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 | 1mo | galaxy brain: a LLM is ffmpeg arxiv.org/abs/2306.04050 | galaxy brain: a LLM is ffmpeg arxiv.org/abs/2306.04050 | ↗Not briefed |
bluesky bluesky-ai | 1mo | Broke: LLMs are just fancy autocomplete Woke: LLMs are a technology unlike anything we've ever invented, more grown than engineered Bespok… | Broke: LLMs are just fancy autocomplete Woke: LLMs are a technology unlike anything we've ever invented, more grown than engineered B… | ↗Not briefed |
hn hn-ai | 1mo | Show HN: Jinx – a link shortener you host free on GitHub Pages | Jinx is just links, a small static website on GitHub to share short links to your stuff. Add your links to one file, turn on GitHub Pa… | ↗Not briefed |
hn hn-ai | 1mo | Amfly: A deterministic simulation of six fruit fly connectomes in a reward loop | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | Yeh - I'd be interested in what the feedback they have was, AI wanted vs AI not wanted. There is no doubt there is a large bunch of users t… | Yeh - I'd be interested in what the feedback they have was, AI wanted vs AI not wanted. There is no doubt there is a large bunch of us… | ↗Not briefed |
bluesky bluesky-ai | 1mo | my llm tells me, "you can't spell fistula without eula" so yeah. As many arseholes as you need. | my llm tells me, "you can't spell fistula without eula" so yeah. As many arseholes as you need. | ↗Not briefed |
bluesky bluesky-ai | 1mo | Today was another day where I let claude chew on a problem for a few hours so i could show my bosses the tokens i spent and tell them how i… | Today was another day where I let claude chew on a problem for a few hours so i could show my bosses the tokens i spent and tell them… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Offensive tools are cool, but the LLM-hacking work is just amazingly reckless because it is so effective just due to parallelism: it is the… | Offensive tools are cool, but the LLM-hacking work is just amazingly reckless because it is so effective just due to parallelism: it i… | ↗Not briefed |
hn hn-ai | 1mo | Hacking OpenAI | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | Anthropic gave their bots hacking tools, and OpenAI was an easy target. QED. | Anthropic gave their bots hacking tools, and OpenAI was an easy target. QED. | ↗Not briefed |
bluesky bluesky-ai | 1mo | Apollo Global Management is in talks with SoftBank Group about boosting the size of a loan to $9 billion from $5.4 billion to help the Japa… | Apollo Global Management is in talks with SoftBank Group about boosting the size of a loan to $9 billion from $5.4 billion to help the… | ↗Not briefed |
arxiv arxiv-ai | 1mo | When Does Reasoning Help in Machine Translation? A Hierarchical Analysis of LRM Reasoning Traces | Large Reasoning Models increasingly use intermediate traces for machine translation, but it remains unclear when such reasoning helps… | ↗Not briefed |
arxiv arxiv-ai | 1mo | VLA-Scope: Shift-Aware Failure Prediction for Vision-Language-Action Models | Vision-language-action (VLA) models map visual observations and natural-language instructions to robotic actions, but distribution shi… | ↗Not briefed |
bluesky bluesky-ai | 1mo | A thought: current AI writing detection cannot always consistently identify current frontier model writing. HOWEVER, future models will a… | A thought: current AI writing detection cannot always consistently identify current frontier model writing. HOWEVER, future models w… | ↗Not briefed |
hn hn-ai | 1mo | Cache-Tax | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | Guessing we'll be seeing a lot of this in the coming months www.wsj.com/tech/ai/hack... | Guessing we'll be seeing a lot of this in the coming months www.wsj.com/tech/ai/hack... | ↗Not briefed |
bluesky bluesky-ai | 1mo | #AiCommunity #aivideo #CreativeAi Longest prompt I’ve ever used—over a dozen paragraphs, cleaned up with a locally run LLM. First image fro… | #AiCommunity #aivideo #CreativeAi Longest prompt I’ve ever used—over a dozen paragraphs, cleaned up with a locally run LLM. First imag… | ↗Not briefed |
bluesky bluesky-ai | 1mo | "“Our AI content strategy has started a ‘doom loop’ that will hurt the performance of our models and the entire web at the same time,” Bren… | "“Our AI content strategy has started a ‘doom loop’ that will hurt the performance of our models and the entire web at the same time,”… | ↗Not briefed |
bluesky bluesky-ai | 1mo | the sudden push for ai safety is a pretty clear and textbook attempt at regulatory capture by incumbent llm providers | the sudden push for ai safety is a pretty clear and textbook attempt at regulatory capture by incumbent llm providers | ↗Not briefed |
arxiv arxiv-ai | 1mo | From Trainability Diagnostics to Optimization Claims: Boundaries and Controls in Variational Quantum Optimization | Barren plateau diagnostics characterize whether gradient signal remains available for training, but surviving signal need not translat… | ↗Not briefed |
bluesky bluesky-ai | 1mo | wait, he still runs it? i thought he fucked off to anthropic | wait, he still runs it? i thought he fucked off to anthropic | ↗Not briefed |
bluesky bluesky-ai | 1mo | 法務向けAI「Astra for Law」が登場、GPT-6 Astraより法務タスクがとても上手 https://gigazine.net/news/20260918-openai-astra-for-law/ | 法務向けAI「Astra for Law」が登場、GPT-6 Astraより法務タスクがとても上手 https://gigazine.net/news/20260918-openai-astra-for-law/ | ↗Not briefed |
arxiv arxiv-ai | 1mo | Multiclass Semantic Segmentation of Wildland Fire Images Using Context-Aware Centralized Copy-Paste Data Augmentation | Producing accurate annotations for deep learning based image segmentation is both costly and labor intensive. This challenge is especi… | ↗Not briefed |
bluesky bluesky-ai | 1mo | So my first point is that since they use sex as recruitment and ideological enforcement, no, sorry, the sex is relevant. See here clearly … | So my first point is that since they use sex as recruitment and ideological enforcement, no, sorry, the sex is relevant. See here cle… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Yudkowsky is the major thought-leader about AI and the paperclipper specifically, and Dario/Anthropic are actually a rival faction in their… | Yudkowsky is the major thought-leader about AI and the paperclipper specifically, and Dario/Anthropic are actually a rival faction in… | ↗Not briefed |
bluesky bluesky-ai | 1mo | So the rationalist tribe has been losing its war for money and attention in the space to EA and Anthropic for a while, and they are getting… | So the rationalist tribe has been losing its war for money and attention in the space to EA and Anthropic for a while, and they are ge… | ↗Not briefed |
bluesky bluesky-ai | 1mo | "...what I find most consistently fascinating about the training process is how both Google and Anthropic speak to a professional audience … | "...what I find most consistently fascinating about the training process is how both Google and Anthropic speak to a professional audi… | ↗Not briefed |
bluesky bluesky-ai | 1mo | This post is how I discovered the waterfox team has used, and will likely continue using claude to develop waterfox Why is every major pro… | This post is how I discovered the waterfox team has used, and will likely continue using claude to develop waterfox Why is every majo… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Hackers Used Anthropic’s Claude to Break Into OpenAI www.wsj.com/tech/ai/hack... | Hackers Used Anthropic’s Claude to Break Into OpenAI www.wsj.com/tech/ai/hack... | ↗Not briefed |
arxiv arxiv-ai | 1mo | Beyond Reference-Based Evaluation: Reward Models for Meta-Evaluation of Grammatical Error Correction | Reference-based metrics for Grammatical Error Correction (GEC) such as M$^2$ and ERRANT assume that the reference set enumerates all v… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Microsoft幹部がAIによるデータスクレイピングを「人類史上最大の労働窃盗」と表現 https://gigazine.net/news/20260918-microsoft-exec-called-ai-scraping-largest-theft-labor/ | Microsoft幹部がAIによるデータスクレイピングを「人類史上最大の労働窃盗」と表現 https://gigazine.net/news/20260918-microsoft-exec-called-ai-scraping-largest-theft-labor/ | ↗Not briefed |
arxiv arxiv-ai | 1mo | KnowDemo: Knowledge-Guided Robot Demonstration Generation from Human Videos | Learning robot manipulation policies typically requires substantial demonstration data, which are costly to collect on real robots. Re… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Settling a bajillion dollar lawsuit with a transformative collaboration agreement that values the height of human skill and wisdom is one w… | Settling a bajillion dollar lawsuit with a transformative collaboration agreement that values the height of human skill and wisdom is… | ↗Not briefed |
arxiv arxiv-ai | 1mo | FOCAL-VLA: Subtask-Guided Geometry Distillation and Implicit World Modeling for Vision-Language-Action Models | Vision-language-action (VLA) models built on pretrained vision-language models have demonstrated strong performance across diverse rob… | ↗Not briefed |
bluesky bluesky-ai | 1mo | I kind of wonder how much of "Astra for Law" is based on OpenAI scraping PlainSite in violation of copyright law. The government records ar… | I kind of wonder how much of "Astra for Law" is based on OpenAI scraping PlainSite in violation of copyright law. The government recor… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Hallucination-R1: Robustness-Oriented Paraphrase Generation for Factual Consistency | Factual hallucination is commonly defined by incorrect factual outputs. We study a paraphrase-induced hallucination setting, where a m… | ↗Not briefed |
bluesky bluesky-ai | 1mo | I understand why the models didn’t come up this way, but as they reach maturity it seems quite obvious to me that every big AI company shou… | I understand why the models didn’t come up this way, but as they reach maturity it seems quite obvious to me that every big AI company… | ↗Not briefed |
hn hn-ai | 1mo | Essays of Dario Amodei are why people are worried about artificial intelligence | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | I’m gonna say something controversial but the smart (but evil) play for the times is to get a big chunk of OpenAI stock and finalize a part… | I’m gonna say something controversial but the smart (but evil) play for the times is to get a big chunk of OpenAI stock and finalize a… | ↗Not briefed |
arxiv arxiv-ai | 1mo | A Fully Differentiable Neuro-Soft-Symbolic Framework for Perceptual Task Planning | Perceptual planning tasks require two key capabilities: accurately perceiving uncertain scenes and planning valid action sequences fol… | ↗Not briefed |
bluesky bluesky-ai | 1mo | The project is no more allegedly thanks to LLM weilding script kiddies... PS5 Linux dev quits after "the only hypervisor bug left" was rep… | The project is no more allegedly thanks to LLM weilding script kiddies... PS5 Linux dev quits after "the only hypervisor bug left" wa… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Anthropic will have you believe that it will create a simulation of the entire world and then delete it one day | Anthropic will have you believe that it will create a simulation of the entire world and then delete it one day | ↗Not briefed |
bluesky bluesky-ai | 1mo | What's the over/under that one of the openAI/Anthropic/etc guys to have made their pet AI sing "daisy, daisy"? #tcm #tcmparty #stanleykubr… | What's the over/under that one of the openAI/Anthropic/etc guys to have made their pet AI sing "daisy, daisy"? #tcm #tcmparty #stanle… | ↗Not briefed |
rss techmeme | 1mo | A federal judge denies OpenAI's bid to see SpaceXAI's confidential settlement with Apple in an antitrust case, finding the materials irrelevant after a review (Hassan Ali Kanu/Politico) | Hassan Ali Kanu / Politico: A federal judge denies OpenAI's bid to see SpaceXAI's confidential settlement with Apple in an antitrust c… | ↗Not briefed |
bluesky bluesky-ai | 1mo | No, I'm sorry, but there is legitimately a difference between using an LLM as a tool to automate certain repetitive tasks while writing cod… | No, I'm sorry, but there is legitimately a difference between using an LLM as a tool to automate certain repetitive tasks while writin… | ↗Not briefed |
hn hn-ai | 1mo | AWS now supports spend limits | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | Llm으로 외국어 공부하기 좋을것 같긴 한데 내가 그렇게 써보니까 괜찮긴 하지만 계속해서 “얘의 설명을 어디까지 믿을 수 있는것인가” 하는 문제가 계속 찝찝하게 남아있어서 좀 방해가 됨… 사람의 설명도 잘못 설명하는 사람이야 많겠지만 llm의 할루시… | Llm으로 외국어 공부하기 좋을것 같긴 한데 내가 그렇게 써보니까 괜찮긴 하지만 계속해서 “얘의 설명을 어디까지 믿을 수 있는것인가” 하는 문제가 계속 찝찝하게 남아있어서 좀 방해가 됨… 사람의 설명도 잘못 설명하는 사람이야 많겠지만 llm… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Fewer Steps, Better Actions: Rethinking Flow-Matching Inference for VLA Policies | Vision-language-action (VLA) policies based on flow matching generate action chunks through repeated evaluations of an action expert.… | ↗Not briefed |
bluesky bluesky-ai | 1mo | "Hackers Used Anthropic’s Claude to Break Into OpenAI" Robert McMillan for @wsj.com | "Hackers Used Anthropic’s Claude to Break Into OpenAI" Robert McMillan for @wsj.com | ↗Not briefed |
arxiv arxiv-ai | 1mo | Ability-Residual Decoupled Modeling for Affective Cognitive Diagnosis | Cognitive diagnosis infers students' concept mastery from response logs. However, students' responses are not determined by mastery al… | ↗Not briefed |