AINews Portal

News

50 items (filtered)

The raw AI news stream — everything we ingest across RSS, Hacker News, arXiv, GitHub, Bluesky, YouTube and more, newest first.

SourceAgeTitleSummaryStatus
arxiv
arxiv-ai
1moGEM-MPC: Balancing Exploration and Exploitation through Expert-Guided PlanningEffective exploration in high-dimensional continuous control remains a central challenge in reinforcement learning. Planning-based met…
↗Not briefed
bluesky
bluesky-ai
1moThe concern expressed by this software engineer isn't merely a jailbreak. The bigger concern is that a swarm of Chatbot agents were able to…The concern expressed by this software engineer isn't merely a jailbreak. The bigger concern is that a swarm of Chatbot agents were ab…
↗Not briefed
hn
hn-ai
1moWhat Using AI Therapy Gets Wrong—
↗Not briefed
bluesky
bluesky-ai
1moTrans rights are popular. When presented without culture war bullshit they just are. AI/LLM models and the resulting data centers have mass…Trans rights are popular. When presented without culture war bullshit they just are. AI/LLM models and the resulting data centers have…
↗Not briefed
bluesky
bluesky-ai
1moWann die KI-Expert uns um die Gefahr von unkontrollierte KI zu warnen, dann wir besser ihren gehört müssen. #KI www.spiegel.de/wirtschaft/…Wann die KI-Expert uns um die Gefahr von unkontrollierte KI zu warnen, dann wir besser ihren gehört müssen. #KI www.spiegel.de/wirtsc…
↗Not briefed
bluesky
bluesky-ai
1moComputer science pioneers Geoffrey Hinton and Stuart Russell sign onto a push for transparent AI risk assessments at OpenAI and Anthropic. …Computer science pioneers Geoffrey Hinton and Stuart Russell sign onto a push for transparent AI risk assessments at OpenAI and Anthro…
↗Not briefed
bluesky
bluesky-ai
1moFound the actual link (was looking at news articles), not xss, heap overflow on libheif letting them get RCE on discourse. www.hacktron.ai/…Found the actual link (was looking at news articles), not xss, heap overflow on libheif letting them get RCE on discourse. www.hacktro…
↗Not briefed
bluesky
bluesky-ai
1moLLMサービス(GenAI)はデータ収集の点ももちろんですが、 さらに怖い点として使用者の倫理観をぶっ壊して 「赤信号みんなで渡れば怖くない」 を本当に実現させた状態にしてしまったことと思います。 (よく使われる『民主化』という言葉を使うなら『悪事の民主化』だと思う・・・)LLMサービス(GenAI)はデータ収集の点ももちろんですが、 さらに怖い点として使用者の倫理観をぶっ壊して 「赤信号みんなで渡れば怖くない」 を本当に実現させた状態にしてしまったことと思います。 (よく使われる『民主化』という言葉を使うなら『悪事の民主化』だと思…
↗Not briefed
bluesky
bluesky-ai
1moTekoäly on jo valjastettu ihmisten tappamiseen sodissa. Ja nyt: "Tekoäly-yhtiö OpenAI kertoi, ettei se pysty hallitsemaan kehittyneimpiä t…Tekoäly on jo valjastettu ihmisten tappamiseen sodissa. Ja nyt: "Tekoäly-yhtiö OpenAI kertoi, ettei se pysty hallitsemaan kehittyneim…
↗Not briefed
bluesky
bluesky-ai
1moi think if you run an unmarked llm agent on here to helpfully respond to people expressing distress in order to build your brand you should…i think if you run an unmarked llm agent on here to helpfully respond to people expressing distress in order to build your brand you s…
↗Not briefed
bluesky
bluesky-ai
1moAnthropic and OpenAI need truly independent safety evaluators, experts say in public letterAnthropic and OpenAI need truly independent safety evaluators, experts say in public letter
↗Not briefed
bluesky
bluesky-ai
1mono, it genuinely just like came to me from a discussion about Anthropic and bioresearch of just '... is this for making more Aella bathwate…no, it genuinely just like came to me from a discussion about Anthropic and bioresearch of just '... is this for making more Aella bat…
↗Not briefed
rss
techmeme
1moA German court says Meta is liable for fake ads posted by third parties on Instagram and Facebook and orders it to remove such content and pay damages (Linda Pasquini/Reuters)Linda Pasquini / Reuters: A German court says Meta is liable for fake ads posted by third parties on Instagram and Facebook and orders…
↗Not briefed
bluesky
bluesky-ai
1moEL PITO DEL SERENO / ¿Nos matará la IA? 🤔 🛸 Empresas como OpenAI y Anthropic se aprovechan de discursos existenciales, casi religioso…EL PITO DEL SERENO / ¿Nos matará la IA? 🤔 🛸 Empresas como OpenAI y Anthropic se aprovechan de discursos existenciales, casi reli…
↗Not briefed
bluesky
bluesky-ai
1modeepseek my beloveddeepseek my beloved
↗Not briefed
bluesky
bluesky-ai
1moAnthropic has quietly set up a laboratory to do physical biology work as it pushes its AI ambitions into the world of drug science. The AI …Anthropic has quietly set up a laboratory to do physical biology work as it pushes its AI ambitions into the world of drug science. Th…
↗Not briefed
arxiv
arxiv-ai
1moCIBuzzBench: A Benchmark for Cross-Lingual Understanding of Chinese Internet BuzzwordsChinese social media has generated a vast and continually evolving lexicon of internet buzzwords whose meanings are often non-literal…
↗Not briefed
hn
hn-ai
1moPutin envoy and far-right AfD prepare talks to get Russian gas back for Germany—
↗Not briefed
bluesky
bluesky-ai
1moLike the Rs in strawberry and full glass of wine problem. An LLM doesn't know what a strawberry is, or what an R is, what "three" is, or wh…Like the Rs in strawberry and full glass of wine problem. An LLM doesn't know what a strawberry is, or what an R is, what "three" is,…
↗Not briefed
arxiv
arxiv-ai
1moTERMon: Detecting Persistent Behavioral Threats in Edge AI via Hardware-Native Ternary Runtime MonitorEdge AI accelerators are increasingly deployed in safety-critical environments, where model outputs may control physical actuators, ma…
↗Not briefed
bluesky
bluesky-ai
1moThe fact that a student can meaningfully use language and discuss ideas, without needing a copy of all the text in the world first, should …The fact that a student can meaningfully use language and discuss ideas, without needing a copy of all the text in the world first, sh…
↗Not briefed
hn
hn-ai
1moShow HN: Using a diffusion model to write docs quicklyHi folks,<p>On my previous teams so many people procrastinated writing docs. And the AI docs are just not good enough right now imho,…
↗Not briefed
hn
hn-ai
1moDoes this mean it's fine to use Claude subscriptions with third party harnesses?I&#x27;ve been curious about using a claude subscription with OpenCode, Pi and some other third party harnesses.<p>If you remember, th…
↗Not briefed
arxiv
arxiv-ai
1moSpecQuant: Speculative Decoding with Multi-Parent Quantization for Adaptive LLM InferenceRunning large language models (LLMs) locally continues to be limited by restrictions of compute and memory on consumer hardware. The p…
↗Not briefed
bluesky
bluesky-ai
1moenquanto isso, a grande imprensa, balizada pelos interesses do grande capital, nos alerta que um agente de IA hackeou uma empresa sem quere…enquanto isso, a grande imprensa, balizada pelos interesses do grande capital, nos alerta que um agente de IA hackeou uma empresa sem…
↗Not briefed
bluesky
bluesky-ai
1moHm, pretty sure I can think of a bigger theft of labor in our history www.nytimes.com/2026/09/17/t...Hm, pretty sure I can think of a bigger theft of labor in our history www.nytimes.com/2026/09/17/t...
↗Not briefed
hn
hn-ai
1moGrep your Claude Code/Codex history and jump back into the exact sessionBuilt a small devtool I wanted for myself:<p>context-find<p>Search all your Claude Code + Codex conversations across projects and mach…
↗Not briefed
bluesky
bluesky-ai
1moIn all the “will AI destroy us/won’t it” storming, we still don’t talk nearly enough about how to the corporations creating it are guilty o…In all the “will AI destroy us/won’t it” storming, we still don’t talk nearly enough about how to the corporations creating it are gui…
↗Not briefed
bluesky
bluesky-ai
1moFor people who think Excel spreadsheets can accurately predict the future then an AI model will seem like a godFor people who think Excel spreadsheets can accurately predict the future then an AI model will seem like a god
↗Not briefed
arxiv
arxiv-ai
1moBayesian classification of astronomical spectra with class uncertaintiesContext: We developed a probabilistic machine learning method with the aim of performing the O(10)-way classification of low- and high…
↗Not briefed
arxiv
arxiv-ai
1moOptimization Geometry of Equivalent Brownian RKHS RepresentationsEquivalent finite parameterizations can represent the same functions and intrinsic norm yet induce different optimization algorithms.…
↗Not briefed
bluesky
bluesky-ai
1moSo, from talking to two early former bay area OpenAI employees I trust, neither of whom were ever True Believing Rationalists (even rationa…So, from talking to two early former bay area OpenAI employees I trust, neither of whom were ever True Believing Rationalists (even ra…
↗Not briefed
bluesky
bluesky-ai
1moWhat are we even doing hereWhat are we even doing here
↗Not briefed
hn
hn-ai
1moNeverLose – Crash-proof any Python function with one decorator—
↗Not briefed
arxiv
arxiv-ai
1moCIPL: A Channel-Aware Framework for Recoverable Privacy Leakage in LLM AgentsPrivacy leakage in LLM agents is commonly evaluated within individual components such as memory, retrieval, or tool-use pipelines, whi…
↗Not briefed
bluesky
bluesky-ai
1moAn excellent thread bringing the realities of LLM behavior into contextAn excellent thread bringing the realities of LLM behavior into context
↗Not briefed
hn
hn-ai
1moRepoGuard – CLI to stop Cursor and Copilot from bypassing architecture—
↗Not briefed
hn
hn-ai
1moShow HN: OpsGraph – evidence-first PostgreSQL investigationsOpen-source, self-hosted PostgreSQL investigations with bounded read-only execution and inspectable evidence behind findings.<p>GitHub…
↗Not briefed
arxiv
arxiv-ai
1moListen Before You Speak: Response Planning from Listener Facial Reactions for Conversational Speech GenerationConversational speech depends on dialogue context and the listener's immediately preceding behavior. We propose ReACT-TTS, a two-stage…
↗Not briefed
hn
hn-ai
1moAllelify: Run commands in parallel while storing logs in /tmp to limit noise—
↗Not briefed
bluesky
bluesky-ai
1mo"“Look inward and accept that if you’re building Frankenstein, No 1, you should stop and No 2, when companies come to you and say: ‘We need…"“Look inward and accept that if you’re building Frankenstein, No 1, you should stop and No 2, when companies come to you and say: ‘We…
↗Not briefed
bluesky
bluesky-ai
1moJevon's Paradox for demand response? No, this is just greed and capitalism, I think. I really hope those making these sorts of claims actua…Jevon's Paradox for demand response? No, this is just greed and capitalism, I think. I really hope those making these sorts of claims…
↗Not briefed
bluesky
bluesky-ai
1modata centre 'demand flex' has been framed as a climate solution, even though their net effect on grids is probably still a worsening of emi…data centre 'demand flex' has been framed as a climate solution, even though their net effect on grids is probably still a worsening o…
↗Not briefed
bluesky
bluesky-ai
1moEinfach immer weiter Effective-Altruism-Spinner als Experten ernst nehmen. Kann ja keiner wissen, dass das ein totalitärer Kult ist. www.s…Einfach immer weiter Effective-Altruism-Spinner als Experten ernst nehmen. Kann ja keiner wissen, dass das ein totalitärer Kult ist.…
↗Not briefed
arxiv
arxiv-ai
1moGUARD: Natural Forgetting in Large Reasoning Models via Guided Answer-Reasoning DistillationRecent advances in large reasoning models (LRMs) have made machine unlearning more challenging, as protected facts or unsafe rationale…
↗Not briefed
arxiv
arxiv-ai
1moThe Spoken Wikipedia Presentation Corpus🔧 ModalWe present the Spoken Wikipedia Presentation Corpus, an extension of the Spoken Wikipedia Corpora featuring LLM-generated slide decks…
↗Not briefed
bluesky
bluesky-ai
1moAgreed. Best post I’ve read on this is from @eryk.bsky.social mail.cyberneticforests.com/models-dont-...Agreed. Best post I’ve read on this is from @eryk.bsky.social mail.cyberneticforests.com/models-dont-...
↗Not briefed
bluesky
bluesky-ai
1molol, someone on the school tech Google group used Claude to code an extension to hide AI in search results for students. Best use of AI yet.lol, someone on the school tech Google group used Claude to code an extension to hide AI in search results for students. Best use of A…
↗Not briefed
bluesky
bluesky-ai
1moHow many more are they not telling us about? If ever there was a time to promote/teach/stress critical thinking and independent thought, it…How many more are they not telling us about? If ever there was a time to promote/teach/stress critical thinking and independent though…
↗Not briefed
hn
hn-ai
1moExclusive – Anthropic sets up biology lab as it ramps AI drug program—
↗Not briefed