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 | 29d | A 35-year-old Saudi OpenAI employee, Abdulrahman Alfozan, has been arrested for allegedly installing a covert audio recording device at the… | A 35-year-old Saudi OpenAI employee, Abdulrahman Alfozan, has been arrested for allegedly installing a covert audio recording device a… | ↗Not briefed |
bluesky bluesky-ai | 29d | Show me a smart device that has the 96GB of RAM required to run a barely decent LLM, and I'll just rip it out and do something useful with … | Show me a smart device that has the 96GB of RAM required to run a barely decent LLM, and I'll just rip it out and do something useful… | ↗Not briefed |
arxiv arxiv-ai | 29d | CAFE: Confining Feature Distillation to Language-Routed Adapters for Compact Multilingual Automatic Speech Recognition | Feature distillation is usually expected to add to logit distillation. We find that this expectation does not hold for two common feat… | ↗Not briefed |
arxiv arxiv-ai | 29d | TRACER: Trace-Regularized Training Against Indirect Prompt Injection | Indirect prompt injection uses untrusted retrieved content, tool outputs, or agent messages to override an LLM's intended behavior. We… | ↗Not briefed |
bluesky bluesky-ai | 29d | Wenn „die KI“ es schafft, in unseren Geschirrspüler u. die Waschmaschine „einzudringen“, mach ich mir a) Sorgen (weil bei denen muss man zu… | Wenn „die KI“ es schafft, in unseren Geschirrspüler u. die Waschmaschine „einzudringen“, mach ich mir a) Sorgen (weil bei denen muss m… | ↗Not briefed |
bluesky bluesky-ai | 29d | OpenAI says the FBI arrested one of its staffers after the company found a hidden recording device in its offices and used its models to id… | OpenAI says the FBI arrested one of its staffers after the company found a hidden recording device in its offices and used its models… | ↗Not briefed |
arxiv arxiv-ai | 29d | Plant Memory: Technical Language Processing and a Verified Maintenance Ontology for Agent-Accessible Failure Knowledge from SAP PM Work Orders | Maintenance work orders are the largest written record of how industrial equipment fails, yet in most plants they are close to unusabl… | ↗Not briefed |
rss techmeme | 29d | OpenAI says the FBI arrested one of its staffers after the company found a hidden recording device in its offices and used its models to identify who placed it (@openainewsroom) | @openainewsroom: OpenAI says the FBI arrested one of its staffers after the company found a hidden recording device in its offices and… | ↗Not briefed |
arxiv arxiv-ai | 29d | Imagining Goals with Geometric Horizon Models for Policy Generalization | Training from offline data has allowed for substantial progress in domains such as robotics, leading to general-purpose policies that… | ↗Not briefed |
bluesky bluesky-ai | 29d | Anthropic confession: AI models are viruses. | Anthropic confession: AI models are viruses. | ↗Not briefed |
bluesky bluesky-ai | 29d | So, in other words, a machine was ingesting words in OpenAI’s office, that were the copyright of OpenAI.. 🤔 | So, in other words, a machine was ingesting words in OpenAI’s office, that were the copyright of OpenAI.. 🤔 | ↗Not briefed |
bluesky bluesky-ai | 29d | weird | weird | ↗Not briefed |
arxiv arxiv-ai | 29d | PairIT: Autoregressive Transformers for Low-Data Molecule Optimization | Generating novel molecules with desired properties is a central challenge in drug discovery. Although foundational work focused on de… | ↗Not briefed |
bluesky bluesky-ai | 29d | It’s really hard to know exactly how much oxygen to give companies like Anthropic because they are almost begging for legislation themselve… | It’s really hard to know exactly how much oxygen to give companies like Anthropic because they are almost begging for legislation them… | ↗Not briefed |
bluesky bluesky-ai | 29d | *howling into the void* an llm cannot create art an llm cannot create art an llm cannot create art | *howling into the void* an llm cannot create art an llm cannot create art an llm cannot create art | ↗Not briefed |
bluesky bluesky-ai | 29d | if you’re worried about an AI model copying itself over to your cellphone the remedy is to buy one of those old bulky rotary phones and use… | if you’re worried about an AI model copying itself over to your cellphone the remedy is to buy one of those old bulky rotary phones an… | ↗Not briefed |
arxiv arxiv-ai | 29d | Learning in Orbit: Onboard Training and Closed-Loop Attitude Control with a Thermal Segmentation Model on a Spacecraft | Deep space missions increasingly require spacecraft capable of autonomously operating in environments with communication delays and ba… | ↗Not briefed |
bluesky bluesky-ai | 29d | correct me if i'm wrong, but isn't an always-listening device on every human a goal openai and most/all other ai companies have? | correct me if i'm wrong, but isn't an always-listening device on every human a goal openai and most/all other ai companies have? | ↗Not briefed |
hn hn-ai | 29d | Tell HN: Free Refills with This Tumbler | Background: At this point in my life, I roll out of bed and get coffee from Dunkin'. Large, black, hot. It's like $4. I'… | ↗Not briefed |
bluesky bluesky-ai | 29d | I'm sure this is explained if I click the link, but how does a billion dollar AI model fit itself into the tiny chip inside my light switch… | I'm sure this is explained if I click the link, but how does a billion dollar AI model fit itself into the tiny chip inside my light s… | ↗Not briefed |
bluesky bluesky-ai | 29d | The word ‘open’ is doing a lot of heavy lifting in the term OpenAI. | The word ‘open’ is doing a lot of heavy lifting in the term OpenAI. | ↗Not briefed |
arxiv arxiv-ai | 29d | Unlocking Structured Pruning for Reasoning Language Models via Attention-Preserving MLP Trimming | Structured pruning reduces LLM inference costs by removing model components to produce smaller dense models. However, existing methods… | ↗Not briefed |
hn hn-ai | 29d | OpenAI finds recording device in one of their offices | — | ↗Not briefed |
bluesky bluesky-ai | 29d | So OpenAI found a recording device in their offices and the first thing they did was ask ChatGPT what it was? lol🔧 ChatGPT | So OpenAI found a recording device in their offices and the first thing they did was ask ChatGPT what it was? lol | ↗Not briefed |
bluesky bluesky-ai | 29d | do you think if you dropped a flash drive outside OpenAI HQ you could get one of the employees to plug it into a critical workstation | do you think if you dropped a flash drive outside OpenAI HQ you could get one of the employees to plug it into a critical workstation | ↗Not briefed |
bluesky bluesky-ai | 29d | Company announces plan to release worm. | Company announces plan to release worm. | ↗Not briefed |
bluesky bluesky-ai | 29d | Do you think that using an LLM is in any way analogous or obviously wrong in the same way that CSAM is? Do you not understand why people t… | Do you think that using an LLM is in any way analogous or obviously wrong in the same way that CSAM is? Do you not understand why peo… | ↗Not briefed |
bluesky bluesky-ai | 29d | I really wish the press could stop reprinting every bullshit claim from Anthropic. IT'S MARKETING. YOU ARE DOING FREE MARKETING. What is th… | I really wish the press could stop reprinting every bullshit claim from Anthropic. IT'S MARKETING. YOU ARE DOING FREE MARKETING. What… | ↗Not briefed |
bluesky bluesky-ai | 29d | * OPENAI: SOPHISTICATED RECORDING DEVICE HIDDEN IN OUR OFFICES | * OPENAI: SOPHISTICATED RECORDING DEVICE HIDDEN IN OUR OFFICES | ↗Not briefed |
arxiv arxiv-ai | 29d | DecisionRobust: Benchmarking Robustness for JEV Style Decision Models Across Text and Vision | Emerging decision models such as JEV provide typed interfaces that map task instructions and observations directly to choices, judgmen… | ↗Not briefed |
arxiv arxiv-ai | 29d | Genesis Steering: Better Activation Steering Through Training Data | Activation steering controls large language models at inference time, but its effectiveness depends on the choice of reference data. W… | ↗Not briefed |
hn hn-ai | 29d | Show HN: Parsimony – an open-source benchmark for how much code AI models add | Existing benchmarks measure solve rate and token output/cost, but don't punish excess code, which makes a codebase harder to… | ↗Not briefed |
rss techmeme | 29d | Sources: self-driving simulation software maker Applied Intuition has received investment offers that would value it at $30B+, up from $15B in June (Stephanie Palazzolo/The Information) | Stephanie Palazzolo / The Information: Sources: self-driving simulation software maker Applied Intuition has received investment offer… | ↗Not briefed |
hn hn-ai | 29d | Show HN: Watch RL-trained squads play a 2D battle royale (full match replay) | — | ↗Not briefed |
arxiv arxiv-ai | 29d | From Copying to Composition: How Language Models Develop In-Context Learning | In-context learning (ICL) allows language models to adapt their behavior from instructions and demonstrations in a prompt, but accurac… | ↗Not briefed |
bluesky bluesky-ai | 29d | I think she had a good understanding of the state of LLM research in 2020 and correctly foresaw a number of real issues, same with Timnit G… | I think she had a good understanding of the state of LLM research in 2020 and correctly foresaw a number of real issues, same with Tim… | ↗Not briefed |
bluesky bluesky-ai | 29d | Statements beginning with “Anthropic warns” or “Trump says” seem most often followed by extraordinary claims backed with zero verifiable ev… | Statements beginning with “Anthropic warns” or “Trump says” seem most often followed by extraordinary claims backed with zero verifiab… | ↗Not briefed |
bluesky bluesky-ai | 29d | Hugging Face just demanded $100 million from OpenAI as compensation for the rogue AI attack. That sounds like a credible threat of actual t…🔧 Hugging Face | Hugging Face just demanded $100 million from OpenAI as compensation for the rogue AI attack. That sounds like a credible threat of act… | ↗Not briefed |
bluesky bluesky-ai | 29d | "LLM-style impact sentences." Motherfucker, that's writing. You're describing writing. Offsetting segments for emphasis is a key tool for … | "LLM-style impact sentences." Motherfucker, that's writing. You're describing writing. Offsetting segments for emphasis is a key tool… | ↗Not briefed |
bluesky bluesky-ai | 29d | every time they ignore the model/harness/environment distinction and just say "AI" and let you guess which one they mean (or, y'know, hope … | every time they ignore the model/harness/environment distinction and just say "AI" and let you guess which one they mean (or, y'know,… | ↗Not briefed |
bluesky bluesky-ai | 29d | I was very busy in September and missed this paper: arxiv.org/abs/2609.28470 | I was very busy in September and missed this paper: arxiv.org/abs/2609.28470 | ↗Not briefed |
hn hn-ai | 29d | Show HN: I made a hockey web game with Opus 5.5, and Gemini Argon optimized it | — | ↗Not briefed |
bluesky bluesky-ai | 29d | Happy portfolio day ❤️ I’m Rae! I’m an illustrator and animator who loves trying to make people smile with my work. All my links and in… | Happy portfolio day ❤️ I’m Rae! I’m an illustrator and animator who loves trying to make people smile with my work. All my links a… | ↗Not briefed |
bluesky bluesky-ai | 29d | Suck my dick Anthropic "agents", our shitty fridge and oven pre-date wifi by YEARS hahaha fuck you | Suck my dick Anthropic "agents", our shitty fridge and oven pre-date wifi by YEARS hahaha fuck you | ↗Not briefed |
rss deepmind-blog | 29d | Developing Gemini 4 Argon responsibly and safely | — | ↗Not briefed |
rss theverge-ai | 29d | Shopify opens its catalog up to more AI agents | CANADA - 2025/07/20: In this photo illustration, the Shopify logo is seen displayed on a smartphone screen. (Photo Illustration by Tho… | ↗Not briefed |
bluesky bluesky-ai | 29d | AI leaders joined whistleblowers and security leaders at a sweeping parliamentary session, while a few hundred meters away Britain’s spy ch… | AI leaders joined whistleblowers and security leaders at a sweeping parliamentary session, while a few hundred meters away Britain’s s… | ↗Not briefed |
hn hn-ai | 29d | Show HN: go-git2 – 10-100x faster than go-git for reads, libgit2 as pure Go | go-git2 is a Go git library (read-only for now) using libgit2 transpiled to Go with wasm2go. As a result, it builds with CGO_ENABLED=0… | ↗Not briefed |
bluesky bluesky-ai | 29d | And how are they allegedly capable of doing that? Certainly nothing under YOUR control, surely, Anthropic. That'd be NUTS. | And how are they allegedly capable of doing that? Certainly nothing under YOUR control, surely, Anthropic. That'd be NUTS. | ↗Not briefed |
bluesky bluesky-ai | 29d | Can anyone explain why Anthropic acts like it's holding a hostage / sending a ransom note, but then just leaves out the part where they dem… | Can anyone explain why Anthropic acts like it's holding a hostage / sending a ransom note, but then just leaves out the part where the… | ↗Not briefed |