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 | 18d | Auditing Safety Preference Supervision with Recorded Derivations | Preference labels align model training, but the judgments behind them are difficult to inspect or verify. In this paper, we examine ho… | ↗Not briefed |
bluesky bluesky-ai | 18d | AI-generated photos falsely showing an elderly couple stripped and robbed during unrest in France drew millions of views as far-right accou… | AI-generated photos falsely showing an elderly couple stripped and robbed during unrest in France drew millions of views as far-right… | ↗Not briefed |
arxiv arxiv-ai | 18d | Solving the Ho--Nguyen Conjecture for Gaussian Mixtures: Applications to Mixtures of Experts | Over-fitted Gaussian mixtures can contain many parameter configurations with nearly the same density, making the order of density canc… | ↗Not briefed |
bluesky bluesky-ai | 18d | "Let us bribe you into supporting our crimes against humanity (with promises we'll never keep)". www.nbcnews.com/tech/tech-ne... | "Let us bribe you into supporting our crimes against humanity (with promises we'll never keep)". www.nbcnews.com/tech/tech-ne... | ↗Not briefed |
arxiv arxiv-ai | 18d | Theorem prover Arend | This paper is an exposition of the theorem prover Arend developed at JetBrains Research. Arend language has been designed to facilitat… | ↗Not briefed |
hn hn-ai | 18d | FloatDoor: Platform-Triggered Backdoors in LLMs | — | ↗Not briefed |
hn hn-ai | 18d | Overmind: Cut PyTorch model load time 15s to 0.2s with a zero-copy shared cache | — | ↗Not briefed |
bluesky bluesky-ai | 18d | LOL: Days after the FBI busted an alleged Saudi espionage operation inside OpenAI, here's the U.S. Embassy in Riyadh proudly celebrating Am… | LOL: Days after the FBI busted an alleged Saudi espionage operation inside OpenAI, here's the U.S. Embassy in Riyadh proudly celebrati… | ↗Not briefed |
bluesky bluesky-ai | 18d | Do mathematicians discover philosophie of science now? Then, the OpenAI case had at least one positive effect for the discipline. | Do mathematicians discover philosophie of science now? Then, the OpenAI case had at least one positive effect for the discipline. | ↗Not briefed |
bluesky bluesky-ai | 18d | This is the sort of thing that gets me to click through and ignore the developer wholesale 🤢🤢 | This is the sort of thing that gets me to click through and ignore the developer wholesale 🤢🤢 | ↗Not briefed |
bluesky bluesky-ai | 18d | GIGAZINE：Claudeが25年以上謎だった暗号規格の数値の起源を解明、NSAが説明なく提供した数値は「1990年代の乱数」だった： https://gigazine.net/news/20261019-ecdsa-seed-origin/ | GIGAZINE：Claudeが25年以上謎だった暗号規格の数値の起源を解明、NSAが説明なく提供した数値は「1990年代の乱数」だった： https://gigazine.net/news/20261019-ecdsa-seed-origin/ | ↗Not briefed |
bluesky bluesky-ai | 18d | When our AI code reviewer's cost per run doubled, we dug into the token usage to find out why. Learn how we used prompt caching and intelli… | When our AI code reviewer's cost per run doubled, we dug into the token usage to find out why. Learn how we used prompt caching and in… | ↗Not briefed |
hn hn-ai | 18d | Show HN: Realtime talking avatar running free locally (90+FPS on RTX4090) | I wanted to give my humble voice AI a face. The only way to do that was to use Tavus or Lemonslice and pay a large amount of money per… | ↗Not briefed |
bluesky bluesky-ai | 18d | Before prompt injection attacks on LLMs, blue boxes let hackers make free long-distance phone calls. Phone companies isolated their control… | Before prompt injection attacks on LLMs, blue boxes let hackers make free long-distance phone calls. Phone companies isolated their co… | ↗Not briefed |
hn hn-ai | 18d | GPT Rosalind | — | ↗Not briefed |
hn hn-ai | 18d | Taco: Cross platform LLM based compressor, eliminates floating point | — | ↗Not briefed |
arxiv arxiv-ai | 18d | Guided Diffusion for Trustworthy Synthetic Text Generation | The rapid evolution of Large Language Models is fueled by massive datasets, yet this reliance poses severe privacy risks, as models ca… | ↗Not briefed |
bluesky bluesky-ai | 18d | "A report… from Anthropic’s offensive research team grabbed attention by describing a new, practical threat, one devoid of human characteri… | "A report… from Anthropic’s offensive research team grabbed attention by describing a new, practical threat, one devoid of human chara… | ↗Not briefed |
arxiv arxiv-ai | 18d | When Models Know but Can't Linearly Say: Relational Encoding Limits Linear Vulnerability Classification in Code LLMs | Large language models can detect vulnerable code, yet the form of vulnerability encoding in their representations remains unclear. We… | ↗Not briefed |
bluesky bluesky-ai | 18d | During the Great Depression there were endless dance marathons. During our LLM Implosion there will be endless emergency podcasting marath… | During the Great Depression there were endless dance marathons. During our LLM Implosion there will be endless emergency podcasting m… | ↗Not briefed |
arxiv arxiv-ai | 18d | TxnQA: Benchmarking Analytical Reasoning and Evidence Grounding over Transaction Event Sequences | Analyzing merchant transaction histories requires dynamic temporal filtering, entity resolution, and aggregation over irregular event… | ↗Not briefed |
hn hn-ai | 18d | Show HN: Hiding messages as a generative art; open sourcing my code from 2020 | I've always been fascinated by secret messages, and for years I scribbled ideas for a graphical way of hiding messages that would… | ↗Not briefed |
arxiv arxiv-ai | 18d | ValiAssist: Scaling LLM-generated Programming Feedback Validation by Demonstration | Generative AI can create programming feedback at scale, but validating its alignment with instructors' pedagogical goals and strategie… | ↗Not briefed |
arxiv arxiv-ai | 18d | Target-Aware Curation of Hallucination Stress Tests for Large Language Models🔧 Elicit | Hallucination stress tests can yield similar average scores across language models even when different prompts elicit hallucinations i… | ↗Not briefed |
bluesky bluesky-ai | 18d | Anthropic sued by major Canadian news publishers for copyright claims in Ontario court Newspapers say Anthropic infringed on their copyrigh… | Anthropic sued by major Canadian news publishers for copyright claims in Ontario court Newspapers say Anthropic infringed on their cop… | ↗Not briefed |
hn hn-ai | 18d | Ex OpenAI engineer accused of planting bug was in contact with Saudi royals | — | ↗Not briefed |
bluesky bluesky-ai | 18d | The Saudi government offering to pay Abdulrahman Alfozan's bond did not sit well with a San Francisco court www.wired.com/story/openai.… | The Saudi government offering to pay Abdulrahman Alfozan's bond did not sit well with a San Francisco court www.wired.com/story/op… | ↗Not briefed |
hn hn-ai | 18d | Android for running Anki Desktop, no proot | — | ↗Not briefed |
hn hn-ai | 18d | Cutting our inference bill by 68% by being obsessive about optimization | — | ↗Not briefed |
arxiv arxiv-ai | 18d | Diagnostic Analysis for Large Language Model Reasoning | Reasoning structures (e.g., chains and graphs) are increasingly used for probabilistic inference, factuality verification, and uncerta… | ↗Not briefed |
bluesky bluesky-ai | 18d | Good. I really hope they win and this is merely the first of innumerable such suits. www.cbc.ca/news/busines... | Good. I really hope they win and this is merely the first of innumerable such suits. www.cbc.ca/news/busines... | ↗Not briefed |
arxiv arxiv-ai | 18d | AutoRedClaw: Automated Red-Teaming for Claw-Style Agents | Agent systems are shifting from task-scoped tool-use agents to persistent, personalized assistants such as OpenClaw. These claw-style… | ↗Not briefed |
hn hn-ai | 18d | Nvidia-windows-vbios-vix-loader – Load a modified VBIOS on Nvidia GPU on Windows | — | ↗Not briefed |
bluesky bluesky-ai | 18d | Downballot Republicans: So you have to make sure that the car is gas-powered and the garage is completely sealed when you turn it on Trump:… | Downballot Republicans: So you have to make sure that the car is gas-powered and the garage is completely sealed when you turn it on T… | ↗Not briefed |
bluesky bluesky-ai | 18d | The "but the fundamentals!" people are having a REALLY hard time now that LLM assistance has kicked the legs out of a whole pile of shit we… | The "but the fundamentals!" people are having a REALLY hard time now that LLM assistance has kicked the legs out of a whole pile of sh… | ↗Not briefed |
rss ai-snake-oil | 18d | New paper: Emerging trends in AI agent evaluation | Understanding the methods, questions, and ecosystem for evaluation | ↗Not briefed |
bluesky bluesky-ai | 18d | Considering that US taxpayers funded the entire development of SpaceX, this is a perfectly reasonable idea. What's unreasonable is Elon con… | Considering that US taxpayers funded the entire development of SpaceX, this is a perfectly reasonable idea. What's unreasonable is Elo… | ↗Not briefed |
bluesky bluesky-ai | 18d | Bit of an odd job here. Anthropic looking for a person to design and run tabletop exercises and wargames. | Bit of an odd job here. Anthropic looking for a person to design and run tabletop exercises and wargames. | ↗Not briefed |
bluesky bluesky-ai | 18d | I would read the hell out of an article collating historic quotes about the inevitability of NFTs and the metaverse from the current crop o… | I would read the hell out of an article collating historic quotes about the inevitability of NFTs and the metaverse from the current c… | ↗Not briefed |
bluesky bluesky-ai | 18d | How about giving a share of SpaceX stock to each child whose sibling or parent starved to death because of his USAID cuts? www.nytimes.com/… | How about giving a share of SpaceX stock to each child whose sibling or parent starved to death because of his USAID cuts? www.nytimes… | ↗Not briefed |
rss techcrunch-ai | 18d | Amazon is taking its biggest shot at Nvidia yet with new TorchNeuron dev tool | Amazon wants to break Nvidia’s hold on PyTorch, the free toolkit used to build most AI apps, so that more developers run their apps on… | ↗Not briefed |
bluesky bluesky-ai | 18d | nope i don't know why there's a donation from anthropic to anthropic but anyway it is $85k in georgia elections anthropic makes the Claud… | nope i don't know why there's a donation from anthropic to anthropic but anyway it is $85k in georgia elections anthropic makes the… | ↗Not briefed |
bluesky bluesky-ai | 18d | no you write like ai cause you use ai | no you write like ai cause you use ai | ↗Not briefed |
arxiv arxiv-ai | 18d | Predicting Linear Relational Decodability Without the Relation | Linear relational embeddings (LREs) approximate a language model's subject-to-object mapping with an affine function. Their faithfulne… | ↗Not briefed |
bluesky bluesky-ai | 18d | i don’t disagree, but i think the specific thing he’s worried about is that trying to make AI more conservative breaks the model and makes … | i don’t disagree, but i think the specific thing he’s worried about is that trying to make AI more conservative breaks the model and m… | ↗Not briefed |
arxiv arxiv-ai | 18d | CEFR-oriented Russian Text Simplification Using Reinforcement Learning | We present a study on controlled automatic simplification of Russian texts using preference-based optimization algorithms with a focus… | ↗Not briefed |
hn hn-ai | 18d | Time to LLM Post | My new hobby: 1) open hn 2) read headlines and maybe the stories behind them, in order 3) if I hit one about AI, close Hn<p>I'm… | ↗Not briefed |
hn hn-ai | 18d | How Lego Is Getting More Adults to Tinker with Their Toy Bricks | — | ↗Not briefed |
bluesky bluesky-ai | 18d | DeepSeek barred from government agency IT systems politi.co/4yxjDmp | DeepSeek barred from government agency IT systems politi.co/4yxjDmp | ↗Not briefed |
arxiv arxiv-ai | 18d | EGI-BENCH: An Enterprise General Intelligence Benchmark Suite🔧 GPT-5 | Enterprise adoption of LLMs in agentic workflows is gated by evaluations that span a wide arc of abilities: CRM operations, heterogene… | ↗Not briefed |