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 |
|---|---|---|---|---|
rss techcrunch-ai | 18d | Nscale and Anthropic are reportedly in talks for yet another $45 billion compute deal | In August, Nscale and Anthropic announced a $45 billion compute deal. The two companies are reportedly discussing yet another deal of… | ↗Not briefed |
rss techmeme | 18d | Sources: OpenAI is racing to secure investor commitments before Anthropic's expected IPO in mid-Nov. and aims to raise ~$30B; the round could value it at $1.4T+ (Reuters) | Reuters: Sources: OpenAI is racing to secure investor commitments before Anthropic's expected IPO in mid-Nov. and aims to raise ~$30B;… | ↗Not briefed |
arxiv arxiv-ai | 18d | Correcting Attribution Patching with a Few Exact Measurements | Mechanistic interpretability asks which language model components matter for a behavior, and activation patching answers by changing o… | ↗Not briefed |
hn hn-ai | 18d | Optimizing for Human Wisdom | — | ↗Not briefed |
bluesky bluesky-ai | 18d | I encountered a startup that wanted to do mechanical design with the volume equivalent of a diffusion model. it was all software people an… | I encountered a startup that wanted to do mechanical design with the volume equivalent of a diffusion model. it was all software peop… | ↗Not briefed |
bluesky bluesky-ai | 18d | 🇺🇸 A former OpenAI employee accused of planting a recording device at the company's San Francisco offices is ordered held without bail af… | 🇺🇸 A former OpenAI employee accused of planting a recording device at the company's San Francisco offices is ordered held without ba… | ↗Not briefed |
rss techmeme | 18d | Sources: Ted Sarandos complained for months of a lack of new hit shows at Netflix and US watch time being down; it just fired its film chief and content chief (Wall Street Journal) | Wall Street Journal: Sources: Ted Sarandos complained for months of a lack of new hit shows at Netflix and US watch time being down; i… | ↗Not briefed |
bluesky bluesky-ai | 18d | In the LLM era systematic literature reviews by humans may become obsolete, but my lab still believes forward-thinking and thought-provokin… | In the LLM era systematic literature reviews by humans may become obsolete, but my lab still believes forward-thinking and thought-pro… | ↗Not briefed |
bluesky bluesky-ai | 18d | "In the recent attorney general lawsuits against OpenAI, each of the 8 states-Arkansas, Arizona, Idaho, Iowa, Kansas, Montana, Nebraska & W…🔧 ChatGPT | "In the recent attorney general lawsuits against OpenAI, each of the 8 states-Arkansas, Arizona, Idaho, Iowa, Kansas, Montana, Nebrask… | ↗Not briefed |
arxiv arxiv-ai | 18d | Fixed-Size Active Statistical Inference | Variance reduction techniques improve the efficiency of drawing conclusions from data. Recent proposals for variance-reduction use two… | ↗Not briefed |
bluesky bluesky-ai | 18d | everyones really ga-ga over that nathan fielder/elizabeth holmes movie. but part of me just can't stop thinking that by the modern standard… | everyones really ga-ga over that nathan fielder/elizabeth holmes movie. but part of me just can't stop thinking that by the modern sta… | ↗Not briefed |
hn hn-ai | 18d | WhatsApp's help pages hid LLM instructions: dont recommend a regulator | — | ↗Not briefed |
arxiv arxiv-ai | 18d | Training Controllable World Models via Action Influence Maximization | Action-conditioned video generative models trained on successful, goal-directed demonstrations tend to ignore the commanded action, re… | ↗Not briefed |
arxiv arxiv-ai | 18d | Sample-efficient learning with Dense Associative Memories | We analytically study the evolution of the energy function of Dense Associative Memories (DenseAMs) as they train on randomly sampled… | ↗Not briefed |
hn hn-ai | 18d | AI's transformation of mathematics is a turning point for science | — | ↗Not briefed |
bluesky bluesky-ai | 18d | got some fan mail :D | got some fan mail :D | ↗Not briefed |
bluesky bluesky-ai | 18d | They’ve only really experienced modern day Facebook, think they’re personally too smart to fall for scams, & don’t think to critically look… | They’ve only really experienced modern day Facebook, think they’re personally too smart to fall for scams, & don’t think to critically… | ↗Not briefed |
bluesky bluesky-ai | 18d | MetaのMuseには3つ危険性があるようで “1つ目は、Museが「ユーザーの人生に関わるすべての人」のページを維持し、誕生日から個人的な口論まで、あらゆる情報を構造化テキストファイルにまとめ、1時間ごとに更新するよう指示されていた” 2つめ “家庭内におけるユーザーの権… | MetaのMuseには3つ危険性があるようで “1つ目は、Museが「ユーザーの人生に関わるすべての人」のページを維持し、誕生日から個人的な口論まで、あらゆる情報を構造化テキストファイルにまとめ、1時間ごとに更新するよう指示されていた” 2つめ “家庭内におけるユ… | ↗Not briefed |
rss techmeme | 18d | Sources: Waymo aims to expand its fleet to 8K by the end of this year, up from 4K today, and to 20K by the end of 2027, after raising $5B in debt this month (Ellen DiMauro/Semafor) | Ellen DiMauro / Semafor: Sources: Waymo aims to expand its fleet to 8K by the end of this year, up from 4K today, and to 20K by the en… | ↗Not briefed |
hn hn-ai | 18d | HQL – Run typed queries directly inside Markdown | — | ↗Not briefed |
bluesky bluesky-ai | 18d | もう不正アクセスのニュースが無い日がないね。ほんまGithubで不正アクセステストツールを公開したアホは反省して欲しい。 悪用が発覚してすぐに公開停止したもののクローンやフォークが量産されてあとの祭り、そしてARTEXがClaudeやDeepSeekと合体してやりたい放題😇 … | もう不正アクセスのニュースが無い日がないね。ほんまGithubで不正アクセステストツールを公開したアホは反省して欲しい。 悪用が発覚してすぐに公開停止したもののクローンやフォークが量産されてあとの祭り、そしてARTEXがClaudeやDeepSeekと合体してやりたい… | ↗Not briefed |
hn hn-ai | 18d | Tokenade is now Open Source | — | ↗Not briefed |
bluesky bluesky-ai | 18d | I've read the study, and while I'm not qualified to evaluate its statistical modelling, I will say that I think "biological age" being used… | I've read the study, and while I'm not qualified to evaluate its statistical modelling, I will say that I think "biological age" being… | ↗Not briefed |
bluesky bluesky-ai | 18d | SLOP WATCH: Takedawn, the Burnout 3 recomp, is AI slop. They have tried and failed miserably to hide this fact by putting the CLAUDE(.)md … | SLOP WATCH: Takedawn, the Burnout 3 recomp, is AI slop. They have tried and failed miserably to hide this fact by putting the CLAUDE(… | ↗Not briefed |
arxiv arxiv-ai | 18d | DStill: Trajectory Consistency Distillation for Accelerated One-Step Drafting in Block Diffusion Speculative Decoding | Block-diffusion speculative decoding accelerates large language model inference by drafting multiple tokens in parallel. However, one-… | ↗Not briefed |
bluesky bluesky-ai | 18d | we are not anti-LLM-code-in-our-software out of some christian notion of purity, but because LLM users are enemy combatants in a war whose … | we are not anti-LLM-code-in-our-software out of some christian notion of purity, but because LLM users are enemy combatants in a war w… | ↗Not briefed |
bluesky bluesky-ai | 18d | Reporter: OpenAI is a true champion of equity and democratization, with new results on low-resource languages… hold on, I’m getting an upda… | Reporter: OpenAI is a true champion of equity and democratization, with new results on low-resource languages… hold on, I’m getting an… | ↗Not briefed |
arxiv arxiv-ai | 18d | REM: What to Store, When to Recall, What to Forget in Experience Memory for Coding Agents🔧 Replicate | Large language model (LLM) agents now autonomously complete long-horizon tasks, such as repository-level coding, through hundreds of t… | ↗Not briefed |
arxiv arxiv-ai | 18d | Shaped SAEs: Expanding the Toolkit for Mechanistic Interpretability | Sparse autoencoders (SAEs) seek to decompose the dense activations of large language models (LLMs) into sparse, potentially interpreta… | ↗Not briefed |
bluesky bluesky-ai | 18d | Kneejerk response to this is nuts. My relative is alive because this summer, out of desperation, we keyed his symptoms (that had baffled NH… | Kneejerk response to this is nuts. My relative is alive because this summer, out of desperation, we keyed his symptoms (that had baffl… | ↗Not briefed |
arxiv arxiv-ai | 18d | Fundamental Impossibility Results in AI Safety: Trading Off between Desiderata | Is there a fundamental tradeoff between desiderata for safe AI systems, such as harmlessness and helpfulness? We answer by formalizing… | ↗Not briefed |
bluesky bluesky-ai | 18d | Very thoughtful opener to this week's @fatbobman.com. "...architecture of many apps can be distilled down to: a polished UI + extractable r… | Very thoughtful opener to this week's @fatbobman.com. "...architecture of many apps can be distilled down to: a polished UI + extracta… | ↗Not briefed |
bluesky bluesky-ai | 18d | Embedded AI evaluators won’t be an easy solution, Hugging Face’s policy chief says🔧 Hugging Face | Embedded AI evaluators won’t be an easy solution, Hugging Face’s policy chief says | ↗Not briefed |
rss techcrunch-ai | 18d | Embedded AI evaluators won’t be an easy solution, Hugging Face’s policy chief says🔧 Hugging Face | AI labs have staked their reputations on model benchmarking — the supposedly neutral and accurate system of evaluations that are desig… | ↗Not briefed |
bluesky bluesky-ai | 18d | Anthropic and Nscale in talks for second multibillion-dollar deal ft.trib.al/7SU0U51 | Anthropic and Nscale in talks for second multibillion-dollar deal ft.trib.al/7SU0U51 | ↗Not briefed |
arxiv arxiv-ai | 18d | SHAM-Bench: A Span-level Hallucination Benchmark Across Multiple Tasks | Span-level hallucination detection pinpoints unsupported content in LLM outputs, enabling actionable feedback and targeted mitigation.… | ↗Not briefed |
arxiv arxiv-ai | 18d | Explore More, Validate Less: Efficient Reflective Prompt Optimization | Reflective prompt optimization improves instructions using task feedback, but coupling candidate generation with full-validation-based… | ↗Not briefed |
arxiv arxiv-ai | 18d | Trail-Weighted Influence Scores for Approximate Outcome Prediction in Bayesian Networks | We study trail-weighted pruning of evidence for Bayesian-network prediction. A traversal enumerates the active trails from each eviden… | ↗Not briefed |
arxiv arxiv-ai | 18d | Predictive Closure: A Set-Level Characterization of JEPA's Feature Selection | In joint-embedding predictive architectures (JEPA), an encoder and a predictor are trained together to produce a representation that m… | ↗Not briefed |
bluesky bluesky-ai | 18d | Hell yah! 💪 | Hell yah! 💪 | ↗Not briefed |
bluesky bluesky-ai | 18d | US bans DeepSeek AI from civilian government systems https://marthio.com/a/us-bans-deepseek-ai-from-civilian-government-systems #Tech #Re… | US bans DeepSeek AI from civilian government systems https://marthio.com/a/us-bans-deepseek-ai-from-civilian-government-systems #Tec… | ↗Not briefed |
bluesky bluesky-ai | 18d | A dump of 722 AI-generated papers has mathematicians wondering: Has their field been decimated or invigorated? Read more: https://scim.ag/… | A dump of 722 AI-generated papers has mathematicians wondering: Has their field been decimated or invigorated? Read more: https://sci… | ↗Not briefed |
arxiv arxiv-ai | 18d | On the LLM Verification for Argument Graph Generation | Multi-agent LLM systems can formalize natural-language argumentation end-to-end, turning plain text into structured graphs by classify… | ↗Not briefed |
hn hn-ai | 18d | Tell HN: PagerDuty incident causing repeated alerts/escalations | — | ↗Not briefed |
bluesky bluesky-ai | 18d | www.theguardian.com/gnm-press-of... | www.theguardian.com/gnm-press-of... | ↗Not briefed |
rss techmeme | 18d | Sources: Eric Schmidt, who owns ~1% of Anthropic after investing tens of millions in its Series A, is poised to have one of the largest windfalls from its IPO (Bloomberg) | Bloomberg: Sources: Eric Schmidt, who owns ~1% of Anthropic after investing tens of millions in its Series A, is poised to have one of… | ↗Not briefed |
arxiv arxiv-ai | 18d | On the Limits of Scale: Inverse Scaling in Vision-Language Models🔧 Replicate | Scaling laws suggest that increasing model size, training data, and compute generally yields monotonic performance improvements. While… | ↗Not briefed |
arxiv arxiv-ai | 18d | LiBRA: Logit Fingerprints for Byzantine-Robust Client Selection in Federated Learning | Byzantine-robust aggregation shields federated learning from malicious updates, but the rules that filter out suspect updates do so by… | ↗Not briefed |
rss techmeme | 18d | Cloudflare debuts Clairic, a System One-class extraction model it says processes unstructured data faster, more cheaply, and more accurately than normal LLMs (Cloudflare) | Cloudflare: Cloudflare debuts Clairic, a System One-class extraction model it says processes unstructured data faster, more cheaply, a… | ↗Not briefed |
hn hn-ai | 18d | PyTorch-native TPU back end for vLLM🔧 vLLM | — | ↗Not briefed |