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 | LLM chatbots on published papers without explicit opt-in from authors is so incredibly disrespectful and completely undermines the idea of … | LLM chatbots on published papers without explicit opt-in from authors is so incredibly disrespectful and completely undermines the ide… | ↗Not briefed |
arxiv arxiv-ai | 1mo | LoopCD: Loop-wise Contrastive Decoding for Improving Reasoning in Looped Language Models | Looped Language Models (LoopLMs) perform "latent reasoning" by recursively refining internal latent representations with shared weight… | ↗Not briefed |
arxiv arxiv-ai | 1mo | When Residualization Helps an Audit: Format Effects, Slice Gains, and Their Limits | Evaluation scores used around LLM systems -- including reward models, rerankers, and LLM judges -- can track surface form instead of t… | ↗Not briefed |
bluesky bluesky-ai | 1mo | I mean, the article you just shared is from 1981 and it literally says "Machines with preprogrammed responses don't have intelligence, unle… | I mean, the article you just shared is from 1981 and it literally says "Machines with preprogrammed responses don't have intelligence,… | ↗Not briefed |
bluesky bluesky-ai | 1mo | AI giants sued for 'safety' cartel ⚖️ Anthropic, OpenAI, SpaceXAI, and Google are facing an antitrust lawsuit, alleging a coordinated agre… | AI giants sued for 'safety' cartel ⚖️ Anthropic, OpenAI, SpaceXAI, and Google are facing an antitrust lawsuit, alleging a coordinated… | ↗Not briefed |
rss openai-news | 1mo | Expanding OpenAI Academy with new learning paths | Explore new OpenAI Academy learning paths for employees, developers, leaders, educators, and students to build and demonstrate practic… | ↗Not briefed |
arxiv arxiv-ai | 1mo | LIMIT: Less Is More for Instruction Tuning in Text-to-SQL | Large language models have achieved remarkable progress on Text-to-SQL through reasoning-enhanced fine-tuning, yet existing approaches… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Swaplse (2026): Uncensored 4K NSFW AI generator running on dedicated RTX 5090s. Text-to-image, image-to-video & AI chat — under 2.5s gens,… | Swaplse (2026): Uncensored 4K NSFW AI generator running on dedicated RTX 5090s. Text-to-image, image-to-video & AI chat — under 2.5s… | ↗Not briefed |
hn hn-ai | 1mo | Ask HN: My recent show HN post got flagged need guidance | hi recently i shared my work which is an opensource platform i made and yesterday i shared its details and its repo link on hackernews… | ↗Not briefed |
bluesky bluesky-ai | 1mo | unfortunately deepseek models are legally blind | unfortunately deepseek models are legally blind | ↗Not briefed |
arxiv arxiv-ai | 1mo | Efficient LLM Distillation for Bangladesh Legal Context: A Smartphone-Compatible Retrieval-Augmented Generation Model | Legal information in Bangladesh is inaccessible to most citizens. Statutory text is English-only, trained lawyers are concentrated in… | ↗Not briefed |
bluesky bluesky-ai | 1mo | pretty much blocking LLM boosters by rote now. I take a pragmatic attitude to using the tools, but enough of these guys are harmful wingnut… | pretty much blocking LLM boosters by rote now. I take a pragmatic attitude to using the tools, but enough of these guys are harmful wi… | ↗Not briefed |
arxiv arxiv-ai | 1mo | CREDO: Variance-Guided Rubric Evolution for Replay-Corrected Credit Assignment | Long-horizon language agents receive sparse terminal feedback, while intermediate rubrics provide structured but potentially misspecif… | ↗Not briefed |
bluesky bluesky-ai | 1mo | This is a fun coda to an FT article and I'm sure partly tongue in cheek. But it's fun because of a super counterintuitive thing: an apocaly… | This is a fun coda to an FT article and I'm sure partly tongue in cheek. But it's fun because of a super counterintuitive thing: an ap… | ↗Not briefed |
rss marktechpost | 1mo | StepFun Launches Step 5 Preview: A 600B-Total, 27B-Active MoE Model With 1M Context for Long-Horizon Agentic Work | StepFun has released Step 5 Preview, a sparse Mixture-of-Experts model with 600B total parameters and 27B active per token. It support… | ↗Not briefed |
bluesky bluesky-ai | 1mo | www.reuters.com/world/anthro... darioamodei.com/post/we-must... www.who.int/publications... garymarcus.substack.com/p/top-three-... | www.reuters.com/world/anthro... darioamodei.com/post/we-must... www.who.int/publications... garymarcus.substack.com/p/top-three-… | ↗Not briefed |
bluesky bluesky-ai | 1mo | "Wat pas echt griezelig zou zijn, is een geautomatiseerd laboratorium met door AI aangestuurde robots die zelfstandig proefjes doen. Dus d… | "Wat pas echt griezelig zou zijn, is een geautomatiseerd laboratorium met door AI aangestuurde robots die zelfstandig proefjes doen.… | ↗Not briefed |
arxiv arxiv-ai | 1mo | APEXA: Execution-Integrity Enforcement for Multi-Agent LLM Automation of Synchrotron Data Reduction | Synchrotron data reduction, detector calibration followed by azimuthal integration of terabyte-scale diffraction series, is a multi-st… | ↗Not briefed |
bluesky bluesky-ai | 1mo | And in a separate piece on valuation models the FT notes ‘Clearly, if AI really does make humans extinct, every investment goes to zero’ ww… | And in a separate piece on valuation models the FT notes ‘Clearly, if AI really does make humans extinct, every investment goes to zer… | ↗Not briefed |
bluesky bluesky-ai | 1mo | some people on here do a version of this where it’s a big crisis that the left is missing out on using claude code to build agentic AI work… | some people on here do a version of this where it’s a big crisis that the left is missing out on using claude code to build agentic AI… | ↗Not briefed |
hn hn-ai | 1mo | Software Optimization Resources | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | also it's so obvious now all these loreslop videos on youtube are scripted by an LLM like it's not even hard to catch. Thanks for ruining e… | also it's so obvious now all these loreslop videos on youtube are scripted by an LLM like it's not even hard to catch. Thanks for ruin… | ↗Not briefed |
bluesky bluesky-ai | 1mo | I love how if I use the bsky app and press on a dm, it teleports me to messages from days ago instead of recent messages. And somehow they'… | I love how if I use the bsky app and press on a dm, it teleports me to messages from days ago instead of recent messages. And somehow… | ↗Not briefed |
hn hn-ai | 1mo | Slower than an SD card under heavy load: iPhone 18 Pro Max's QLC storage tested | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | MCP-GRANITE Benchmark: GRANularity Interface TEsting for MCP-Based LLM Agents | As LLM agents increasingly interact with external tools through standardized protocols such as MCP, tool-interface design becomes a cr… | ↗Not briefed |
bluesky bluesky-ai | 1mo | This shit drives me nuts when people are like “don’t you want quote nontechnical people unquote to be able to automate stuff” yes! I do and… | This shit drives me nuts when people are like “don’t you want quote nontechnical people unquote to be able to automate stuff” yes! I d… | ↗Not briefed |
arxiv arxiv-ai | 1mo | TAC-Time: Texts as Channels For Multimodal Time Series Forecasting🔧 Modal | Most existing time series forecasting methods rely solely on numerical observations, overlooking rich contextual information from auxi… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Graded-Relevance Composed Multimodal Retrieval for E-commerce Visual Search at Scale | Visual search on large e-commerce catalogs must serve both "similarity" queries that ask for items resembling an uploaded image and "m… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Buenos días almas diversas! Por si os apetece echarle un ojo hoy escribo en @Conversation_E sobre el miedo que debe (o no) tener ma comunid… | Buenos días almas diversas! Por si os apetece echarle un ojo hoy escribo en @Conversation_E sobre el miedo que debe (o no) tener ma co… | ↗Not briefed |
arxiv arxiv-ai | 1mo | STAR: Scene- and Task-Aware 4D Radar Preprocessing Towards End-to-End Cognitive Radar | Four-dimensional (4D) Radar has emerged as a key sensor for environmental perception, providing range, azimuth, elevation, and Doppler… | ↗Not briefed |
rss techmeme | 1mo | China collected $32.3B of stamp duty on stock sales between January and August, up 80%+ YoY, as average daily trading value rose 72%, amid an AI trading frenzy (Zhang Shidong/South China Morning Post) | Zhang Shidong / South China Morning Post: China collected $32.3B of stamp duty on stock sales between January and August, up 80%+ YoY,… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Acceptance-Aware Draft Model Training for Speculative Decoding | Speculative decoding accelerates large language model (LLM) inference by using a lightweight draft model to generate multiple candidat… | ↗Not briefed |
bluesky bluesky-ai | 1mo | FT asking the big questions here - could Anthropic maintain its growth trajectory if AI destroys humanity? www.ft.com/content/96d0... | FT asking the big questions here - could Anthropic maintain its growth trajectory if AI destroys humanity? www.ft.com/content/96d0... | ↗Not briefed |
arxiv arxiv-ai | 1mo | Mind or Message? Auditing Theory of Mind in Multi-Agent Social Simulation | Language model agents are increasingly used to simulate social interaction, and the resulting transcripts read as though the agents un… | ↗Not briefed |
hn hn-ai | 1mo | Find-jevable-code – audit a repo for Jev-replaceable decisions | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | Luck Is Not Skill: When Do Paired Rollouts Help Group-Relative RL of LLM Agents? | Group-relative reinforcement learning compares rollouts of the same prompt, but independent environment noise can obscure these compar… | ↗Not briefed |
rss techmeme | 1mo | A researcher used GPT-6 Astra to decipher a WWI German radio transmission from 1918, one of the 50 famous unsolved ciphers listed on a German science blog (prinz) | prinz: A researcher used GPT-6 Astra to decipher a WWI German radio transmission from 1918, one of the 50 famous unsolved ciphers list… | ↗Not briefed |
bluesky bluesky-ai | 1mo | SoftBank Group is seeking the equivalent of more than $11 billion in what would be one of the biggest junk bond deals ever as the conglomer… | SoftBank Group is seeking the equivalent of more than $11 billion in what would be one of the biggest junk bond deals ever as the cong… | ↗Not briefed |
hn hn-ai | 1mo | OpenAI Degrading Their Models, Also for Paid Users | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | Kokotajlo wrote the forecast about the trajectory of AI that has proven to be the most accurate so far. He was also willing to forgo $2 mi… | Kokotajlo wrote the forecast about the trajectory of AI that has proven to be the most accurate so far. He was also willing to forgo… | ↗Not briefed |
bluesky bluesky-ai | 1mo | kan maar beter gewoon de hele nieuws sectie in de adblocker stoppen want ze zijn tot het naieve toe pro-llm | kan maar beter gewoon de hele nieuws sectie in de adblocker stoppen want ze zijn tot het naieve toe pro-llm | ↗Not briefed |
arxiv arxiv-ai | 1mo | CLOOPD: Closing the Learner Loop in On-Policy Distillation | On-policy distillation (OPD) pays twice for each fresh batch: the student generates trajectories and a stronger teacher scores them. E… | ↗Not briefed |
arxiv arxiv-ai | 1mo | P2Flow: Phoneme-aware Progressive Flow Matching for Extreme Speech Super-Resolution | Generative models have recently demonstrated considerable promise in speech super-resolution (SSR). Nevertheless, the majority of exis… | ↗Not briefed |
hn hn-ai | 1mo | What Sun Got Wrong | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | Data Agents: Agentic Data Systems | Traditional data systems face profound limitations in the AI era, relying on human-crafted pipelines, lacking semantic understanding o… | ↗Not briefed |
bluesky bluesky-ai | 1mo | My #newsletter #TechLetters ☕️. OpenAI security was a text string. Claude hacked OpenAI. Russia’s election system leaked 111M voters. AWS l… | My #newsletter #TechLetters ☕️. OpenAI security was a text string. Claude hacked OpenAI. Russia’s election system leaked 111M voters.… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Self-Healing Harness for Runtime Oversight of Agent Self-Modification | LLM agents can change their own future behavior, raising a basic control question of which self-generated changes should be allowed to… | ↗Not briefed |
bluesky bluesky-ai | 1mo | as soon as they started giving LLM models hot girl names like "fable" and "astra" the models got a lot better. really makes you think | as soon as they started giving LLM models hot girl names like "fable" and "astra" the models got a lot better. really makes you think | ↗Not briefed |
bluesky bluesky-ai | 1mo | Kokotajlo wrote the forecast about the trajectory of AI that has proven to be the most accurate so far. He was also willingly to forgo $2 … | Kokotajlo wrote the forecast about the trajectory of AI that has proven to be the most accurate so far. He was also willingly to forg… | ↗Not briefed |
arxiv arxiv-ai | 1mo | OSCAR: Order-aware Scoring and Calibration for AI Rankings | Judge-specific sensitivity is useful for aggregating pairwise LLM evaluations, but its interpretation depends on which systematic pres… | ↗Not briefed |