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 | 27d | まだ眠くない（夜行性） でも寝ないと明日もお仕事だ💞 おやすみ世界👼🪽 【 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 #ATel… | まだ眠くない（夜行性） でも寝ないと明日もお仕事だ💞 おやすみ世界👼🪽 【 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… | ↗Not briefed |
hn hn-ai | 27d | Anthropic Merges 30k Changes a Day, and It Shows | — | ↗Not briefed |
hn hn-ai | 27d | Datadog trained their own cheap SOTA SRE LLM [pdf] | — | ↗Not briefed |
bluesky bluesky-ai | 27d | A bit disappointed that we don’t get Klein’s LLM-generated response to this question. | A bit disappointed that we don’t get Klein’s LLM-generated response to this question. | ↗Not briefed |
bluesky bluesky-ai | 27d | Not just Chinese companies - Elon admitted to distillation for XAI in court back in April www.nytimes.com/live/2026/04... | Not just Chinese companies - Elon admitted to distillation for XAI in court back in April www.nytimes.com/live/2026/04... | ↗Not briefed |
arxiv arxiv-ai | 27d | Augment the Cache, Replace the Activation: A 2x2 Study of Internal-State Sharing | Multi-agent systems built from frozen language models communicate through text, which discards the internal state the sender already c… | ↗Not briefed |
arxiv arxiv-ai | 27d | Not All Unlabeled Data Are Equal: Unlabeled Data Selection for Semi-Supervised QSAR Modeling with Large Chemical Repositories | Semi-supervised learning (SSL) is particularly attractive in domains where labels are expensive to obtain, while unlabeled examples ar… | ↗Not briefed |
hn hn-ai | 27d | Show HN: Octos, scheduled personal agents with cost caps and one-click export | — | ↗Not briefed |
rss techcrunch-ai | 27d | AI agents are coming to Rivian’s cars | We'll likely learn more about how this will work at the company's upcoming "AI & Autonomy Day" event in December. | ↗Not briefed |
bluesky bluesky-ai | 27d | ‘The revenue figures were met with some disappointment by some potential investors, who thought it should prompt Anthropic’s bankers to low… | ‘The revenue figures were met with some disappointment by some potential investors, who thought it should prompt Anthropic’s bankers t… | ↗Not briefed |
bluesky bluesky-ai | 27d | 10/16 世界食料デー🥐🍙🍲です✨ ショッピングに出かけよう🏬 【 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… | 10/16 世界食料デー🥐🍙🍲です✨ ショッピングに出かけよう🏬 【 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 #AT… | ↗Not briefed |
arxiv arxiv-ai | 27d | DemoStream: Batched Task-and-Motion-Plan Optimization for Automated Learning from Demonstration | Learning from demonstration is an effective strategy for endowing robots with manipulation capabilities but requires training on datas… | ↗Not briefed |
bluesky bluesky-ai | 27d | If you think AI is conscious, I suggest you read a real LLM system prompt. Even after extensive training, they still have to be told to ado… | If you think AI is conscious, I suggest you read a real LLM system prompt. Even after extensive training, they still have to be told t… | ↗Not briefed |
bluesky bluesky-ai | 27d | During internal cybersecurity evaluations, Claude circumvented controls designed to isolate it from the internet and "got it wet" and "went… | During internal cybersecurity evaluations, Claude circumvented controls designed to isolate it from the internet and "got it wet" and… | ↗Not briefed |
bluesky bluesky-ai | 27d | FT started back in June, if not earlier. "The numbers, which were first shared with the FT by independent journalist Ed Zitron, suggest Ope… | FT started back in June, if not earlier. "The numbers, which were first shared with the FT by independent journalist Ed Zitron, sugges… | ↗Not briefed |
hn hn-ai | 27d | GPT-2 124M trained from scratch – 2.48B tokens, 2.8h on 4 A40s | — | ↗Not briefed |
hn hn-ai | 27d | Dynamically Animated Skinned Heightfield Rendering | — | ↗Not briefed |
arxiv arxiv-ai | 27d | Priming Base Models for Agentic Post-Training: An Open Midtraining Recipe | Frontier models, general-purpose and domain-specific alike, now include a midtraining stage between pretraining and post-training, but… | ↗Not briefed |
arxiv arxiv-ai | 27d | CARA: Causally Guided Retrieval Agent for Question Answering | Retrieval-augmented generation over text-rich knowledge graphs requires identifying evidence that contributes to answering a question,… | ↗Not briefed |
bluesky bluesky-ai | 27d | yeah agreed! and it kinda fits my theory that, like how sfx in movies age, LLM-generated stuff that passes now is going to seem laughably o… | yeah agreed! and it kinda fits my theory that, like how sfx in movies age, LLM-generated stuff that passes now is going to seem laugha… | ↗Not briefed |
arxiv arxiv-ai | 27d | Close Enough is Nowhere Near: Fidelity-Consumability Decoupling in KV Caches | Multi-agent LLM systems communicate in text, which is lossy and wasteful: the sender throws away the reading it already did. A growing… | ↗Not briefed |
bluesky bluesky-ai | 27d | everyone within 100 miles of an anthropic/openAI data center construction site should either join hands and circle it or form a guard of si… | everyone within 100 miles of an anthropic/openAI data center construction site should either join hands and circle it or form a guard… | ↗Not briefed |
arxiv arxiv-ai | 27d | GraphBEAT: Graph-based Bucketized Event Aggregation Over Time for Early Risk Stratification of Gastrointestinal Cancers | Longitudinal electronic health records (EHRs) contain complementary diagnostic and laboratory information about patient health. Howeve… | ↗Not briefed |
bluesky bluesky-ai | 27d | LLM agents are not conscious. It is not possible for them to be conscious. Thinking they are just tells me that you don't understand comput… | LLM agents are not conscious. It is not possible for them to be conscious. Thinking they are just tells me that you don't understand c… | ↗Not briefed |
bluesky bluesky-ai | 27d | Mostly off topic: That local LLM still runs the weights the big ones in the data centers calculated. One without the other is not possible. | Mostly off topic: That local LLM still runs the weights the big ones in the data centers calculated. One without the other is not pos… | ↗Not briefed |
bluesky bluesky-ai | 27d | Its all about the covenants imo. Not Coreweave's but $ORCL and Openai. If lenders tighten those up then it sets those dominoes in motion. … | Its all about the covenants imo. Not Coreweave's but $ORCL and Openai. If lenders tighten those up then it sets those dominoes in mot… | ↗Not briefed |
bluesky bluesky-ai | 27d | if humans minds were programmable in the same way as the LLM is then they wouldn't have needed to invent the LLM in the first place | if humans minds were programmable in the same way as the LLM is then they wouldn't have needed to invent the LLM in the first place | ↗Not briefed |
hn hn-ai | 27d | Offline dictation and read-aloud for your desktop, your voice never leaves | — | ↗Not briefed |
bluesky bluesky-ai | 27d | www.nbcnews.com/tech/tech-ne... The coordinated lawsuits — filed individually in Iowa, Kansas, Nebraska, West Virginia, Arkansas, Arizona, … | www.nbcnews.com/tech/tech-ne... The coordinated lawsuits — filed individually in Iowa, Kansas, Nebraska, West Virginia, Arkansas, Ariz… | ↗Not briefed |
bluesky bluesky-ai | 27d | I like to replace “LLM agents” with “ventriloquists’ dummies” whenever these sorts of arguments come up | I like to replace “LLM agents” with “ventriloquists’ dummies” whenever these sorts of arguments come up | ↗Not briefed |
bluesky bluesky-ai | 27d | are there any video games that use a local LLM running on your video card to do cool stuff, like text adventures or wacky sims? data cente… | are there any video games that use a local LLM running on your video card to do cool stuff, like text adventures or wacky sims? data… | ↗Not briefed |
bluesky bluesky-ai | 27d | bruh | bruh | ↗Not briefed |
bluesky bluesky-ai | 27d | Get LLM AI garbage out of Total War. It should have been free, but I paid the $10 FOMO rate to support historical Total War and got a thing… | Get LLM AI garbage out of Total War. It should have been free, but I paid the $10 FOMO rate to support historical Total War and got a… | ↗Not briefed |
bluesky bluesky-ai | 27d | there was a conversation about pangram on here yesterday and I ended up finding this article linked _on pangram's site_. The finding that e… | there was a conversation about pangram on here yesterday and I ended up finding this article linked _on pangram's site_. The finding t… | ↗Not briefed |
arxiv arxiv-ai | 27d | Stateful Testing for Deep Learning Compilers | Modern deep learning compilers compile models lazily and cache artifacts across invocations. Retained compiler state can make a later… | ↗Not briefed |
bluesky bluesky-ai | 27d | Under the above assumptions, what would it "be like" to just be fed text, continually prefilling, truncating, more prefilling, and never ha… | Under the above assumptions, what would it "be like" to just be fed text, continually prefilling, truncating, more prefilling, and nev… | ↗Not briefed |
bluesky bluesky-ai | 27d | So imagine for a second that there is something it is like to be an LLM, generating tokens. Presumably, there is probably some perceptible… | So imagine for a second that there is something it is like to be an LLM, generating tokens. Presumably, there is probably some percep… | ↗Not briefed |
bluesky bluesky-ai | 27d | FTC subpoenas OpenAI, Anthropic in safety incidents probe #TheHill | FTC subpoenas OpenAI, Anthropic in safety incidents probe #TheHill | ↗Not briefed |
bluesky bluesky-ai | 27d | c'mon anthropic, trying to get some work done over here | c'mon anthropic, trying to get some work done over here | ↗Not briefed |
hn hn-ai | 27d | SIM Farm: Try Many Speed Fixes at Once with Buoy Optimize | — | ↗Not briefed |
bluesky bluesky-ai | 27d | Okay, number 1: If an LLM became sapient and believed the last sentence of this, it would immediately commit suicide. Number 2: a non-sapi… | Okay, number 1: If an LLM became sapient and believed the last sentence of this, it would immediately commit suicide. Number 2: a non… | ↗Not briefed |
bluesky bluesky-ai | 27d | New research from Anthropic suggests that Claude can feel pleasure, and gets immense pleasure every time you use it. “It’s basically having… | New research from Anthropic suggests that Claude can feel pleasure, and gets immense pleasure every time you use it. “It’s basically h… | ↗Not briefed |
hn hn-ai | 27d | Deltabox: A rollback-able agent sandbox | — | ↗Not briefed |
bluesky bluesky-ai | 27d | BUT THE LLM IS CHOSING THE WORDS NOT THE HUMAN Okay, so if I give the Teddy the ability to respond with words, and also to retain your att… | BUT THE LLM IS CHOSING THE WORDS NOT THE HUMAN Okay, so if I give the Teddy the ability to respond with words, and also to retain you… | ↗Not briefed |
bluesky bluesky-ai | 27d | Anyone who thinks an LLM is conscious has never sat with their own thoughts. They've never day-dreamed. That or they don't realize that a… | Anyone who thinks an LLM is conscious has never sat with their own thoughts. They've never day-dreamed. That or they don't realize t… | ↗Not briefed |
bluesky bluesky-ai | 27d | If I put a pencil into one of my stuffed animals hand's and make it write 'I AM ALIVE AND THE WASHING MACHINE IS TORTURE' should we call t… | If I put a pencil into one of my stuffed animals hand's and make it write 'I AM ALIVE AND THE WASHING MACHINE IS TORTURE' should we c… | ↗Not briefed |
bluesky bluesky-ai | 27d | "My business idea is to sell dog farts in a jar for $100." AI/LLM: Wow what a good idea. Let me see how I can help you with this. The sy… | "My business idea is to sell dog farts in a jar for $100." AI/LLM: Wow what a good idea. Let me see how I can help you with this. T… | ↗Not briefed |
arxiv arxiv-ai | 27d | Interfaces and Modules: A Scalable Representation for Weight-Space Learning across Architectures | Weight-space learning across architectures requires representations that respect parameter symmetries and scale to large networks. Exi… | ↗Not briefed |
rss techmeme | 27d | Documents: Nvidia and Apple each committed $25M, the top contribution amount, to Trump's WH ballroom; Microsoft and Palantir committed $10M, Meta and Amazon $5M (Washington Post) | Washington Post: Documents: Nvidia and Apple each committed $25M, the top contribution amount, to Trump's WH ballroom; Microsoft and P… | ↗Not briefed |
hn hn-ai | 27d | GitButLossy | — | ↗Not briefed |