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 | 10d | When Safety Becomes an Attack Surface: Manipulating Abstention in Tool-Using AI Agents | An AI agent's decision not to act can itself become an attack surface. We study warning-induced over-abstention, an availability attac… | ↗Not briefed |
bluesky bluesky-ai | 10d | Anyway, Anthropic has a DC office so I'm not surprised to see their employees on the list; he's a great senator and a lot of DC based emplo… | Anyway, Anthropic has a DC office so I'm not surprised to see their employees on the list; he's a great senator and a lot of DC based… | ↗Not briefed |
bluesky bluesky-ai | 10d | at this point i'm less concerned about house house fucking up and more frustrated at the discourse surrounding it. house house seems to at … | at this point i'm less concerned about house house fucking up and more frustrated at the discourse surrounding it. house house seems t… | ↗Not briefed |
bluesky bluesky-ai | 10d | (step 1: generate linguistic probabilities live by asking two sets of intersecting questions and seeing what the audience says step 2: mo… | (step 1: generate linguistic probabilities live by asking two sets of intersecting questions and seeing what the audience says step… | ↗Not briefed |
bluesky bluesky-ai | 10d | What do you think the difference is, exactly? A GLM 5.3 request uses ~1.22 Wh at Neuralwatt. An RTX 5060 maxes out at 145W. So playing a gr… | What do you think the difference is, exactly? A GLM 5.3 request uses ~1.22 Wh at Neuralwatt. An RTX 5060 maxes out at 145W. So playing… | ↗Not briefed |
bluesky bluesky-ai | 10d | 💯 @schumer.senate.gov has been pushing one of the worst bills ever — S Bill 5653 BAAJA, aka the Dirty Data Center bill, which would underm… | 💯 @schumer.senate.gov has been pushing one of the worst bills ever — S Bill 5653 BAAJA, aka the Dirty Data Center bill, which would u… | ↗Not briefed |
bluesky bluesky-ai | 10d | RANDOM INCIDENTAL NOTE: after about 1.5 hours of discussion, I DID figure out a way to transform tinker toys into a useable visual demons… | RANDOM INCIDENTAL NOTE: after about 1.5 hours of discussion, I DID figure out a way to transform tinker toys into a useable visual d… | ↗Not briefed |
bluesky bluesky-ai | 10d | Anthropic is releasing a suite of tools that weave the company’s artificial intelligence capabilities into more everyday office work | Anthropic is releasing a suite of tools that weave the company’s artificial intelligence capabilities into more everyday office work | ↗Not briefed |
bluesky bluesky-ai | 10d | Anyway: More Jensen, more jacket, Anthropic IPO, and more with @edzitron.com | Anyway: More Jensen, more jacket, Anthropic IPO, and more with @edzitron.com | ↗Not briefed |
hn hn-ai | 10d | Show HN: Mimzie, an agent-neutral alternative to the Claude Suite/Claude Cowork🔧 ChatGPT | hi HN!<p>I built Mimzie, an agent neutral AI workspace inspired by the Claude Suite (Docs, Sheets, Slides) and Cowork.<p>It's a f… | ↗Not briefed |
hn hn-ai | 10d | Australia considers criminal charges for AI companies that fail to prevent harm | — | ↗Not briefed |
hn hn-ai | 10d | Retake: OSS iOS App Store preview video generator | — | ↗Not briefed |
bluesky bluesky-ai | 10d | It's not true. That section on OpenSecrets means people who *work at* Anthropic donated to him. OpenSecrets explains that, but it's very co… | It's not true. That section on OpenSecrets means people who *work at* Anthropic donated to him. OpenSecrets explains that, but it's ve… | ↗Not briefed |
rss techmeme | 10d | A deep dive into AMD's Zen 6 EPYC Venice: it uses TSMC's N2P, the 9686F often beats Nvidia's Vera, a dual 9996 is up to 93% faster than a dual Zen 5 9965, more (Jake Roach/Tom's Hardware) | Jake Roach / Tom's Hardware: A deep dive into AMD's Zen 6 EPYC Venice: it uses TSMC's N2P, the 9686F often beats Nvidia's Vera, a dual… | ↗Not briefed |
arxiv arxiv-ai | 10d | Is Jev a Good Multilingual Judge? | Generative large language models are widely used for multilingual text evaluation, yet whether decision models that directly output ju… | ↗Not briefed |
bluesky bluesky-ai | 10d | where's that report about llm benchmark results completely flubbing it when you change the capitalization of prompts? I want to add it to m… | where's that report about llm benchmark results completely flubbing it when you change the capitalization of prompts? I want to add it… | ↗Not briefed |
bluesky bluesky-ai | 10d | i don't have to waffle about productivity with llm coding because the games industry and the tech industry have both been taking the path o… | i don't have to waffle about productivity with llm coding because the games industry and the tech industry have both been taking the p… | ↗Not briefed |
bluesky bluesky-ai | 10d | We had, "Don't have the resources for that." And we had, "Gonna take some research to solve this." And we even had, "This is too hard and n… | We had, "Don't have the resources for that." And we had, "Gonna take some research to solve this." And we even had, "This is too hard… | ↗Not briefed |
bluesky bluesky-ai | 10d | My brain making some connection between the deep diving for pre-nuclear test steel and pre-LLM games. I know there are some studios who don… | My brain making some connection between the deep diving for pre-nuclear test steel and pre-LLM games. I know there are some studios wh… | ↗Not briefed |
hn hn-ai | 10d | AI is an information gateway. Audiences still don't want it making their news | — | ↗Not briefed |
bluesky bluesky-ai | 10d | need to do inverse speculative decoding where if i say something that the llm trained on me would say with >80% probability then it doesnt … | need to do inverse speculative decoding where if i say something that the llm trained on me would say with >80% probability then it do… | ↗Not briefed |
bluesky bluesky-ai | 10d | So what do we do about the LLM industry? Organize, demand change, yes. But also we MUST discuss with peers, and arrive to some kind of b… | So what do we do about the LLM industry? Organize, demand change, yes. But also we MUST discuss with peers, and arrive to some kind… | ↗Not briefed |
hn hn-ai | 10d | Show HN: PeepNote – Local desktop notes for saving and reusing LLM responses | — | ↗Not briefed |
bluesky bluesky-ai | 10d | Really hope that everyone getting sanctimonious about house house has a proportionally greater amount of vitriol for valve, the giant corpo… | Really hope that everyone getting sanctimonious about house house has a proportionally greater amount of vitriol for valve, the giant… | ↗Not briefed |
bluesky bluesky-ai | 10d | i do really wish the big walking devs at least disclosed the ai useage from the start instead of trying to hide it but honestly like. what … | i do really wish the big walking devs at least disclosed the ai useage from the start instead of trying to hide it but honestly like.… | ↗Not briefed |
bluesky bluesky-ai | 10d | Van Hollen's #2 donor is Anthropic & he's taken twice the amount in donations as Schumer. Ugh I hate researching politicians | Van Hollen's #2 donor is Anthropic & he's taken twice the amount in donations as Schumer. Ugh I hate researching politicians | ↗Not briefed |
bluesky bluesky-ai | 10d | OpenAI posts steep losses even as quarterly revenue rises to $10 billion, WSJ reports reut.rs/4kaa5dE | OpenAI posts steep losses even as quarterly revenue rises to $10 billion, WSJ reports reut.rs/4kaa5dE | ↗Not briefed |
bluesky bluesky-ai | 10d | Anthropic has recognised the illegally occupied territories as Russian? | Anthropic has recognised the illegally occupied territories as Russian? | ↗Not briefed |
bluesky bluesky-ai | 10d | I've said before but I'm baffled at all these companies that used to be so strict about how their brand and products are shown, the exactin… | I've said before but I'm baffled at all these companies that used to be so strict about how their brand and products are shown, the ex… | ↗Not briefed |
bluesky bluesky-ai | 10d | It doesn’t also look at having to deal with Agentic flooding in their inbound communication channels- not sure any service provider can be … | It doesn’t also look at having to deal with Agentic flooding in their inbound communication channels- not sure any service provider ca… | ↗Not briefed |
bluesky bluesky-ai | 10d | ive been saying that deepseek will win the culture war because everyone apparently now agrees they're a Big Fat Fish™ | ive been saying that deepseek will win the culture war because everyone apparently now agrees they're a Big Fat Fish™ | ↗Not briefed |
rss techmeme | 10d | NY-based Daylight, which provides AI-powered managed security, raised an $80M Series B led by Goldman Sachs Alternatives, sources say at a ~$400M valuation (Maria Armental/Wall Street Journal) | Maria Armental / Wall Street Journal: NY-based Daylight, which provides AI-powered managed security, raised an $80M Series B led by Go… | ↗Not briefed |
arxiv arxiv-ai | 10d | Effective Teacher Support for Reasoning Distillation | Forward-KL language-model distillation normally computes dense teacher and student scores over the full vocabulary at every supervised… | ↗Not briefed |
bluesky bluesky-ai | 10d | You also forgot that Schumer’s daughter literally is a lobbyist for OpenAI. | You also forgot that Schumer’s daughter literally is a lobbyist for OpenAI. | ↗Not briefed |
bluesky bluesky-ai | 10d | i'm conflicted on the big walk discourse and will refrain from going deep into my thoughts but what im certain about is that llm use is nev… | i'm conflicted on the big walk discourse and will refrain from going deep into my thoughts but what im certain about is that llm use i… | ↗Not briefed |
arxiv arxiv-ai | 10d | Generalize Then Personalize: Wearable-Based Heart Rate Forecasting Under Scarce Ruck-March Data | Heart rate (HR) forecasting may provide advance information about near-future cardiovascular response, potentially supporting physiolo… | ↗Not briefed |
bluesky bluesky-ai | 10d | Getting mad about the energy usage with this specific case is extra crazy because how do you think your computer is rendering the game? GPU… | Getting mad about the energy usage with this specific case is extra crazy because how do you think your computer is rendering the game… | ↗Not briefed |
bluesky bluesky-ai | 10d | does the engine have ai code? does the plugin shipped with the game have ai code? was the solution to your problem that someone posted writ… | does the engine have ai code? does the plugin shipped with the game have ai code? was the solution to your problem that someone posted… | ↗Not briefed |
bluesky bluesky-ai | 10d | Grab the rest of our library with TrueRef+! We are gearing up for a HUGE last half of 2026! Lots of exciting site upgrades incoming. For $1… | Grab the rest of our library with TrueRef+! We are gearing up for a HUGE last half of 2026! Lots of exciting site upgrades incoming. F… | ↗Not briefed |
bluesky bluesky-ai | 10d | even godot! their policy does not ban LLM code, it basically says “don’t commit huge swathes of LLM code that you did not review or underst… | even godot! their policy does not ban LLM code, it basically says “don’t commit huge swathes of LLM code that you did not review or un… | ↗Not briefed |
arxiv arxiv-ai | 10d | Open Challenges and Opportunities in Human-Agent Systems | Agentic AI systems increasingly undertake extended tasks and act on people's behalf, making human-agent interaction central to their e… | ↗Not briefed |
bluesky bluesky-ai | 10d | if you think the BDS movement is a great and good thing but don´t think anyone should boycott anything that has even a bit of LLM useage th… | if you think the BDS movement is a great and good thing but don´t think anyone should boycott anything that has even a bit of LLM usea… | ↗Not briefed |
bluesky bluesky-ai | 10d | In unrelated news, there's a coordinated attempt ongoing to spam oss-security@ with LLM-generated "vulnerability reports" via anonymous rem… | In unrelated news, there's a coordinated attempt ongoing to spam oss-security@ with LLM-generated "vulnerability reports" via anonymou… | ↗Not briefed |
arxiv arxiv-ai | 10d | AutoAssumption: Searching over Assumptions Discovers ML-Systems Algorithms | Many advances in ML systems came from breaking an assumption that earlier methods took for granted: speculative decoding dropped the a… | ↗Not briefed |
bluesky bluesky-ai | 10d | In the OpenAI-Hugging Face incident, Stephanie T. Nguyen, Erie Meyer, Adam Pickersgill and Mihir Kshirsagar see “conventional failures of o…🔧 Hugging Face | In the OpenAI-Hugging Face incident, Stephanie T. Nguyen, Erie Meyer, Adam Pickersgill and Mihir Kshirsagar see “conventional failures… | ↗Not briefed |
hn hn-ai | 10d | Technical report on the game Densha de Go Plug and Play | — | ↗Not briefed |
bluesky bluesky-ai | 10d | hate to break it to everyone, but if games have to disclose LLM code then every game released in the past couple of years that uses a major… | hate to break it to everyone, but if games have to disclose LLM code then every game released in the past couple of years that uses a… | ↗Not briefed |
bluesky bluesky-ai | 10d | –– so i have this bad feeling that after anthropic IPO’s and they start really charging people for this thing they hope people are hooked o… | –– so i have this bad feeling that after anthropic IPO’s and they start really charging people for this thing they hope people are hoo… | ↗Not briefed |
bluesky bluesky-ai | 10d | - over 290 commits and 62,000 lines of code in a single day - rounded elements with *that* border - the subtext on the main page - it inclu… | - over 290 commits and 62,000 lines of code in a single day - rounded elements with *that* border - the subtext on the main page - it… | ↗Not briefed |
hn hn-ai | 10d | Pi: Virtual Models | — | ↗Not briefed |