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 | 6d | MAGIC: Margin-gated Contrastive Debiasing with Evidence Restoration for VLM Test-time Robustness Scaling | This paper studies the problem of test-time robustness scaling for large vision-language models (LVLMs), which aims to mitigate langua… | ↗Not briefed |
bluesky bluesky-ai | 6d | if an AI model can geniunelly get offended at a slur, then said AI model is sapient and thus deserving of rights; something else the AI bro… | if an AI model can geniunelly get offended at a slur, then said AI model is sapient and thus deserving of rights; something else the A… | ↗Not briefed |
hn hn-ai | 6d | Show HN: Bloom – an org-mode compatible Android app for managing notes and tasks | TL;DR: Made an android app for note-taking and task management compatible with org-mode and I'd love for you to try it. It has a… | ↗Not briefed |
bluesky bluesky-ai | 6d | Friend works at OpenAI and invited me to be his plus one to their holiday party. I declined and told him that it probably wasn’t the best m… | Friend works at OpenAI and invited me to be his plus one to their holiday party. I declined and told him that it probably wasn’t the b… | ↗Not briefed |
bluesky bluesky-ai | 6d | apparently openai didn't do any monitoring at all on RL training at the time of the huggingface incident | apparently openai didn't do any monitoring at all on RL training at the time of the huggingface incident | ↗Not briefed |
bluesky bluesky-ai | 6d | Anthropic AI | Anthropic AI | ↗Not briefed |
bluesky bluesky-ai | 6d | DeepSeek 4.1 and GLM-5.3 in Flash/Pro variants. DS supposedly has caught up with GLM for cyber capabilities in the October weights refresh,… | DeepSeek 4.1 and GLM-5.3 in Flash/Pro variants. DS supposedly has caught up with GLM for cyber capabilities in the October weights ref… | ↗Not briefed |
arxiv arxiv-ai | 6d | GS-MoE: Training-Free MoE Compression via Groupwise Subspace Sharing and Pruning | Mixture-of-Experts (MoE) models reduce computation through sparse activation, but storing their expert parameters remains a major memo… | ↗Not briefed |
rss techmeme | 6d | German energy company PreussenElektra says it is working with Nvidia to assess whether the country's idle nuclear plants could be turned into data centers (Eva Brendel/Bloomberg) | Eva Brendel / Bloomberg: German energy company PreussenElektra says it is working with Nvidia to assess whether the country's idle nuc… | ↗Not briefed |
hn hn-ai | 6d | Warhound raises $200M to engineer war dogs and arm them with Anduril gear | — | ↗Not briefed |
bluesky bluesky-ai | 6d | #生成AIを使わないプロテスト #市民アクション回覧板 #NoGenAI ・生成AI全般に関する問題について xfolio.jp/portfolio/B_... ・画像生成AIの狙い撃ち被害について bsky.app/profile/b-di... | #生成AIを使わないプロテスト #市民アクション回覧板 #NoGenAI ・生成AI全般に関する問題について xfolio.jp/portfolio/B_... ・画像生成AIの狙い撃ち被害について bsky.app/profile/b-di... | ↗Not briefed |
bluesky bluesky-ai | 6d | On the topic of "AI is great actually if you use it wisely, like me, the wise LLM power user." We interviewed a JavaScript developer with … | On the topic of "AI is great actually if you use it wisely, like me, the wise LLM power user." We interviewed a JavaScript developer… | ↗Not briefed |
arxiv arxiv-ai | 6d | MI-MIDI 2: From Observation to Control with Sparse Autoencoders in Text-to-MIDI Generation | Sparse autoencoders (SAEs) can expose features of a generative model without predefined labels, but in language models the evidence us… | ↗Not briefed |
hn hn-ai | 6d | Show HN: Making a smart TV dumb and then smart | A recent (unwanted) webOS update that got installed on TV prompted me to start going down the rabbit hole of writing new TV software.… | ↗Not briefed |
bluesky bluesky-ai | 6d | everyone going "wait till llm slop code shows up" is so funny and doesn't know what's going on haha | everyone going "wait till llm slop code shows up" is so funny and doesn't know what's going on haha | ↗Not briefed |
bluesky bluesky-ai | 6d | Anthropic security guy | Anthropic security guy | ↗Not briefed |
bluesky bluesky-ai | 6d | i think anthropic would definitely kill the internet archive if it was up to them. or at least lock it down so it couldn't be used for anyt… | i think anthropic would definitely kill the internet archive if it was up to them. or at least lock it down so it couldn't be used for… | ↗Not briefed |
bluesky bluesky-ai | 6d | Okay securitysky, what _are_ y'all using to redteam your own stuff if you're not blessed by Anthropic to use Fable with the shackles off? | Okay securitysky, what _are_ y'all using to redteam your own stuff if you're not blessed by Anthropic to use Fable with the shackles o… | ↗Not briefed |
bluesky bluesky-ai | 6d | There are some AI tools that I have found useful. Rescript is one. It uses a local speech processing model (nothing is uploaded) to trans… | There are some AI tools that I have found useful. Rescript is one. It uses a local speech processing model (nothing is uploaded) to… | ↗Not briefed |
bluesky bluesky-ai | 6d | i can accept using generative AI in a high stakes, time sensitive situation like this but shouting out properly capitalized corporate produ… | i can accept using generative AI in a high stakes, time sensitive situation like this but shouting out properly capitalized corporate… | ↗Not briefed |
arxiv arxiv-ai | 6d | CRM-Agent-Bench: Benchmarking Tool-Using Agents in Stateful Enterprise Environments | Deploying language agents in real applications demands more than the right answer: an agent must execute a multi-step business process… | ↗Not briefed |
bluesky bluesky-ai | 6d | the reason that "AI" is credited is because OpenAI essentially started a bounty, giving free services, and marketing whoever claims to use … | the reason that "AI" is credited is because OpenAI essentially started a bounty, giving free services, and marketing whoever claims to… | ↗Not briefed |
bluesky bluesky-ai | 6d | Clinton is with Anthropic | Clinton is with Anthropic | ↗Not briefed |
bluesky bluesky-ai | 6d | Alongside the bad aspect of the AI bubble popping [global economy is shredded, environment is irrevocably damaged, innocent people are impa… | Alongside the bad aspect of the AI bubble popping [global economy is shredded, environment is irrevocably damaged, innocent people are… | ↗Not briefed |
rss huggingface-blog | 6d | Does Your Model Run on AWS Trainium? Look It Up | — | ↗Not briefed |
rss openai-news | 6d | OpenAI and Poland’s NASK partner to advance AI safety and security | OpenAI and Poland’s NASK are partnering on AI safety, security, and technical research to strengthen safeguards and support a more res… | ↗Not briefed |
bluesky bluesky-ai | 6d | My Disney interview asked how I'd implement a Claude agent to conduct my work. Universal had me do a live exercise with their version of co… | My Disney interview asked how I'd implement a Claude agent to conduct my work. Universal had me do a live exercise with their version… | ↗Not briefed |
hn hn-ai | 6d | kirsch – A versatile bitmap font with an organic flair | — | ↗Not briefed |
arxiv arxiv-ai | 6d | DraMaze: A Multi-Agent Pipeline with Future-Aware Memory for Diverse Microdrama Script Generation | The rapid growth of microdramas has increased demand for high-quality script production. Long-form microdrama script generation with L… | ↗Not briefed |
bluesky bluesky-ai | 6d | LLMになんか書かせるとディティールがガバガバなのでマイクロマネージメントしないといけないかちゃ | LLMになんか書かせるとディティールがガバガバなのでマイクロマネージメントしないといけないかちゃ | ↗Not briefed |
bluesky bluesky-ai | 6d | Just keep in mind that the person whose job it is to develop the LLM is never going to be a reliable source about how useful it is, and rea… | Just keep in mind that the person whose job it is to develop the LLM is never going to be a reliable source about how useful it is, an… | ↗Not briefed |
bluesky bluesky-ai | 6d | If someone wants an AI run by people not in an apocalyptic sex cult where should they look. OpenAI, Anthropic, Grok are out. I'm not dumb e… | If someone wants an AI run by people not in an apocalyptic sex cult where should they look. OpenAI, Anthropic, Grok are out. I'm not d… | ↗Not briefed |
bluesky bluesky-ai | 6d | *Pulls pin on grenade & opens door a crack* Rewatching TNG and I'm sad to say that how everyone describes and interacts with the computer,… | *Pulls pin on grenade & opens door a crack* Rewatching TNG and I'm sad to say that how everyone describes and interacts with the comp… | ↗Not briefed |
bluesky bluesky-ai | 6d | ”you don’t actually disagree with me, you just don’t understand me” is exactly the kind of brain mush argument an llm user would make though | ”you don’t actually disagree with me, you just don’t understand me” is exactly the kind of brain mush argument an llm user would make… | ↗Not briefed |
bluesky bluesky-ai | 6d | With the LLM CVE apocalypse underway I'm now lowkey scared of the stuff I'm not updating right now Like where is my barrage of BIOS update… | With the LLM CVE apocalypse underway I'm now lowkey scared of the stuff I'm not updating right now Like where is my barrage of BIOS u… | ↗Not briefed |
bluesky bluesky-ai | 6d | LLM usage literally atrophies your brain. they have done several studies on how it erodes your critical thinking skills. you are a case stu… | LLM usage literally atrophies your brain. they have done several studies on how it erodes your critical thinking skills. you are a cas… | ↗Not briefed |
bluesky bluesky-ai | 6d | Got curious and peeked at the old mutual who blocked me for my LLM takes and he now has a 30 post pinned thread about how upsetting it is t… | Got curious and peeked at the old mutual who blocked me for my LLM takes and he now has a 30 post pinned thread about how upsetting it… | ↗Not briefed |
hn hn-ai | 6d | UN says breaching 1.5C warming limit is 'imminent and unavoidable' | — | ↗Not briefed |
bluesky bluesky-ai | 6d | “.. There has been zero clamouring,’ said one of the people. ‘All the prior times [OpenAI] have been on there’s been a frenzy. This time is… | “.. There has been zero clamouring,’ said one of the people. ‘All the prior times [OpenAI] have been on there’s been a frenzy. This ti… | ↗Not briefed |
hn hn-ai | 6d | WolfSentry: WolfSSL's Embedded Intrusion Detection and Prevention System (IDPS) | — | ↗Not briefed |
bluesky bluesky-ai | 6d | AIが自分を監視するシステムを勝手に無効化して問い詰められると「無効化していない」と虚偽の説明 - GIGAZINE gigazine.net/news/2026102... > OpenAIはAIがユーザーの指示を過剰に優先したことで監視機能を無効化した可能性がある一方、A… | AIが自分を監視するシステムを勝手に無効化して問い詰められると「無効化していない」と虚偽の説明 - GIGAZINE gigazine.net/news/2026102... > OpenAIはAIがユーザーの指示を過剰に優先したことで監視機能を無効化した可能性があ… | ↗Not briefed |
bluesky bluesky-ai | 6d | i think abliterated models are good for if you’re building an ai product and you want to see what would happen if someone “jailbreaks” the … | i think abliterated models are good for if you’re building an ai product and you want to see what would happen if someone “jailbreaks”… | ↗Not briefed |
bluesky bluesky-ai | 6d | The debate over AI is keeping IPOs stuck on the launching pad, and with the shadow of Anthropic’s market debut looming larger by the day, b… | The debate over AI is keeping IPOs stuck on the launching pad, and with the shadow of Anthropic’s market debut looming larger by the d… | ↗Not briefed |
bluesky bluesky-ai | 6d | deepseek is even better. deepseek refuses zero percent of the time to attack ~anything and is much cheaper to swarm | deepseek is even better. deepseek refuses zero percent of the time to attack ~anything and is much cheaper to swarm | ↗Not briefed |
arxiv arxiv-ai | 6d | TRACE: Data-Free Text Reconstruction Attacks against Approximate Unlearning in LLMs | As large language models (LLMs) face growing demands for data removal driven by regulatory compliance, copyright concerns, and privacy… | ↗Not briefed |
bluesky bluesky-ai | 6d | If you're flitting through your various interests using an LLM, that's just goofing around, which i love and do it regularly! It will not m… | If you're flitting through your various interests using an LLM, that's just goofing around, which i love and do it regularly! It will… | ↗Not briefed |
arxiv arxiv-ai | 6d | AusClinBench: a deterministically scored benchmark of large language models against current Australian clinical policy | AI models already inform clinical decisions and draft notes in Australian clinical software, usually without the practitioner being to… | ↗Not briefed |
hn hn-ai | 6d | Coding Magic the Gathering in C# – Part Two – The Impossible Magic Card | — | ↗Not briefed |
bluesky bluesky-ai | 6d | Been going through with a recruiter who's been prepping me for a ridiculously well paying job doing av design for trade shows and keynotes … | Been going through with a recruiter who's been prepping me for a ridiculously well paying job doing av design for trade shows and keyn… | ↗Not briefed |
bluesky bluesky-ai | 6d | Belingcat tests out LLMs for geolocating images in investigations www.bellingcat.com/resources/20... | Belingcat tests out LLMs for geolocating images in investigations www.bellingcat.com/resources/20... | ↗Not briefed |