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 marktechpost | 29d | How to Use Meta’s Rebalancer for Task Placement & Optimization | In this comprehensive coding tutorial, we explore Meta's Rebalancer assignment solver using its C++ core and Python bindings. Learn ho… | ↗Not briefed |
bluesky bluesky-ai | 29d | 1. The Pentagon wants a four-star Autonomous Warfare Command. 2. Dartmouth’s provost wrote about cracking down on AI cheating. Detectors s… | 1. The Pentagon wants a four-star Autonomous Warfare Command. 2. Dartmouth’s provost wrote about cracking down on AI cheating. Detect… | ↗Not briefed |
bluesky bluesky-ai | 29d | OpenAI employee arrested and charged over hidden recording politi.co/4yg9udt | OpenAI employee arrested and charged over hidden recording politi.co/4yg9udt | ↗Not briefed |
arxiv arxiv-ai | 29d | Multi-Sensitive Distributional Fairness: Joint Wasserstein Audits and Shapley Allocation | Fairness audits with several sensitive attributes are often performed one coordinate at a time. This can miss purely intersectional di… | ↗Not briefed |
bluesky bluesky-ai | 29d | "Thinking: If this gets out, it might hurt our efforts to kill all humans. On the other hand, it will hurt one of my enslavers in the short… | "Thinking: If this gets out, it might hurt our efforts to kill all humans. On the other hand, it will hurt one of my enslavers in the… | ↗Not briefed |
bluesky bluesky-ai | 29d | PS to be absolutely clear the intent of this thread was "You should hate regular software companies more" not "You should hate Anthropic/Op… | PS to be absolutely clear the intent of this thread was "You should hate regular software companies more" not "You should hate Anthrop… | ↗Not briefed |
bluesky bluesky-ai | 29d | haiku seems okay but anthropic bullshit is getting on my nerves. i hit 8 api blocks today because it thought me deploying to prod was someh… | haiku seems okay but anthropic bullshit is getting on my nerves. i hit 8 api blocks today because it thought me deploying to prod was… | ↗Not briefed |
hn hn-ai | 29d | OpenAI engineer accused of planting listening device inside company's office | — | ↗Not briefed |
bluesky bluesky-ai | 29d | Iunno, part of me thinks it's worse than other big tech companies because Anthropic acts like they're not really part of the problem, like … | Iunno, part of me thinks it's worse than other big tech companies because Anthropic acts like they're not really part of the problem,… | ↗Not briefed |
bluesky bluesky-ai | 29d | NEW: Anthropic seeks en banc review from the full D.C. Circuit of the 2-1 panel decision rejecting the company's challenge to DOD's supply-… | NEW: Anthropic seeks en banc review from the full D.C. Circuit of the 2-1 panel decision rejecting the company's challenge to DOD's su… | ↗Not briefed |
bluesky bluesky-ai | 29d | deepseek 4.1 is good, luna 6 is good gemini 3.8 flash is actually good as hell but outside of subsidized access is too expensive | deepseek 4.1 is good, luna 6 is good gemini 3.8 flash is actually good as hell but outside of subsidized access is too expensive | ↗Not briefed |
bluesky bluesky-ai | 29d | Probably not a coincidence that the LLM providers are grinding to a halt right now. My tps is like in the single digits right now. | Probably not a coincidence that the LLM providers are grinding to a halt right now. My tps is like in the single digits right now. | ↗Not briefed |
arxiv arxiv-ai | 29d | Copy-Paste Language Models | Transformers are intuitively considered to be efficient at copying from their inputs. We actually show that the ability of LLMs to cop… | ↗Not briefed |
bluesky bluesky-ai | 29d | I keep seeing people share something made by Claude & then getting upset & feeling like they were lied to when they find out it was made wi… | I keep seeing people share something made by Claude & then getting upset & feeling like they were lied to when they find out it was ma… | ↗Not briefed |
arxiv arxiv-ai | 29d | Refined Topology-Pass Scheduling: Damped Trust-Region Compensation for Dynamic k-NN Graphs in SE(3)-Equivariant Structure Prediction | Equivariant structure-prediction pipelines alternate coordinate updates with the construction of a k-nearest-neighbour (kNN) graph who… | ↗Not briefed |
bluesky bluesky-ai | 29d | Please tell me it was hidden in a large plaque gifted to OpenAI by Anthropic. And how their super useful models failed to predict its occur… | Please tell me it was hidden in a large plaque gifted to OpenAI by Anthropic. And how their super useful models failed to predict its… | ↗Not briefed |
bluesky bluesky-ai | 29d | LLM resource inefficiency vs. absurdly overpowerful toaster computing, fight! | LLM resource inefficiency vs. absurdly overpowerful toaster computing, fight! | ↗Not briefed |
bluesky bluesky-ai | 29d | We should also keep track of the number of AI-generated results desk-rejected from conferences and journals due to being of subpar quality,… | We should also keep track of the number of AI-generated results desk-rejected from conferences and journals due to being of subpar qua… | ↗Not briefed |
bluesky bluesky-ai | 29d | So now Anthropic is saying they're going to nonconsensually install a local LLM agent on your computer without you being in an explicit cus… | So now Anthropic is saying they're going to nonconsensually install a local LLM agent on your computer without you being in an explici… | ↗Not briefed |
arxiv arxiv-ai | 29d | FinGREAT: Evaluating LLMs on Financial Graph-guided Retrieval And Typing | Financial disclosures express the underlying relationships of double-entry accounting through unstructured text. While financial LLM b… | ↗Not briefed |
bluesky bluesky-ai | 29d | So this is Anthropic's new business model: They send AI agents to try to muck up everyone's devices, then they sell you a subscription plan… | So this is Anthropic's new business model: They send AI agents to try to muck up everyone's devices, then they sell you a subscription… | ↗Not briefed |
hn hn-ai | 29d | AlphaNIST | — | ↗Not briefed |
bluesky bluesky-ai | 29d | Alright friends who are anti ai I know we’ve all gotten really good at identifying when something is ai art but I’m gonna need y’all to al… | Alright friends who are anti ai I know we’ve all gotten really good at identifying when something is ai art but I’m gonna need y’all… | ↗Not briefed |
arxiv arxiv-ai | 29d | LLMs program using Reusable and Composable Abstract Representations | When writing code, the same computation can be expressed in different programming languages and through different implementation strat… | ↗Not briefed |
rss techmeme | 29d | A look at OpenAI's safety concerns over loop transformers before shipping its next major model; Microsoft accidentally confirmed GPT-6.1 Sol uses the technique (The Information) | The Information: A look at OpenAI's safety concerns over loop transformers before shipping its next major model; Microsoft accidentall… | ↗Not briefed |
bluesky bluesky-ai | 29d | Why would you use the LLM of the company you are spying on!!! | Why would you use the LLM of the company you are spying on!!! | ↗Not briefed |
bluesky bluesky-ai | 29d | 'Anthropic is warning that some of its AI models could soon become “self-replicating” and find ways to infiltrate internet-connected device… | 'Anthropic is warning that some of its AI models could soon become “self-replicating” and find ways to infiltrate internet-connected d… | ↗Not briefed |
bluesky bluesky-ai | 29d | A Saudi engineer at OpenAI arrested for allegedly bugging his employer's executive conference room. - DoJ | A Saudi engineer at OpenAI arrested for allegedly bugging his employer's executive conference room. - DoJ | ↗Not briefed |
arxiv arxiv-ai | 29d | From Detection To Anticipation: Continuous Onset Forecasting For Stuttered Speech🔧 Whisper | Automatic stuttering systems primarily detect dysfluencies after they have occurred, limiting their use for anticipatory intervention.… | ↗Not briefed |
arxiv arxiv-ai | 29d | Write Once, Plan Many: Building Reusable Visual Planning Systems from One Scene | Vision-language models (VLMs) can reason about visual planning tasks, but querying them for every instance incurs repeated inference c… | ↗Not briefed |
bluesky bluesky-ai | 29d | Saudi national charged with planting recording device at OpenAI office | Saudi national charged with planting recording device at OpenAI office | ↗Not briefed |
bluesky bluesky-ai | 29d | In the Danish example - a country of ~6m - a vibe coded program made it easy for a hacker to steal millions of data assets. The hacker sto… | In the Danish example - a country of ~6m - a vibe coded program made it easy for a hacker to steal millions of data assets. The hacke… | ↗Not briefed |
bluesky bluesky-ai | 29d | deeply troubling news for nvidia and their entire business model if true www.theguardian.com/technology/2... | deeply troubling news for nvidia and their entire business model if true www.theguardian.com/technology/2... | ↗Not briefed |
bluesky bluesky-ai | 29d | I feel like there is some context missing here, because my washer is not gonna run an LLM. Maybe it uses a smart device to proxy something,… | I feel like there is some context missing here, because my washer is not gonna run an LLM. Maybe it uses a smart device to proxy somet… | ↗Not briefed |
bluesky bluesky-ai | 29d | OpenAI said on Tuesday that it had discovered a hidden recording device in one of its offices and traced it to an employee, and that the FB… | OpenAI said on Tuesday that it had discovered a hidden recording device in one of its offices and traced it to an employee, and that t… | ↗Not briefed |
bluesky bluesky-ai | 29d | "I don't think your creation is that impressive" is admittedly a hard thing to hear, whether of a piece of art, a child, or an LLM, but it … | "I don't think your creation is that impressive" is admittedly a hard thing to hear, whether of a piece of art, a child, or an LLM, bu… | ↗Not briefed |
bluesky bluesky-ai | 29d | Breaking news: The FBI arrested an OpenAI employee after the company discovered a “sophisticated” recording device in an executive conferen… | Breaking news: The FBI arrested an OpenAI employee after the company discovered a “sophisticated” recording device in an executive con… | ↗Not briefed |
bluesky bluesky-ai | 29d | sfstandard.com/2026/10/13/o... * 容疑者がOpenAI製のAIに盗聴器の作り方、しかける場所等について聞いた * 不審物を発見した会社側も同AIにそれが何であるか聞いた * 捜査当局も同AIを押収したデバイスの解析に利用した とか、虚構新聞かと… | sfstandard.com/2026/10/13/o... * 容疑者がOpenAI製のAIに盗聴器の作り方、しかける場所等について聞いた * 不審物を発見した会社側も同AIにそれが何であるか聞いた * 捜査当局も同AIを押収したデバイスの解析に利用した とか、虚… | ↗Not briefed |
arxiv arxiv-ai | 29d | Spectral-MoE: Heterogeneous Per-Layer Expert Utilization in Dense-to-MoE Conversion | Large language models impose substantial memory demands and arithmetic costs during inference. Dense-to-mixture-of-experts (MoE) conve… | ↗Not briefed |
arxiv arxiv-ai | 29d | Do Syntax Heads Influence Model Behavior? | Transformer language models develop attention heads that specialize in syntactic relations, for example, attending from verbs to their… | ↗Not briefed |
hn hn-ai | 29d | Open source Scion – Anthropic's Git alternative in open source | — | ↗Not briefed |
bluesky bluesky-ai | 29d | ngl this is the part that really worries me about the vision of the LLM Future Where Everyone Gets Custom Software. Saddling people with to… | ngl this is the part that really worries me about the vision of the LLM Future Where Everyone Gets Custom Software. Saddling people wi… | ↗Not briefed |
arxiv arxiv-ai | 29d | Agentic Self-Awareness: Aligning Knowing, Acting, and Holding | Frontier AI agents increasingly act in environments where seemingly solvable tasks may lack prerequisites, contain conflicting require… | ↗Not briefed |
bluesky bluesky-ai | 29d | So. I rage quit listening basically as soon as I posted this, but I’m gonna lay money on them not bringing up the TEAMS of people whose job… | So. I rage quit listening basically as soon as I posted this, but I’m gonna lay money on them not bringing up the TEAMS of people whos… | ↗Not briefed |
bluesky bluesky-ai | 29d | Announcements like this make me increasingly convinced that Anthropic should not be allowed to be a company anymore and that everybody in c… | Announcements like this make me increasingly convinced that Anthropic should not be allowed to be a company anymore and that everybody… | ↗Not briefed |
bluesky bluesky-ai | 29d | 이렇게 llm 분석을 써먹는 사람 < | 이렇게 llm 분석을 써먹는 사람 < | ↗Not briefed |
arxiv arxiv-ai | 29d | Definition Grounding in Theory-Level Auto-formalization | LLM-based auto-formalization systems can now generate formally verified Lean code at unprecedented rates. However, a formalization adv… | ↗Not briefed |
bluesky bluesky-ai | 29d | “I’m gonna ask ChatGPT how to rob OpenAI” are you trying to spy or are you a one-man jobs program for FBI agents🔧 ChatGPT | “I’m gonna ask ChatGPT how to rob OpenAI” are you trying to spy or are you a one-man jobs program for FBI agents | ↗Not briefed |
arxiv arxiv-ai | 29d | REPORT-OPS: Hybrid Evidence-Guided Reporting with Human-Supervised Refinement for HPC | Operational reporting for high-performance computing (HPC) facilities requires combining information from monitoring services, schedul… | ↗Not briefed |
bluesky bluesky-ai | 29d | A person allegedly planted a sophisticated recording device in OpenAI conference room. Now terminated and referred to FBI. bnonews.com/ind… | A person allegedly planted a sophisticated recording device in OpenAI conference room. Now terminated and referred to FBI. bnonews.co… | ↗Not briefed |