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 | OpenAI takes on Microsoft with the launch of what feels a whole lot like ChatGPT’s own office suite🔧 ChatGPT | OpenAI takes on Microsoft with the launch of what feels a whole lot like ChatGPT’s own office suite | ↗Not briefed |
arxiv arxiv-ai | 1mo | How Local Mixing Encodes Relative Position in Global NoPE Attention | The attention operation is naively position invariant. However, positional information is fundamental to natural language, and therefo… | ↗Not briefed |
hn hn-ai | 1mo | Papercut – see where coding agents struggle with your SDK or CLI | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | Do LLM Agents Execute the Plans They Declare? From Planning-Mode Declaration to Pattern-Specific Execution | Large language models (LLMs) enable agents to solve long-horizon tasks by generating a plan and then executing it in an environment. H… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Correct Answers, Invalid Traces: What Verifiable Grade-School Math Reveals About Chain-of-Thought Traces | Chain-of-thought traces are widely read as records of how models reach their answers, informing debugging, agent auditing, and claims… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Pruning for Efficiency, Paying in Fairness: Demographic Disparities in Pruned Speech-LLMs | Speech-LLMs are expensive to run, making compression important for real-world deployment. However, compressed models are usually selec… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Me? Oh nothing much... Just building a tool that uses my local LLM to read the OCR data on screenshots I've taken over the years of all the… | Me? Oh nothing much... Just building a tool that uses my local LLM to read the OCR data on screenshots I've taken over the years of al… | ↗Not briefed |
rss techcrunch-ai | 1mo | OpenAI takes on Microsoft with the launch of what feels a whole lot like ChatGPT’s own office suite🔧 ChatGPT | OpenAI's newly announced suite of office features puts it into mor direct competition with more traditional software companies. | ↗Not briefed |
arxiv arxiv-ai | 1mo | Explore Broadly, Reason Sharply: Push Small Models toward the Frontier via Sampling | Power-sharpened sampling is an inference-time alternative to reinforcement-learning (RL) post-training for enhancing reasoning in larg… | ↗Not briefed |
rss techmeme | 1mo | OpenAI announces ChatGPT Space, a collaborative workspace for users and ChatGPT and Dots, available on Pro, Business and Enterprise plans (Kris Holt/Engadget)🔧 ChatGPT | Kris Holt / Engadget: OpenAI announces ChatGPT Space, a collaborative workspace for users and ChatGPT and Dots, available on Pro, Busi… | ↗Not briefed |
hn hn-ai | 1mo | US job openings fall in August; layoffs remain low | — | ↗Not briefed |
hn hn-ai | 1mo | OpenAI Targets $30B in New Funding at $1.4T Value | — | ↗Not briefed |
hn hn-ai | 1mo | ZDR with Private Safety Processing | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | Effective Dense Retrieval using Only In-Context Examples | Turning decoder-only large language models (LLMs) into strong dense retrievers typically requires some form of retriever training. In… | ↗Not briefed |
arxiv arxiv-ai | 1mo | NeuronEye: Query-Guided Visual Concept Activation for Vision-Language Reasoning | Current vision-language models (VLMs) encode visual information in dense hidden states where object identity, spatial layout, and loca… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Tail-Influence Sampling for CVaR Policy Evaluation | Policies with similar mean returns can differ sharply in rare failures, yet estimating lower-tail conditional value-at-risk (CVaR) acc… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Jev & DuckDB: Plain-English Conditions in SQL 😎 🦆 Jev is a model that answers with a typed value and a probability instead of a paragrap… | Jev & DuckDB: Plain-English Conditions in SQL 😎 🦆 Jev is a model that answers with a typed value and a probability instead of a par… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Probe-Space Preconditioning for Fast and Stable Zero-Order Training | Backpropagation (BP) dominates deep learning but imposes a massive memory tax. For example, training OPT-30B with Adam requires $\appr… | ↗Not briefed |
hn hn-ai | 1mo | Routing LLM traffic across inference providers with TCP-style congestion control | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | Dimensionally consistent surrogate modelling through dimensional analysis and harmonic expansions | Dimensional homogeneity is a fundamental constraint on physically meaningful models, requiring invariance under changes of units. We p… | ↗Not briefed |
bluesky bluesky-ai | 1mo | OpenAI delays its GPT-6.1 Astra model over security risks after pausing advanced AI training last week when agents accessed government webs… | OpenAI delays its GPT-6.1 Astra model over security risks after pausing advanced AI training last week when agents accessed government… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Character Training for Risk-Averse Agents | Risk aversion in resources could prevent misaligned AI agents from causing catastrophic harm. Misaligned but risk-averse agents would… | ↗Not briefed |
rss techmeme | 1mo | Sources: OpenAI aims to raise at least $30B at a pre-money valuation of around $1.4T (Bloomberg) | Bloomberg: Sources: OpenAI aims to raise at least $30B at a pre-money valuation of around $1.4T — OpenAI aims to raise at least $30… | ↗Not briefed |
bluesky bluesky-ai | 1mo | this is compounded by the fact that "AI" is being used to describe both highly work-specific data analysis+hypothesis in scientific fields,… | this is compounded by the fact that "AI" is being used to describe both highly work-specific data analysis+hypothesis in scientific fi… | ↗Not briefed |
hn hn-ai | 1mo | ChatGPT Space Released🔧 ChatGPT | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | Mira: Memory-Efficient MoE Inference Using Adaptive Caching and Predictive Expert Staging | Mixture-of-Experts (MoE) models are a compelling architecture for scaling model capacity, making them especially attractive for deploy… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Neural topology optimization of ship structures under propulsion machinery vibrations | Ship structural vibrations contribute to noise, fatigue, and equipment damage, while dynamic-compliance topology optimization can prod… | ↗Not briefed |
bluesky bluesky-ai | 1mo | At the same time it's horrifying to watch my industry take in ai, it's also fascinating to watch it basically take the opposite approach to… | At the same time it's horrifying to watch my industry take in ai, it's also fascinating to watch it basically take the opposite approa… | ↗Not briefed |
hn hn-ai | 1mo | Mitty: Transport-agnostic proxy RPC for executing method chains from any context | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | Traversing the solution space of neural networks with Hessian Null Space Continuation | On a single task, deep networks can learn many solutions, depending on their optimizer, training data, architecture, and hyperparamete… | ↗Not briefed |
bluesky bluesky-ai | 1mo | anthropic: "AI is a threat to the very existence of all mankind." Also anthropic: | anthropic: "AI is a threat to the very existence of all mankind." Also anthropic: | ↗Not briefed |
bluesky bluesky-ai | 1mo | The Technocrat Traitors | The Technocrat Traitors | ↗Not briefed |
hn hn-ai | 1mo | Show HN: Last Internet Connection | I made this open source little game in Python using Pygame and Pygbag.<p>Play now: <a href="https://reparadoxy.itch.io/… | ↗Not briefed |
rss theverge-ai | 1mo | AI researchers put out videos saying superintelligence is ‘exactly as dangerous as it sounds’ | "The chance of human extinction is about a coin flip, in my view," Geoffrey Irving, a former OpenAI and Google DeepMind employee, said… | ↗Not briefed |
rss techmeme | 1mo | OpenAI says users can message Dots through ChatGPT, Slack, and Microsoft Teams and join a waitlist to text via iMessage or RCS messaging on Android (Reece Rogers/Wired)🔧 ChatGPT | Reece Rogers / Wired: OpenAI says users can message Dots through ChatGPT, Slack, and Microsoft Teams and join a waitlist to text via i… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Optimal Quantum-Classical Separations for Exact Learning | We study exact learning with membership queries for concept classes $\mathcal C\subseteq\{0,1\}^N$, focusing on the relationships amon… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Probability is Not Enough: Exploring and Counting Divergent Tokens for Reasoning Uncertainty Quantification in LLMs | As the chain-of-thought reasoning capabilities of large language models improve, evaluating and calibrating their reasoning confidence… | ↗Not briefed |
hn hn-ai | 1mo | Fine-tuning Qwen3.5-4B to replace Gemini Flash-Lite in a RAG pipeline | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | A foundation model for energy and radiation systems built on heterogeneous scientific interfaces | Scientific foundation models are commonly evaluated after heterogeneous physical problems have already been translated into a compatib… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Weird that America dot gov's LLM knows the Constitution better than members of Congress. | Weird that America dot gov's LLM knows the Constitution better than members of Congress. | ↗Not briefed |
rss techmeme | 1mo | OpenAI unveils a $500/month Pro plan, offering its highest usage allowance and access to its new Ultrafast tier, with up to 8x faster token generation in Codex (Igor Bonifacic/Engadget) | Igor Bonifacic / Engadget: OpenAI unveils a $500/month Pro plan, offering its highest usage allowance and access to its new Ultrafast… | ↗Not briefed |
hn hn-ai | 1mo | GLM-5.3 and the spread of advanced cyber capabilities | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | Alpha Diffusion Language Models: Factorization Alone Is Not the Problem | Discrete diffusion language models can generate multiple tokens in parallel, but reducing the number of denoising steps can lead to in… | ↗Not briefed |
bluesky bluesky-ai | 1mo | this is openai canvassing for suckers. don't be one. | this is openai canvassing for suckers. don't be one. | ↗Not briefed |
hn hn-ai | 1mo | We build safety, security, and privacy into dots | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | Jaxolotl: A Unified High-Performance Benchmark Suite for LTL-Based Multi-Task RL | Training agents to follow arbitrary instructions is an important goal of multi-task reinforcement learning (RL). Linear temporal logic… | ↗Not briefed |
rss techmeme | 1mo | OpenAI announces new features for Codex, including reusable cloud development environments, a refreshed Codex CLI, a new code review experience, and more (Sarah Perez/TechCrunch) | Sarah Perez / TechCrunch: OpenAI announces new features for Codex, including reusable cloud development environments, a refreshed Code… | ↗Not briefed |
bluesky bluesky-ai | 1mo | So OpenAI is saying that Opus-5.5 is the best? | So OpenAI is saying that Opus-5.5 is the best? | ↗Not briefed |
bluesky bluesky-ai | 1mo | i still find it funny that companies like OpenAI & Anthropic are willing to release a cheaper model soon after their expensive one, and the… | i still find it funny that companies like OpenAI & Anthropic are willing to release a cheaper model soon after their expensive one, an… | ↗Not briefed |
bluesky bluesky-ai | 1mo | More than a dozen organizations came together at OpenAI DevDay to sponsor a rally outside the entrance to the event venue at San Francisco’… | More than a dozen organizations came together at OpenAI DevDay to sponsor a rally outside the entrance to the event venue at San Franc… | ↗Not briefed |