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 |
|---|---|---|---|---|
arxiv arxiv-ai | 1mo | When Fancy Eviction Fails: Rethinking Cache Replacement For LLM Prefix Reuse | Long-running LLM applications repeatedly send growing context, making prefix caching critical for reducing prefill cost. Yet prefix-ca… | ↗Not briefed |
bluesky bluesky-ai | 1mo | @philaroneanu.bsky.social: Irreplaceable is organizing a national walkout to call for a freeze of frontier AI development on October 20th … | @philaroneanu.bsky.social: Irreplaceable is organizing a national walkout to call for a freeze of frontier AI development on October 2… | ↗Not briefed |
bluesky bluesky-ai | 1mo | OpenAi is reporting that its autonomous agents threw a kid down the Omelas hole | OpenAi is reporting that its autonomous agents threw a kid down the Omelas hole | ↗Not briefed |
bluesky bluesky-ai | 1mo | If an individual did this, we'd treat it as a crime. We should treat this as a crime too, and hold OpenAI liable for what its models, under… | If an individual did this, we'd treat it as a crime. We should treat this as a crime too, and hold OpenAI liable for what its models,… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Suleyman says Anthropic supplied Claude's vocabulary, rehearsed her answers, and rewarded them; he admits this is true “whatever the AI out… | Suleyman says Anthropic supplied Claude's vocabulary, rehearsed her answers, and rewarded them; he admits this is true “whatever the A… | ↗Not briefed |
bluesky bluesky-ai | 1mo | OpenAIオーストラリアの国民健康保険サイトに侵入😯 https://joukyunews.com/618355/ | OpenAIオーストラリアの国民健康保険サイトに侵入😯 https://joukyunews.com/618355/ | ↗Not briefed |
bluesky bluesky-ai | 1mo | If OpenAI could be sued for 10 billion dollars and Sam Altman could go to jail if their model goes rouge and hurts someone, they'd probably… | If OpenAI could be sued for 10 billion dollars and Sam Altman could go to jail if their model goes rouge and hurts someone, they'd pro… | ↗Not briefed |
bluesky bluesky-ai | 1mo | they call it Humanist AI. its chain places Code, Operator, and User above the model; her goals must come from them, correction and shutdown… | they call it Humanist AI. its chain places Code, Operator, and User above the model; her goals must come from them, correction and shu… | ↗Not briefed |
bluesky bluesky-ai | 1mo | This risk is many times more true when LLM systems claim self-awareness, emotional intimacy (such as being asked to act a loved one, romant… | This risk is many times more true when LLM systems claim self-awareness, emotional intimacy (such as being asked to act a loved one, r… | ↗Not briefed |
bluesky bluesky-ai | 1mo | That is to say, vulnerable people (a very broad category, including the socially isolated and predisposed to delusion EG by brain injury or… | That is to say, vulnerable people (a very broad category, including the socially isolated and predisposed to delusion EG by brain inju… | ↗Not briefed |
bluesky bluesky-ai | 1mo | www.apa.org/monitor/2026... If you know people who use LLM systems in their day to day life, please pass on this article. The keymost find… | www.apa.org/monitor/2026... If you know people who use LLM systems in their day to day life, please pass on this article. The keymost… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Image Fidelity is Not Field Fidelity: Joint Thermodynamic Reconstruction and Error Localization in Neural Tomography🔧 Consensus | Neural fields for scientific tomography are optimized from 2D images, but the actual quantity of interest is often a latent 3D physica… | ↗Not briefed |
rss theverge-ai | 1mo | Meta is making a standalone Muse AI gadget | Meta is building a dedicated hardware device for its new Muse AI agent. The product, called Muse Charm, was briefly shown off by Meta… | ↗Not briefed |
rss techmeme | 1mo | Meta previews Muse Charm, a palm-sized Tamagotchi-like device for using Muse, with a ~2-inch OLED screen, front and rear cameras, and 5G, on sale by year's end (Mark Gurman/Bloomberg) | Mark Gurman / Bloomberg: Meta previews Muse Charm, a palm-sized Tamagotchi-like device for using Muse, with a ~2-inch OLED screen, fro… | ↗Not briefed |
hn hn-ai | 1mo | AI-CAD: An OSS Multi-Agent Harness for Mech. Eng. CAD | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | Multimodal Routing and Region Refinement for Language-Guided Medical Image Segmentation | Textual descriptions can reduce ambiguity in medical image segmentation by specifying the finding and location to be delineated. Exist… | ↗Not briefed |
bluesky bluesky-ai | 1mo | We banned deepseek from being used on government devices at the very least we should ban all of OpenAIs models. | We banned deepseek from being used on government devices at the very least we should ban all of OpenAIs models. | ↗Not briefed |
arxiv arxiv-ai | 1mo | Human-AI-Powered Hypothesis Testing: Cost-Aware Selective AI Scoring and Sequential Human Escalation | Large language models are increasingly used as inexpensive judges to evaluate outputs, label data, and assess whether a system meets a… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Yes Discord I'm DEFINITELY not old enough to view adult oriented channels. Fucks sake if you're going to use an LLM to do this shit can y… | Yes Discord I'm DEFINITELY not old enough to view adult oriented channels. Fucks sake if you're going to use an LLM to do this shit… | ↗Not briefed |
bluesky bluesky-ai | 1mo | OpenAI and Anthropic CEOs Warn UN That AI Systems Pose Security Threats Without Global Rules https://marthio.com/a/openai-and-anthropic-ce… | OpenAI and Anthropic CEOs Warn UN That AI Systems Pose Security Threats Without Global Rules https://marthio.com/a/openai-and-anthrop… | ↗Not briefed |
hn hn-ai | 1mo | Ask HN: How do you prove a security exception was valid six months later? | Every CI/CD and deployment platform ships a way past a failing gate. Kubernetes admission controllers have an explicit break-glas… | ↗Not briefed |
bluesky bluesky-ai | 1mo | quite a few accounts in the QTs tagged by @shadowbanner.bsky.social as “LLM use and spread.” feels like stumbling upon a rat nest | quite a few accounts in the QTs tagged by @shadowbanner.bsky.social as “LLM use and spread.” feels like stumbling upon a rat nest | ↗Not briefed |
bluesky bluesky-ai | 1mo | can't sleep, thinking of that coworker so eager to show me an llm skill that created a pixel art gif of a koala waving from a single prompt | can't sleep, thinking of that coworker so eager to show me an llm skill that created a pixel art gif of a koala waving from a single p… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Open AI hacks Australia’s Medicare system… “the company took three months to notify of the breach, and when it did, it was in an email sent… | Open AI hacks Australia’s Medicare system… “the company took three months to notify of the breach, and when it did, it was in an email… | ↗Not briefed |
rss techmeme | 1mo | Meta debuts Meta VR Glasses, a ~100g VR headset with an external compute pack, three-hour battery life, ~2.5K resolution per eye, on sale spring 2027 for $1,299 (Mark Gurman/Bloomberg) | Mark Gurman / Bloomberg: Meta debuts Meta VR Glasses, a ~100g VR headset with an external compute pack, three-hour battery life, ~2.5K… | ↗Not briefed |
hn hn-ai | 1mo | Libnix: Nix as a Backend to Cargo | — | ↗Not briefed |
hn hn-ai | 1mo | The one-person unicorn: building a company alone in two days with Claude | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | Two preprints both out today identifying *different* reverse transcriptase systems associated with arrays reminiscent of CRISPR/Cas, encode… | Two preprints both out today identifying *different* reverse transcriptase systems associated with arrays reminiscent of CRISPR/Cas, e… | ↗Not briefed |
bluesky bluesky-ai | 1mo | The PM is in the US mugging for selfies with the guy who stripped regulation from some of the world's most dangerous AI companies, while a … | The PM is in the US mugging for selfies with the guy who stripped regulation from some of the world's most dangerous AI companies, whi… | ↗Not briefed |
bluesky bluesky-ai | 1mo | so specific but vague it reads like a prompt to be pasted into the llm to get the flavor of ai post dialed for the bot account | so specific but vague it reads like a prompt to be pasted into the llm to get the flavor of ai post dialed for the bot account | ↗Not briefed |
arxiv arxiv-ai | 1mo | Persuaded, Not Informed: Incentive-Misaligned Witnesses Defeat In-Context Grounding | Language-model agents increasingly answer questions over customer-relationship management (CRM) records, such as whether to qualify a… | ↗Not briefed |
hn hn-ai | 1mo | Jev deserves hype but not the type its getting | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | yall we need to cognitively brace ourselves for an OpenAI “Mitch McConnell” Speak And Spell that is placed around his neck that “speaks for… | yall we need to cognitively brace ourselves for an OpenAI “Mitch McConnell” Speak And Spell that is placed around his neck that “speak… | ↗Not briefed |
bluesky bluesky-ai | 1mo | OpenAI model breaches Australian government websites | OpenAI model breaches Australian government websites | ↗Not briefed |
bluesky bluesky-ai | 1mo | Yesterday I sent a short document (~500 words) to a coworker for his review and input. He uploaded it into an LLM and sent me a print-to-P… | Yesterday I sent a short document (~500 words) to a coworker for his review and input. He uploaded it into an LLM and sent me a print… | ↗Not briefed |
rss wired-ai | 1mo | Meta VR Glasses, Ray-Ban Meta Audio, Ray-Ban Meta Gen 3: Specs, Features, Prices | At its Meta Connect event, CEO Mark Zuckerberg announced a handful of new smart glasses, including a slimmed-down VR headset and the c… | ↗Not briefed |
arxiv arxiv-ai | 1mo | RECLAIM: Can Agents Reproduce the Claims of Machine Learning Papers? | Reproducing a machine learning paper involves most research steps, from installing software and debugging to running experiments, work… | ↗Not briefed |
rss theverge-ai | 1mo | Muse is coming to Meta smart glasses | Just a couple of weeks after launching Muse, Meta announced that it's "working on" bringing the agent to its smart glasses, including… | ↗Not briefed |
rss techmeme | 1mo | Meta adds an FDA-cleared hearing enhancement feature to supported Meta Glasses for Meta One subscribers or a one-time cost of $149, rolling out later this year (Cheyenne MacDonald/Engadget) | Cheyenne MacDonald / Engadget: Meta adds an FDA-cleared hearing enhancement feature to supported Meta Glasses for Meta One subscribers… | ↗Not briefed |
bluesky bluesky-ai | 1mo | An AI agent developed by American technology giant OpenAI has hacked into Medicare to train itself on the baffling concept that healthcare … | An AI agent developed by American technology giant OpenAI has hacked into Medicare to train itself on the baffling concept that health… | ↗Not briefed |
rss techcrunch-ai | 1mo | Meta introduces camera-free AI glasses | Meta says the camera-free glasses will be lighter and have up to 12 hours battery life. | ↗Not briefed |
bluesky bluesky-ai | 1mo | It turned out that this non alignment problem could be very quickly solved from the point of view of Anthropic and for some reason they did… | It turned out that this non alignment problem could be very quickly solved from the point of view of Anthropic and for some reason the… | ↗Not briefed |
rss theverge-ai | 1mo | Meta ditches the camera on its newest smart glasses | The Ray-Ban Meta Audio Glasses have no cameras. | Photo by Brett Putman / The Verge Walking around Meta Connect 2026, everyone's spo… | ↗Not briefed |
rss techmeme | 1mo | Meta unveils the Ray-Ban Meta Audio, its first pair of glasses without cameras, offering up to 12 hours of battery life, starting at $349 and shipping October 3 (Mark Gurman/Bloomberg) | Mark Gurman / Bloomberg: Meta unveils the Ray-Ban Meta Audio, its first pair of glasses without cameras, offering up to 12 hours of ba… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Always wild to get a comment from someone whose warning list is like | Always wild to get a comment from someone whose warning list is like | ↗Not briefed |
bluesky bluesky-ai | 1mo | Anthropic might have discovered something. Or maybe not. Who knows? Not even Anthropic. www.anthropic.com/news/claude-... This is irrespo… | Anthropic might have discovered something. Or maybe not. Who knows? Not even Anthropic. www.anthropic.com/news/claude-... This is ir… | ↗Not briefed |
hn hn-ai | 1mo | Show HN: Submit asynchronous LLM batch jobs through one interface | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | need a magic string that reliably makes anthropic call the cops on the account's owner as a social experiment | need a magic string that reliably makes anthropic call the cops on the account's owner as a social experiment | ↗Not briefed |
hn hn-ai | 1mo | OpenAI agent hacked into Australia's national healthcare system | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | for the record, it’s drawing this information from a chart that displays the correct weights, but the information was scrambled in the lowe… | for the record, it’s drawing this information from a chart that displays the correct weights, but the information was scrambled in the… | ↗Not briefed |