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 | 昭和SFの人工知能が実現したら人間要らんので、経済成長の終焉による終末を迎えるんだが、「LLMは人工知能にすらなれない」ので、現在のLLM企業の掲げる"便利""良いこと"はそもそもが嘘なので、欺瞞情報の拡大再生産によりエントロピーが増大仕切って宇宙より先に熱的に死にインターネッ… | 昭和SFの人工知能が実現したら人間要らんので、経済成長の終焉による終末を迎えるんだが、「LLMは人工知能にすらなれない」ので、現在のLLM企業の掲げる"便利""良いこと"はそもそもが嘘なので、欺瞞情報の拡大再生産によりエントロピーが増大仕切って宇宙より先に熱的に死にイ… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Characterizing Overconfident Failure in LLM-Based Code Generation | Large language models (LLMs) are increasingly used for automated code generation, but generated programs can appear syntactically plau… | ↗Not briefed |
arxiv arxiv-ai | 1mo | DuplexAgent-RSI: Recursive Harness Improvement for Full-Duplex Voice Agent Collaboration | Voice agents are converging on a collaboration pattern: a full-duplex interaction model stays on the live channel as the entry to the… | ↗Not briefed |
bluesky bluesky-ai | 1mo | そして今の生成ＡＩ（ＬＬＭ）はそういう設計思想で出来てる。 だから使い方以前の問題。 「技術も使い方次第」って論点で語るなら、ツールそのものがもう"悪い使い方の実例"なんです。 | そして今の生成ＡＩ（ＬＬＭ）はそういう設計思想で出来てる。 だから使い方以前の問題。 「技術も使い方次第」って論点で語るなら、ツールそのものがもう"悪い使い方の実例"なんです。 | ↗Not briefed |
hn hn-ai | 1mo | Researcher Gets $300 from OpenAI for Major AI Security Flaw | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | ORDO: Operation-level Round-aware Dynamic Ordering for MIP Presolve | Presolve strongly affects mixed-integer programming (MIP) performance, yet learning-based methods only optimize parameter configuratio… | ↗Not briefed |
hn hn-ai | 1mo | Clojure in the Age of Language Models | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | When Do We Need On-Policy Distillation? Distilling on Offline Student Rollouts Is Often Better | On-policy distillation (OPD) has become increasingly popular for transferring teacher capabilities to student models. In this work, we… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Ofcourse it had to be an Italian that creates malware as poetry. www.theregister.com/security/202... | Ofcourse it had to be an Italian that creates malware as poetry. www.theregister.com/security/202... | ↗Not briefed |
arxiv arxiv-ai | 1mo | Open-ended Scientific Discovery with Possibilistic Reasoning | Autonomous scientific discovery with LLMs requires generating and testing hypotheses adaptively as evidence accumulates while maintain… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Memorization and Malign Generalization in Conditional Diffusion Models with Random Features | Conditional diffusion models generate diverse, novel, and high-quality samples under prescribed conditions. However, theoretical under… | ↗Not briefed |
arxiv arxiv-ai | 1mo | REMORY: Learning Residual Memory for Context Compaction | Long-horizon agents compact their history to continue within a finite context window, but a textual summary alone may not support ever… | ↗Not briefed |
arxiv arxiv-ai | 1mo | DynaTE: Accelerating Diffusion LLMs via Dynamic Token Execution | Diffusion-based LLMs (dLLMs) have recently emerged as a promising alternative to autoregressive (AR) LLMs by enabling bidirectional pa… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Being-M0.7: A Latent World-Action Model for Humanoid Robots | Humanoid loco-manipulation requires coordinated locomotion and manipulation informed by future scene evolution and whole-body motion,… | ↗Not briefed |
bluesky bluesky-ai | 1mo | After testifying in the parliamentary inquiry that he didn’t believe AI had been used to write the notifying email "Guardian Australia und… | After testifying in the parliamentary inquiry that he didn’t believe AI had been used to write the notifying email "Guardian Australi… | ↗Not briefed |
arxiv arxiv-ai | 1mo | How to post-train on a surrogate: Envelope sampling mitigates reward hacking | Large language models (LLMs) are commonly post-trained against LLM judges and other cheap surrogates because the true reward, such as… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Tired of juggling 10 AI apps? CurvetAI raises ₹6 Cr pre-seed to build a unified multi-model workspace. 100k+ users already on board. Full … | Tired of juggling 10 AI apps? CurvetAI raises ₹6 Cr pre-seed to build a unified multi-model workspace. 100k+ users already on board.… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Our lab received an AnthropicAI AI for Science award: $25,000 in Claude API credits to accelerate multidimensional measurement of resilienc… | Our lab received an AnthropicAI AI for Science award: $25,000 in Claude API credits to accelerate multidimensional measurement of resi… | ↗Not briefed |
rss marktechpost | 1mo | Perplexity AI Releases pplx-embed-v2-late: A 0.6B Edge Model and a 9B Model Scoring 92.4% on MADQA | Perplexity's pplx-embed-v2-late comes in 2 sizes: a 0.6B model built to run on edge devices, and a 9B model for building high-quality… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Aaaaabsolutely not. | Aaaaabsolutely not. | ↗Not briefed |
arxiv arxiv-ai | 1mo | Phonological Interference in Multilingual Speech Models | Phoneme-level models transcribe or generate speech as a sequence of phonemes, the smallest sound units that distinguish words. These m… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Just what I’d expect an LLM based on your posts to say | Just what I’d expect an LLM based on your posts to say | ↗Not briefed |
bluesky bluesky-ai | 1mo | “…landing a big maths story was something that happened [rarely] – the Venn diagram of maths results that are both interesting and explaina… | “…landing a big maths story was something that happened [rarely] – the Venn diagram of maths results that are both interesting and exp… | ↗Not briefed |
bluesky bluesky-ai | 1mo | OpenAI used AI to help write email warning Australian government AI had hacked its websites www.theguardian.com/australia-ne... | OpenAI used AI to help write email warning Australian government AI had hacked its websites www.theguardian.com/australia-ne... | ↗Not briefed |
bluesky bluesky-ai | 1mo | File under inevitable “Exclusive: Revelation comes after executive told parliamentary inquiry he did not believe AI had been used to writ… | File under inevitable “Exclusive: Revelation comes after executive told parliamentary inquiry he did not believe AI had been used to… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Gated Memory: Admission-Controlled Memory Formation for Conversational AI | Personalized conversational AI relies on long-term memory systems that extract facts from user utterances and store them in persistent… | ↗Not briefed |
bluesky bluesky-ai | 1mo | and like all millennarian movements the promised day of deliverance can keep being postponed as long as you need. Ed Zitron can keep saying… | and like all millennarian movements the promised day of deliverance can keep being postponed as long as you need. Ed Zitron can keep s… | ↗Not briefed |
arxiv arxiv-ai | 1mo | TaReD: Tool-Aware Recursive Decomposition for Long-Horizon Tasks | Agents combine reasoning with tools to interact with external systems and complete real-world tasks. Early agents typically interleave… | ↗Not briefed |
rss techmeme | 1mo | SemiAnalysis: China has 24 GW of operational compute capacity and another 50 GW planned or under construction, compared to 56 GW operational in the US (Financial Times) | Financial Times: SemiAnalysis: China has 24 GW of operational compute capacity and another 50 GW planned or under construction, compar… | ↗Not briefed |
bluesky bluesky-ai | 1mo | A very good piece that, among other things, walks you through how they model the impact of automation on economic growth. aleximas.substac… | A very good piece that, among other things, walks you through how they model the impact of automation on economic growth. aleximas.su… | ↗Not briefed |
bluesky bluesky-ai | 1mo | found the original, they just ripped it off haha. this ratio is nuts if true — I thought it was closer to 50-50 inference vs. training news… | found the original, they just ripped it off haha. this ratio is nuts if true — I thought it was closer to 50-50 inference vs. training… | ↗Not briefed |
bluesky bluesky-ai | 1mo | If a piece of shit uploads some art I made to an LLM did I take an attack of slopportunity? | If a piece of shit uploads some art I made to an LLM did I take an attack of slopportunity? | ↗Not briefed |
bluesky bluesky-ai | 1mo | Hier ein paar zuletzt häufig geteilte #News: Viel zu viele Bäume gefällt: Baustopp für Rechenzentren von Google in Finnland LibreOffice u… | Hier ein paar zuletzt häufig geteilte #News: Viel zu viele Bäume gefällt: Baustopp für Rechenzentren von Google in Finnland LibreOff… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Q-Capsule: A Localized Capsule-Based Quantum Neural Architecture for Barren Plateau Mitigation | Variational quantum algorithms are often limited by barren plateaus: gradients vanish as circuit size and depth increase, making quant… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Tired of juggling 10 AI apps? CurvetAI raises ₹6 Cr pre-seed to build a unified multi-model workspace. 100k+ users already on board. Full … | Tired of juggling 10 AI apps? CurvetAI raises ₹6 Cr pre-seed to build a unified multi-model workspace. 100k+ users already on board.… | ↗Not briefed |
hn hn-ai | 1mo | Terence Tao Responds to the OpenAI Math Drop | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | Whenever I say Claude I am talking about one of the two Fire Emblem guys. NOT the “AI” model. Know this | Whenever I say Claude I am talking about one of the two Fire Emblem guys. NOT the “AI” model. Know this | ↗Not briefed |
rss techmeme | 1mo | A theoretical computer scientist, citing sources, says AI labs have quietly started probing whether their models can break important cryptographic protocols (Scott Aaronson/Shtetl-Optimized) | Scott Aaronson / Shtetl-Optimized: A theoretical computer scientist, citing sources, says AI labs have quietly started probing whether… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Residual spectral instabilities in representation learning | Learned representations can lose latent degrees of freedom successively, suggesting a cascade of transitions whose underlying stabilit… | ↗Not briefed |
hn hn-ai | 1mo | Built a ChatGPT-to-Cloudflare publisher, looking for supplychain/security input🔧 ChatGPT | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | Microsoft and Nvidia: You can run an AI model on this Surface Laptop Ultra Us: Look at that sweet magnetic USB-C port. | Microsoft and Nvidia: You can run an AI model on this Surface Laptop Ultra Us: Look at that sweet magnetic USB-C port. | ↗Not briefed |
rss techmeme | 1mo | Vitalik Buterin says "we should take the risks to cryptography from AI-accelerated math seriously", backing a "bunker mode" call for the blockchain industry (Felix Ng/Cointelegraph) | Felix Ng / Cointelegraph: Vitalik Buterin says “we should take the risks to cryptography from AI-accelerated math seriously”, backing… | ↗Not briefed |
arxiv arxiv-ai | 1mo | LLM-IDEA: Identifiability-Driven Experimental Agent for Autonomous Discovery of Mechanistic World Models | Large language model agents are being increasingly deployed as autonomous scientists, designing experiments and inferring mechanistic… | ↗Not briefed |
hn hn-ai | 1mo | Apple Music Is Getting a New Logo | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | www.theguardian.com/australia-ne.... Soon AI will be communicating with other AI in Government deciding what's best for us all. May as well… | www.theguardian.com/australia-ne.... Soon AI will be communicating with other AI in Government deciding what's best for us all. May as… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Neuro-Memory Fuzzy Inference System for Mimicking Human-like Car Following Behavior | This study presents the Neuro-Memory Fuzzy Inference System (NeMeFIS), a hierarchical machine learning architecture that asymmetricall… | ↗Not briefed |
arxiv arxiv-ai | 1mo | V-CoLA: Vision Token Compression with Linear Attention | Vision-language models (VLMs) have demonstrated impressive capabilities but suffer from substantial computational overhead, as vision… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Les Américains ne vont pas se moquer longtemps de Mistral IA, car l'entreprise française d'intelligence artificielle générative revient en … | Les Américains ne vont pas se moquer longtemps de Mistral IA, car l'entreprise française d'intelligence artificielle générative revien… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Why On-Policy Distillation Sometimes Fails: Vanishing Learning Signals | On-policy distillation (OPD) enables effective capability transfer between language models, yet the mechanisms underlying its failures… | ↗Not briefed |
rss techmeme | 1mo | German search engine Ecosia, used by government services, ditches Mistral for open-source models, including Chinese ones, citing Mistral's disappointing quality (Politico) | Politico: German search engine Ecosia, used by government services, ditches Mistral for open-source models, including Chinese ones, ci… | ↗Not briefed |