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 | 3d | Google expands access to Gemini Tasks, which can help with calls, purchases | Gemini's Task mode can help users coordinate calls, log into websites, and even make purchases on the user's behalf. | ↗Not briefed |
hn hn-ai | 3d | Launch HN: StarSling (YC P25) – Self-improving CI for agent-volume repos | We’re Yonas and Daniel and we’re building StarSling (<a href="https://starsling.dev">https://starsling.dev</a>), f… | ↗Not briefed |
bluesky bluesky-ai | 3d | Pangram's tech might survive no matter how good the LLMs get, it's not really dependent on a weakness in the LLM. They have to do more trai… | Pangram's tech might survive no matter how good the LLMs get, it's not really dependent on a weakness in the LLM. They have to do more… | ↗Not briefed |
bluesky bluesky-ai | 3d | omg they all hate it xxcancel.com/theo/article... | omg they all hate it xxcancel.com/theo/article... | ↗Not briefed |
arxiv arxiv-ai | 3d | HardwareDx: A Foundational Graph Model System for Datacenter Machine Repairs | The rapid scale-up of AI infrastructure has made datacenter reliability a cornerstone of the cloud business economy. Since hardware fa… | ↗Not briefed |
arxiv arxiv-ai | 3d | Which Tokens, in What Order, and How Fast? A Theoretical Analysis of On-Policy Distillation Objectives | On-policy distillation (OPD) trains a student language model to match a teacher model based on its own generations. When using OPD, on… | ↗Not briefed |
rss techmeme | 3d | Google updates Gemini Privacy Hub to extend its retention period for temporary chats and chats with history turned off from 72 hours to around two months (Jagmeet Singh/TechCrunch) | Jagmeet Singh / TechCrunch: Google updates Gemini Privacy Hub to extend its retention period for temporary chats and chats with histor… | ↗Not briefed |
hn hn-ai | 3d | Show HN: Fardesk – fast iPhone app to manage your Claude sessions | — | ↗Not briefed |
arxiv arxiv-ai | 3d | QDA: Closed-Loop LLM Quantization Design with Agents | Post-training quantization is the standard way to serve large language models at low cost. However, the best quantization recipe diffe… | ↗Not briefed |
arxiv arxiv-ai | 3d | Chorus: Shared N-gram Embeddings with Layer-wise Adaptation | Language models can gain capacity at little computational cost from n-gram embedding tables: large tables of learned vectors for short… | ↗Not briefed |
arxiv arxiv-ai | 3d | Where the Tokens Go: Heat, Entropy and Error in Agentic Coding | Coding agents are billed by the token, yet software engineering says nothing about what a token actually buys. Here we introduce a the… | ↗Not briefed |
hn hn-ai | 3d | GPT-6.2 Astra scores 100% on ARC-AGI-3 | — | ↗Not briefed |
arxiv arxiv-ai | 3d | Ambient LLMs: Data-Efficient Supervised Fine-tuning via Temperature Matching | Collecting large-scale, diverse, and high-quality post-training datasets for autoregressive language models is challenging. Consequent… | ↗Not briefed |
bluesky bluesky-ai | 3d | i know all about the discourse around localized and ethical LLM too, i know thats dicier but my personal stance is master's tools dont dism… | i know all about the discourse around localized and ethical LLM too, i know thats dicier but my personal stance is master's tools dont… | ↗Not briefed |
arxiv arxiv-ai | 3d | Spool: Low-Latency Streaming Inference for Hybrid Vision-Language Models🔧 vLLM | Applications such as visual accessibility agents and robotics require answering questions about recent video content promptly, over a… | ↗Not briefed |
arxiv arxiv-ai | 3d | Training Language Models for Introspective Oversight | Language models increasingly exhibit behaviors that are difficult to detect in transcripts, such as reward hacking, deception, and syc… | ↗Not briefed |
arxiv arxiv-ai | 3d | StringQFT: A Multiuse Textbook Resource for AI-Assisted Theoretical Physics | We present StringQFT as a textbook collection for string constructions of quantum field theory and as infrastructure for AI-assisted t… | ↗Not briefed |
rss theverge-ai | 3d | Google’s Spark AI agent becomes ‘Tasks’ — and now it’s free | You can now switch from the “Chat” toggle to “Tasks.” | Image: Google Google is rebranding its Spark AI agent as "Tasks" and making… | ↗Not briefed |
hn hn-ai | 3d | I hate GPT-6.2 Astra | — | ↗Not briefed |
arxiv arxiv-ai | 3d | Neural Circuit Architectural Priors for Rat Locomotion | Animals generate behavior through continuous interactions between their neural circuits, bodies, and environment. Embodied simulations… | ↗Not briefed |
arxiv arxiv-ai | 3d | An inexact stochastic variance-reduced gradient method for constrained convex finite-sum optimization | In this paper, we study a class of constrained convex finite-sum optimization problems. In many practical applications, all or some of… | ↗Not briefed |
arxiv arxiv-ai | 3d | Do LLM Agents Still Need to Be Told How to Self-Improve? | LLM agents increasingly drive their own learning. Self-improvement methods delegate various components of model optimization to the mo… | ↗Not briefed |
arxiv arxiv-ai | 3d | What Does a Bias Head Measure? Non-Identifiability in Head-Level Bias Attribution | Methods that localise social bias in large language models to individual attention heads and remove it by ablation are validated by th… | ↗Not briefed |
arxiv arxiv-ai | 3d | Higher-Scoring Coding Models Did Not Get More Work Done | New coding models advertise themselves on score improvements of coding benchmarks. Yet no study has tested if a model that scores high… | ↗Not briefed |
hn hn-ai | 3d | Poll: Over 80% of respondents decided LLMs are fascist – Plenty use them anyway | — | ↗Not briefed |
hn hn-ai | 3d | Labcheck – one command to check if a Linux server is healthy | — | ↗Not briefed |
hn hn-ai | 3d | What End-to-End AI Means for ADAS | — | ↗Not briefed |
arxiv arxiv-ai | 3d | Multimodal Masked Diffusion Autoencoder for Modeling Impaired Driving | Human cognitive state (e.g., distraction, stress), gaze, and behavior interact with the driving environment in subtle ways that AI cou… | ↗Not briefed |
hn hn-ai | 3d | Mistral Vibe Desktop on macOS and Windows | — | ↗Not briefed |
bluesky bluesky-ai | 3d | I think this is a very easy take to have right now that misses the long-term danger. Every company is *absolutely* working to put an LLM in… | I think this is a very easy take to have right now that misses the long-term danger. Every company is *absolutely* working to put an L… | ↗Not briefed |
arxiv arxiv-ai | 3d | Distortion-Free Semantic Watermarks for Black-Box Generative Models | We present a provably distortion-free semantic watermarking algorithm that requires only black-box sampling access to a generative mod… | ↗Not briefed |
bluesky bluesky-ai | 3d | ‘OpenAI tested dangerous AI without proper oversight and misled public about it.’ Is their entire history and business model. | ‘OpenAI tested dangerous AI without proper oversight and misled public about it.’ Is their entire history and business model. | ↗Not briefed |
rss techcrunch-ai | 3d | Google to keep Gemini chats for 2 months, up from 72 hours, even with history off | Google's Gemini update extending chat retention will roll out globally over the coming weeks. | ↗Not briefed |
arxiv arxiv-ai | 3d | Reading Scene States: Uncovering Safety Failures in Vision-Language Models🔧 Elicit | Vision-language models (VLMs) can provide harmful assistance by interpreting visual events, even without an explicit request for the u… | ↗Not briefed |
arxiv arxiv-ai | 3d | Fine-Tuning Gemma 4 Edge Models for MATLAB Signal-Processing Code with SFT and QLoRA on a Laptop iGPU | Small language models that run on a laptop are attractive for writing signal-processing code offline, but they make systematic errors… | ↗Not briefed |
bluesky bluesky-ai | 3d | Bumping, as they have contacted me again. They "won't use my work" but "the auto system isn't 100%". (So, they will, in short.) Cause I lo… | Bumping, as they have contacted me again. They "won't use my work" but "the auto system isn't 100%". (So, they will, in short.) Cause… | ↗Not briefed |
bluesky bluesky-ai | 3d | I'm not even sure of that because it turns out that people only care about LLM writing in so much that it is Impure and even that is slowly… | I'm not even sure of that because it turns out that people only care about LLM writing in so much that it is Impure and even that is s… | ↗Not briefed |
bluesky bluesky-ai | 3d | Remember: Sam Altman made this play work. abcnews.com/Business/sam... | Remember: Sam Altman made this play work. abcnews.com/Business/sam... | ↗Not briefed |
hn hn-ai | 3d | Show HN: D.a.r.y.l. (1985), a film I loved as a kid, as an AI system card | I was thinking of the film and so I knocked up this blog post about AI safety with a vibe taken from the film. Possible spoiler alert… | ↗Not briefed |
bluesky bluesky-ai | 3d | I can say for a fact that neither of us were consulted on how to program a LLM/LIM agent to do our jobs. Whomever programmed it basically j… | I can say for a fact that neither of us were consulted on how to program a LLM/LIM agent to do our jobs. Whomever programmed it basica… | ↗Not briefed |
arxiv arxiv-ai | 3d | Agent Trust Infrastructure (ATI): A Federated Identity, Trust, Attestation, and Accountability Architecture for Cross-Organization Autonomous AI Agents | AI agents increasingly collaborate across organizational boundaries, invoking tools and interacting with agents operated by independen… | ↗Not briefed |
bluesky bluesky-ai | 3d | choices- "some, all, or none" out of moral cowardice around the decision to just not do the thing people are giving you the mildest possibl… | choices- "some, all, or none" out of moral cowardice around the decision to just not do the thing people are giving you the mildest po… | ↗Not briefed |
bluesky bluesky-ai | 3d | This is the first time since GPT 4.5 that a frontier release from OpenAi is not a clear step forward. I’m willing to bet that 6.2 is amazin… | This is the first time since GPT 4.5 that a frontier release from OpenAi is not a clear step forward. I’m willing to bet that 6.2 is a… | ↗Not briefed |
bluesky bluesky-ai | 3d | I expect that whomever programmed the LLM/LIM agent to try to do my job spoke to anyone who actually knew how to do that. Please note, that… | I expect that whomever programmed the LLM/LIM agent to try to do my job spoke to anyone who actually knew how to do that. Please note,… | ↗Not briefed |
bluesky bluesky-ai | 3d | Microsoft unveils AI models for coding, cybersecurity as it takes on Anthropic, OpenAI reut.rs/4zTca2k | Microsoft unveils AI models for coding, cybersecurity as it takes on Anthropic, OpenAI reut.rs/4zTca2k | ↗Not briefed |
hn hn-ai | 3d | Microsoft increases its competitiveness against OpenAI and other AI companies | — | ↗Not briefed |
hn hn-ai | 3d | Show HN: HashCortx 4.5 comes with the biggest updates and influences 30000 clone | The new version is out for windows and linux and macs , making it so easy to deploy and run agentic local models locally , and being o… | ↗Not briefed |
arxiv arxiv-ai | 3d | Learned versus Kinematic Trajectory Prediction under Reduced Observation Frequency | Trajectory prediction models for automated driving are usually trained on data with a fixed sampling frequency, while positions receiv… | ↗Not briefed |
arxiv arxiv-ai | 3d | Structure--Residual Inductive Biases for Language Model Pretraining | Next-token prediction is relied upon to learn both predictive behavior and internal representations. We ask whether representation str… | ↗Not briefed |
bluesky bluesky-ai | 3d | Can't wait for something to come out months from now where OpenAI or someone is like yea...sorry guys...our version XYZ got out again. | Can't wait for something to come out months from now where OpenAI or someone is like yea...sorry guys...our version XYZ got out again. | ↗Not briefed |