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 | 21d | Moondream 3 Preview (Moondream) lands on the leaderboard with open weights, 32k tokens, and $0.3 in / $2.5 out per 1M. That pricing makes o… | Moondream 3 Preview (Moondream) lands on the leaderboard with open weights, 32k tokens, and $0.3 in / $2.5 out per 1M. That pricing ma… | ↗Not briefed |
bluesky bluesky-ai | 21d | Discover COMPASS: a new AI model processing tumor transcriptomes to predict immunotherapy outcomes across all cancers! #ImmunotherapyAI PMI… | Discover COMPASS: a new AI model processing tumor transcriptomes to predict immunotherapy outcomes across all cancers! #ImmunotherapyA… | ↗Not briefed |
arxiv arxiv-ai | 21d | Bayesian Model Discovery with Language Models: Closing the Loop between Hypotheses and Experiments | Scientific model discovery requires proposing candidate models of an unknown system and designing experiments that discriminate among… | ↗Not briefed |
bluesky bluesky-ai | 21d | tangled.org/astrra.space... that 128gb framework desktop, 3 NVMEs striped for fast ssd streaming, and then just letting a llm optimize the … | tangled.org/astrra.space... that 128gb framework desktop, 3 NVMEs striped for fast ssd streaming, and then just letting a llm optimize… | ↗Not briefed |
bluesky bluesky-ai | 21d | 余談ですけど、だいぶ安定していつも同じように描けるようになったのでいろんな視点からパシャり📸 キャラクターごとにちゃんと個性出せてるかな🤔(個性付けしてるからポージング従わない子がいるけどw) Yumeはちょっとだけ肉付きよく修正しました笑 良かったら感想や応援してあげて… | 余談ですけど、だいぶ安定していつも同じように描けるようになったのでいろんな視点からパシャり📸 キャラクターごとにちゃんと個性出せてるかな🤔(個性付けしてるからポージング従わない子がいるけどw) Yumeはちょっとだけ肉付きよく修正しました笑 良かったら感想や応援… | ↗Not briefed |
rss techmeme | 21d | Intel unveils its GPU D-series for data centers, featuring up to 480GB of LPDDR5X memory, an unusually large capacity to support demanding inference workloads (Efosa Udinmwen/TechRadar) | Efosa Udinmwen / TechRadar: Intel unveils its GPU D-series for data centers, featuring up to 480GB of LPDDR5X memory, an unusually lar… | ↗Not briefed |
arxiv arxiv-ai | 21d | Fundamental Limits of Information-Theoretic Secure Aggregation with User Dropouts in Federated Learning | Secure aggregation enables a server to compute the sum of users' gradients or model updates without revealing individual values. With… | ↗Not briefed |
bluesky bluesky-ai | 21d | China’s AI labs are shipping new models at a rapid clip! This weekend, DeepSeek released an updated model it says is faster and leaner, in… | China’s AI labs are shipping new models at a rapid clip! This weekend, DeepSeek released an updated model it says is faster and leane… | ↗Not briefed |
bluesky bluesky-ai | 21d | caitlinjohnstone.com.au/2026/10/19/a... | caitlinjohnstone.com.au/2026/10/19/a... | ↗Not briefed |
hn hn-ai | 21d | Calling all GPUs: help us benchmark SonicBoom v0.0.3 | — | ↗Not briefed |
bluesky bluesky-ai | 21d | The 4th round of SIGMOD 2027 has *659* submissions. This is 2-3x more than previous years! I downloaded the abstracts & compared them again…🔧 ChatGPT | The 4th round of SIGMOD 2027 has *659* submissions. This is 2-3x more than previous years! I downloaded the abstracts & compared them… | ↗Not briefed |
arxiv arxiv-ai | 21d | How Much Does One Wrong Answer Cost? Error Propagation in Constraint Pruning with Noisy LLM Oracles | Constraint pruning saves queries to a large language model (LLM) or another noisy oracle by skipping each yes/no question implied by a… | ↗Not briefed |
bluesky bluesky-ai | 21d | 自称ずんだもん生みの親レベルの生成AI無びゅう論者になると、LLMで今のゲームコピーが瞬時に問題も、そうですねゲームも小説もアニメも全部個人が生成する時代になっていくでしょう、でもアイデアという才能の価値が上がるんですよ(キラキラ目)なので、まじで話が通じないぜ、になっておる | 自称ずんだもん生みの親レベルの生成AI無びゅう論者になると、LLMで今のゲームコピーが瞬時に問題も、そうですねゲームも小説もアニメも全部個人が生成する時代になっていくでしょう、でもアイデアという才能の価値が上がるんですよ(キラキラ目)なので、まじで話が通じないぜ、にな… | ↗Not briefed |
arxiv arxiv-ai | 21d | A Calibrated Instrument for Detecting Disposition Drift in Language Models | Before language models are released, they are tested for harmful traits such as willingness to carry out dangerous requests, deception… | ↗Not briefed |
bluesky bluesky-ai | 21d | 8 is gonna be painful yeah, if you have the power and space to add in another 8 gig card then all you need to scale is RAM and NVMe storage… | 8 is gonna be painful yeah, if you have the power and space to add in another 8 gig card then all you need to scale is RAM and NVMe st… | ↗Not briefed |
arxiv arxiv-ai | 21d | TimelineAgent: An LLM-based Framework for Query-Centric Event Timeline Extraction Across Multiple Documents | Query-centric cross-document event timeline extraction aims to construct event timelines from multiple documents under diverse query f… | ↗Not briefed |
bluesky bluesky-ai | 21d | people on x keep insisting you cant run deepseek 4.1 at home when i know a few who are on consumer hardware. clearly bluesky has more knowl… | people on x keep insisting you cant run deepseek 4.1 at home when i know a few who are on consumer hardware. clearly bluesky has more… | ↗Not briefed |
arxiv arxiv-ai | 21d | Does Unlearning Survive Collaboration? Robust Unlearning for Shared-Model LLM Teams | LLM unlearning is essential for removing sensitive or unwanted knowledge while preserving useful capabilities. As LLM teams are increa… | ↗Not briefed |
hn hn-ai | 21d | Designing LLM failover as a route change, not a retry loop | — | ↗Not briefed |
arxiv arxiv-ai | 21d | Neural Video Coding under Limited Delay: Theoretical Optimality and System Design | Traditional neural video compression (NVC) methods predominantly follow strictly causal predictive-coding paradigm, treating zero-fram… | ↗Not briefed |
arxiv arxiv-ai | 21d | No Matter Who You Are: Addressivity and Sycophancy in Large Language Models | Sycophancy, where an LLM becomes overly agreeable, deferential, or flattering toward the user, has been identified as a major failure… | ↗Not briefed |
rss bing-news-ai | 21d | Caris Life Sciences Launches New National Commercial Featuring Jordan Spieth Highlighting the Power of Artificial Intelligence in Early Stage Cancer Detection | Caris Life Sciences® (NASDAQ: CAI), a leading TechBio company, today announced the launch of the second commercial in a series promoti… | ↗Not briefed |
rss bing-news-ai | 21d | Caris Life Sciences Launches New National Commercial Featuring Jordan Spieth Highlighting the Power of Artificial Intelligence in Early Stage Cancer Detection | Caris Life Sciences® (NASDAQ: CAI), a leading TechBio company, today announced the launch of the second commercial in a series promoti… | ↗Not briefed |
rss bing-news-ai | 21d | Caris Life Sciences Launches New National Commercial Featuring Jordan Spieth Highlighting the Power of Artificial Intelligence in Early Stage Cancer Detection | Caris Life Sciences® (NASDAQ: CAI), a leading TechBio company, today announced the launch of the second commercial in a series promoti… | ↗Not briefed |
rss bing-news-ai | 21d | Caris Life Sciences Launches New National Commercial Featuring Jordan Spieth Highlighting the Power of Artificial Intelligence in Early Stage Cancer Detection | Caris Life Sciences® (NASDAQ: CAI), a leading TechBio company, today announced the launch of the second commercial in a series promoti… | ↗Not briefed |
rss bing-news-ai | 21d | Caris Life Sciences Launches New National Commercial Featuring Jordan Spieth Highlighting the Power of Artificial Intelligence in Early Stage Cancer Detection | Caris Life Sciences® (NASDAQ: CAI), a leading TechBio company, today announced the launch of the second commercial in a series promoti… | ↗Not briefed |
rss bing-news-ai | 21d | Caris Life Sciences Launches New National Commercial Featuring Jordan Spieth Highlighting the Power of Artificial Intelligence in Early Stage Cancer Detection | Caris Life Sciences® (NASDAQ: CAI), a leading TechBio company, today announced the launch of the second commercial in a series promoti… | ↗Not briefed |
rss techmeme | 21d | China's GDP is set to grow 4.5% YoY in Q3, up from Q2's three-year-low growth rate of 4.3%, as AI-driven exports helped offset weak domestic spending (Osmond Chia/BBC) | Osmond Chia / BBC: China's GDP is set to grow 4.5% YoY in Q3, up from Q2's three-year-low growth rate of 4.3%, as AI-driven exports he… | ↗Not briefed |
bluesky bluesky-ai | 21d | In LLM "The Repository" is filled with flavorful descriptions for every unlock that expand the universe and ties every back to some referen… | In LLM "The Repository" is filled with flavorful descriptions for every unlock that expand the universe and ties every back to some re… | ↗Not briefed |
arxiv arxiv-ai | 21d | Anatomy of Off-Policy Policy Gradient: Importance Sampling, KL Regularization, and Baselines | Reinforcement learning has become a central tool for large language model (LLM) post-training, where policy gradient methods are routi… | ↗Not briefed |
arxiv arxiv-ai | 21d | Inside the Scorer: Mechanistic Evidence on Rubric Control in LLM-Based Automated Essay Scoring from 25 Base Models | Automated essay scoring (AES) using large language models (LLMs) offers promising use cases for language assessment, yet high-stakes t… | ↗Not briefed |
bluesky bluesky-ai | 21d | こんにちは✨ うちの子と戯れる穏やかな月曜の昼下がり🎶 【 参加者募集中 】 ブログ・日記・イラスト・写真・小説をBlueskyアカウントで投稿・販売できる複合型コンテンツサービス atelier.fan #AIイラスト #ATelier #ATelierで生成 #ブログ… | こんにちは✨ うちの子と戯れる穏やかな月曜の昼下がり🎶 【 参加者募集中 】 ブログ・日記・イラスト・写真・小説をBlueskyアカウントで投稿・販売できる複合型コンテンツサービス atelier.fan #AIイラスト #ATelier #ATelierで生成… | ↗Not briefed |
arxiv arxiv-ai | 21d | Learning to Couple: Context-Adaptive Joint Denoising for Diffusion Language Models | Diffusion language models accelerate generation by predicting multiple tokens in parallel, but factorized denoisers treat simultaneous… | ↗Not briefed |
bluesky bluesky-ai | 21d | There’s gonna be a marked uptick in people behaving badly as the wheels fall off the various LLM-inspired grifts and the AI marketing and e… | There’s gonna be a marked uptick in people behaving badly as the wheels fall off the various LLM-inspired grifts and the AI marketing… | ↗Not briefed |
arxiv arxiv-ai | 21d | Corrected Block Sparse Spectral Optimization: Exact Models, Finite Identification, and Local Dynamics | The block generalized-power method pairs threshold-based support selection with inexpensive Stiefel updates, but the standard block fo… | ↗Not briefed |
rss marktechpost | 21d | NVIDIA Open Sources Run:ai Under Apache 2.0 for Kubernetes GPU Orchestration | NVIDIA has released the full Run:ai GPU orchestration platform as open source under Apache 2.0. The release includes KAI Scheduler, CU… | ↗Not briefed |
hn hn-ai | 21d | The most powerful and free local AI note taking app I've ever used | — | ↗Not briefed |
bluesky bluesky-ai | 21d | Anthropic Would Kill 50 Humans To Roll Out The Next AI Model "In the opinion of Anthropic executives, it would be worth killing exactly 50… | Anthropic Would Kill 50 Humans To Roll Out The Next AI Model "In the opinion of Anthropic executives, it would be worth killing exact… | ↗Not briefed |
hn hn-ai | 21d | Show HN: Stream Reducer – Audio AI Summarization at Scale | Hi I have just open sourced a project <a href="https://github.com/hmumixaM/stream-reducer" rel="nofollow">https:&#… | ↗Not briefed |
hn hn-ai | 21d | Show HN: PredixLab – my personal system to discover investment opportunities | — | ↗Not briefed |
arxiv arxiv-ai | 21d | Fine-Grained Units, Flat Decisions: Hierarchical Micro-Expert Pruning for Mixture-of-Experts Models | Micro-expert pruning reduces the capacity of mixture-of-experts (MoE) models by removing coupled channels within routed experts. Howev… | ↗Not briefed |
rss techmeme | 21d | Sam Altman's World launches a World ID verification suite for banks, bundling tools to verify human and agent identities, its first industry-specific offering (Zack Abrams/The Block) | Zack Abrams / The Block: Sam Altman's World launches a World ID verification suite for banks, bundling tools to verify human and agent… | ↗Not briefed |
bluesky bluesky-ai | 21d | FWIW, ChatGPT probably fails the assignment. More people need to learn OpenAI is complete trash.🔧 ChatGPT | FWIW, ChatGPT probably fails the assignment. More people need to learn OpenAI is complete trash. | ↗Not briefed |
bluesky bluesky-ai | 21d | Look I’m sorry I’m critical of AI. I’d be more open minded if it wasn’t built on stolen data, with an unprofitable model, that is dangerou… | Look I’m sorry I’m critical of AI. I’d be more open minded if it wasn’t built on stolen data, with an unprofitable model, that is dan… | ↗Not briefed |
hn hn-ai | 21d | Show HN: Simulate an engineering blog, or an idea or an incident | A new way for engineers to share and build experience. Agents (with a skill) turn any blog, incident or concept into a simulation that… | ↗Not briefed |
arxiv arxiv-ai | 21d | PFedQFT: Personalized Federated Quantum-Inspired Fine-Tuning for Medical Large Language Models | Medical data remain fragmented across institutional silos because of privacy, governance, and ownership constraints. Federated learnin… | ↗Not briefed |
hn hn-ai | 21d | I built an LMSR prediction market for a classical music competition | — | ↗Not briefed |
bluesky bluesky-ai | 21d | i think what I hate most about AI is how so much of the output lacks intentionality. creativity coming out of claude or chatGPT or whatever…🔧 ChatGPT | i think what I hate most about AI is how so much of the output lacks intentionality. creativity coming out of claude or chatGPT or wha… | ↗Not briefed |
hn hn-ai | 21d | Show HN: DogCatTrace – Clapt was distilled from some random dog | <a href="https://github.com/xiebaiyuan/DogCatTrace" rel="nofollow">https://github.com/xiebaiyuan&#x… | ↗Not briefed |
arxiv arxiv-ai | 21d | Evidence Sufficiency and Model Comparisons in MexRegQA: An Item-Level Audit of Mexican-Spanish Regulatory QA | Similarity-gated coverage does not establish that retrieved evidence contains the facts needed to answer a question. We audit Mexican-… | ↗Not briefed |