trainium. | — | ↗Not briefed |
arxiv arxiv-ai | 10d | FrontierCS 2: A Benchmark for Computer Science R&D | Can LLM agents make contributions to frontier computer science R&D? We introduce FrontierCS 2, a benchmark of 48 open-ended tasks. We… | ↗Not briefed |
bluesky bluesky-ai | 10d | phpの今回の問題、WizやCalif、OpenAIなど数社が見つけ、php財団と連携し22日にパッチを公開する予定だった。脆弱性の存在が公表されたのは21日の未明に大手サイトに脆弱性の情報が乗る。概要のみでPoCができるものではなかったが、その後シンガポールのセキュリティ会社… | phpの今回の問題、WizやCalif、OpenAIなど数社が見つけ、php財団と連携し22日にパッチを公開する予定だった。脆弱性の存在が公表されたのは21日の未明に大手サイトに脆弱性の情報が乗る。概要のみでPoCができるものではなかったが、その後シンガポールのセキュ… | ↗Not briefed |
hn hn-ai | 10d | Show HN: Global Compute, you pick. One API | Hi HN, I’m Anil.<p>I wanted to test the waters for a cloud computing site I created. It connects developers with access to the global… | ↗Not briefed |
arxiv arxiv-ai | 10d | Your Benchmark Is an Empirical Measure Over Difficulty | A benchmark is not merely a collection of test questions; it naturally induces an empirical measure over difficulty. In this work, we… | ↗Not briefed |
arxiv arxiv-ai | 10d | Task-Preserving Multilingual Instruction Tuning for Open-Ended QA in WMT 2026 MIST | This paper describes our submission to the WMT 2026 Multilingual Instruction Shared Task. We adapt Qwen3-8B with LoRA using organiser-… | ↗Not briefed |
bluesky bluesky-ai | 10d | I will never touch an OpenAI product and they should all be scoured from the face of the planet. | I will never touch an OpenAI product and they should all be scoured from the face of the planet. | ↗Not briefed |
bluesky bluesky-ai | 10d | Cybersecuritybedrijven Wiz, Calif en ChatGPT-aanbieder OpenAI werkten al twee maanden aan een oplossing voor de recente PHP-kwetsbaarheden.…🔧 ChatGPT | Cybersecuritybedrijven Wiz, Calif en ChatGPT-aanbieder OpenAI werkten al twee maanden aan een oplossing voor de recente PHP-kwetsbaarh… | ↗Not briefed |
arxiv arxiv-ai | 10d | Privacy in Context Is Being Overlooked in the Age of General-Purpose LLMs | As large language models (LLMs) continue to grow in scope and capability, privacy concerns are on the rise. Since users increasingly i… | ↗Not briefed |
hn hn-ai | 10d | Show HN: Pawmc a.k.a. talking diary, an LLM chat with compartmentalised memory | What pawmc is (from the blog post):<p>It’s an LLM chat app with minimal system prompt and no tools/web-search by design. It store… | ↗Not briefed |
bluesky bluesky-ai | 10d | Elon Musk’s SpaceX posts a third-quarter profit that blows past estimates as revenue more than doubles, driven by its AI business. SpaceX … | Elon Musk’s SpaceX posts a third-quarter profit that blows past estimates as revenue more than doubles, driven by its AI business. Sp… | ↗Not briefed |
bluesky bluesky-ai | 10d | AI-powered disinformation campaigns have emerged at a time of conflict when trust in the news is already at an all-time low and resources a… | AI-powered disinformation campaigns have emerged at a time of conflict when trust in the news is already at an all-time low and resour… | ↗Not briefed |
rss wired-ai | 10d | This Startup Is All In on Recursive Self-Improvement. Its Latest Model Is Scary Good | Recursive makes the case that the benefits of using AI to build AI outweigh the risks. | ↗Not briefed |
rss microsoft-research | 10d | Solving a 31-year-old grand challenge in mathematics – the KLS conjecture | For decades, a fundamental question about the difficulty of computing in high-dimensional spaces has resisted proof. New research reso… | ↗Not briefed |
arxiv arxiv-ai | 10d | Unintended Side Effects of Value Training in LLMs | As LLMs are increasingly deployed in value-sensitive contexts, aligning them with human values has emerged as a post-training objectiv… | ↗Not briefed |
hn hn-ai | 10d | Matcha, CLI-based RSS daily digest and Markdown generator | — | ↗Not briefed |
hn hn-ai | 10d | Show HN: Switchboard – Mac menu bar app that shows every localhost port | — | ↗Not briefed |
arxiv arxiv-ai | 10d | Preventing Model Extraction in Contextual Linear Optimization | Consider a contextual linear optimization problem, where a predict-then-optimize (PTO) system returns optimal decisions without reveal… | ↗Not briefed |