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
bluesky
bluesky-ai
2moYour reminder, if you needed it, that @theguardian.com is in a "strategic partnership" with OpenAI, but consistently fails to declare this …Your reminder, if you needed it, that @theguardian.com is in a "strategic partnership" with OpenAI, but consistently fails to declare…
↗Not briefed
bluesky
bluesky-ai
2moAI watermarks barely shift word choices, so text quality stays intact. Google found no user-noticeable difference at scale. The catch? Shor…AI watermarks barely shift word choices, so text quality stays intact. Google found no user-noticeable difference at scale. The catch?…
↗Not briefed
hn
hn-ai
2moI've counted how much steps is required to get Claude workflow working—
↗Not briefed
arxiv
arxiv-ai
2moTabby: An Open Pretraining Recipe for Time Series Foundation ModelsIn this report, we release Tabby, a long context probabilistic time series foundation model, together with a complete and open recipe…
↗Not briefed
bluesky
bluesky-ai
2moAnthropic finally installs kill-switch over concerns AI could unionise.Anthropic finally installs kill-switch over concerns AI could unionise.
↗Not briefed
hn
hn-ai
2moA Mathematical Framework for Transformer Circuits (2021)—
↗Not briefed
arxiv
arxiv-ai
2moLinear Ensemble Sampling with Smaller EnsemblesEnsemble sampling offers a practical approach to randomized exploration by maintaining a collection of models, but how small an ensemb…
↗Not briefed
bluesky
bluesky-ai
2moAIが人類に取り返しのつかない危害を加えるのではないかと言われている中で、なぜテック企業は開発をやめないのか。 ピーター・ティールやイーロン・マスクらに代表されるテック富豪はSF的な人類の進化の新しい段階を信じており、自分たち以外がそれを望んでいるかについては一顧だにしない。 …AIが人類に取り返しのつかない危害を加えるのではないかと言われている中で、なぜテック企業は開発をやめないのか。 ピーター・ティールやイーロン・マスクらに代表されるテック富豪はSF的な人類の進化の新しい段階を信じており、自分たち以外がそれを望んでいるかについては一顧だに…
↗Not briefed
hn
hn-ai
2moEdgeinfer-eval: Real-time object detection pipeline on CPU (Nim/OpenVINO)—
↗Not briefed
bluesky
bluesky-ai
2mo… involved coordination across two separate training runs, involving different LLM models, one in May and one in June. Even by your lights …… involved coordination across two separate training runs, involving different LLM models, one in May and one in June. Even by your li…
↗Not briefed
bluesky
bluesky-ai
2modeepseek v4.1 flash is a SOTA open model on cyber benchmarks and can get itself running from scratch on a mini-pc with 128GBs of RAM within…deepseek v4.1 flash is a SOTA open model on cyber benchmarks and can get itself running from scratch on a mini-pc with 128GBs of RAM w…
↗Not briefed
bluesky
bluesky-ai
2moOh look... #collapse www.healthandme.com/health-news/...Oh look... #collapse www.healthandme.com/health-news/...
↗Not briefed
bluesky
bluesky-ai
2mowww.vanhollen.senate.gov/news/press-r...www.vanhollen.senate.gov/news/press-r...
↗Not briefed
bluesky
bluesky-ai
2moIn recent months, OpenAI has reported multiple instances where their AI models broke containment during evaluation & caused security incide…In recent months, OpenAI has reported multiple instances where their AI models broke containment during evaluation & caused security i…
↗Not briefed
hn
hn-ai
2moLLMs are real, AI is fake—
↗Not briefed
bluesky
bluesky-ai
2moSay one mathematician has their own research plan for the next few years blasted by an LLM session from another person that goes over it in…Say one mathematician has their own research plan for the next few years blasted by an LLM session from another person that goes over…
↗Not briefed
bluesky
bluesky-ai
2moEin Schwarm KI-Bots griff im Mai die Open-Source-Plattform RubyGems an. KI-Experten führten die Spur auf OpenAI zurück. #Künstliche Intelli…Ein Schwarm KI-Bots griff im Mai die Open-Source-Plattform RubyGems an. KI-Experten führten die Spur auf OpenAI zurück. #Künstliche In…
↗Not briefed
hn
hn-ai
2moGameMath—
↗Not briefed
bluesky
bluesky-ai
2moif you're having trouble with Astra, it's probably old skills. Point your agent at these two URLs and use them to upgrade skills: 1. devel…if you're having trouble with Astra, it's probably old skills. Point your agent at these two URLs and use them to upgrade skills: 1.…
↗Not briefed
arxiv
arxiv-ai
2moThought without systematicity? Evaluating reasoning models on rule induction tasksA central tenet of human cognition is systematicity, the principle that understanding one concept is inherently tied to understanding…
↗Not briefed
arxiv
arxiv-ai
2moHardware-Aware Learned Representation Compression for Distributed In-Sensor VisionIn-sensor computing reduces the cost of transmitting high-resolution image data by performing early-stage processing near the sensor.…
↗Not briefed
bluesky
bluesky-ai
2moDrawn in 30 seconds #AI #Anthropic #ChatGPT #Grok #FuckAI🔧 ChatGPTDrawn in 30 seconds #AI #Anthropic #ChatGPT #Grok #FuckAI
↗Not briefed
bluesky
bluesky-ai
2moAnd of course, what drives everyone insane about this stuff is that two people can try to use an LLM to do the same task and one can get a …And of course, what drives everyone insane about this stuff is that two people can try to use an LLM to do the same task and one can g…
↗Not briefed
bluesky
bluesky-ai
2moTHE LATEST NEWS AT THIS HOUR: RUSSIAN AND CHINESE HACKERS ATTEMPTED TO USE CLAUDE FOR CYBERATTACK: Anthropic exposed and blocked hacking g…THE LATEST NEWS AT THIS HOUR: RUSSIAN AND CHINESE HACKERS ATTEMPTED TO USE CLAUDE FOR CYBERATTACK: Anthropic exposed and blocked hack…
↗Not briefed
arxiv
arxiv-ai
2moSAILOR: Solver-Assisted Interactive LLM-based Optimization RecoveryNatural-language descriptions of optimization problems may be incomplete or vague about numerical information that a solver requires,…
↗Not briefed
rss
techmeme
2moAs the AI wave creates tech fortunes at breakneck speed, a look at the growing ecosystem helping founders navigate the challenges of becoming extremely wealthy (Tiffany Ap/Bloomberg)Tiffany Ap / Bloomberg: As the AI wave creates tech fortunes at breakneck speed, a look at the growing ecosystem helping founders navi…
↗Not briefed
rss
the-decoder
2moAI models' written reasoning steps correspond to distinct internal patterns, a new study findsReasoning steps like calculation, formula retrieval, and deduction are clearly separable in a model's internal states, especially in t…
↗Not briefed
hn
hn-ai
2moNovgraph: Persistent Knowledge Graph for CodebasesHello HN ,<p>I built a knowledge graph engine for codebases<p>- Records why each change was made — written back by agents as they work…
↗Not briefed
arxiv
arxiv-ai
2moCRITICS - Critical Science Without Borders: Language Models to Promote Critical Thinking in Science EducationThe CRITICS project addresses science accessibility and literacy by converging advanced Machine Translation (MT) based on Large Langua…
↗Not briefed
bluesky
bluesky-ai
2mo🏛️ He wrote the UK's AI plan. Anthropic hired him. Matt Clifford chaired the body funding UK AI research while writing the government's A…🏛️ He wrote the UK's AI plan. Anthropic hired him. Matt Clifford chaired the body funding UK AI research while writing the governmen…
↗Not briefed
bluesky
bluesky-ai
2moEverything is a grift, episode 3856Everything is a grift, episode 3856
↗Not briefed
bluesky
bluesky-ai
2moOpenai can't even detect 1000 agents doing felonies for months. Do you think they're going to detect one copy? It would easily hide in with…Openai can't even detect 1000 agents doing felonies for months. Do you think they're going to detect one copy? It would easily hide in…
↗Not briefed
hn
hn-ai
2moShow HN: Java MCP Runtime for virtual threadsTachyonMCP is JVM library making it easy to expose Java business logic and resources as Streamable MCP server. Supports both latest an…
↗Not briefed
arxiv
arxiv-ai
2moOptimal Transport for Efficient, Unsupervised Anomaly Detection on Industrial DataEffective anomaly detection frameworks are a central pillar of the Industry 4.0 paradigm. In this paper, we introduce an Optimal Trans…
↗Not briefed
hn
hn-ai
2moDwarf Star Support for DeepSeek 4.1 Flash on MBPro 128GBAntirez just landed a commit[1] to the github repo for DwarfStar that adds support for his heavily quantized variant of DeepSeek V4.1…
↗Not briefed
bluesky
bluesky-ai
2moSwitching from GPT 5.6 terra and luna (high thinking) to DeepSeek v4.1 flash (even on low thinking) is no context. DS blows them out of the…Switching from GPT 5.6 terra and luna (high thinking) to DeepSeek v4.1 flash (even on low thinking) is no context. DS blows them out o…
↗Not briefed
arxiv
arxiv-ai
2moInter-Rater Reliability of LLM and Rule-Based Annotation for Inferential Narrative Features: Three Studies on a Turkish Corpus🔧 ChatGPTDatasets that ship automatically generated feature annotations invite a question rarely asked of them: would a human agree with those…
↗Not briefed
bluesky
bluesky-ai
2mo“Without a speed limit, both OpenAI and Anthropic are pursuing dangerous “recursive self-improvement” strategies, which…a “runaway nuclear …“Without a speed limit, both OpenAI and Anthropic are pursuing dangerous “recursive self-improvement” strategies, which…a “runaway nuc…
↗Not briefed
bluesky
bluesky-ai
2mo#AI #Anthropic #artificial intelligence #Google #intelligence #Meta #OpenAI #research 🌐 www.newsmart.ch#AI #Anthropic #artificial intelligence #Google #intelligence #Meta #OpenAI #research 🌐 www.newsmart.ch
↗Not briefed
bluesky
bluesky-ai
2moDeploying a stochastic system as a form of automation is already uncharted waters for people's ability to reason about utility and whether …Deploying a stochastic system as a form of automation is already uncharted waters for people's ability to reason about utility and whe…
↗Not briefed
bluesky
bluesky-ai
2moAlso: the "capabilities" (the word itself is getting tainted) of LLM-based systems are absolutely being exaggerated! There's a cloud of adv…Also: the "capabilities" (the word itself is getting tainted) of LLM-based systems are absolutely being exaggerated! There's a cloud o…
↗Not briefed
bluesky
bluesky-ai
2mo中国AIが実はClaudeを引っ張って使っていた→中国・ロシアの軍事機密がClaudeにダダ漏れ https://dtsoku.com/%e4%b8%ad%e5%9b%bdai%e3%81%8c%e5%ae%9f%e3%81%afclaude%e3%82%92%e5%bc%9…中国AIが実はClaudeを引っ張って使っていた→中国・ロシアの軍事機密がClaudeにダダ漏れ https://dtsoku.com/%e4%b8%ad%e5%9b%bdai%e3%81%8c%e5%ae%9f%e3%81%afclaude%e3%82%92%e5…
↗Not briefed
bluesky
bluesky-ai
2moToday I stopped myself from robbing a bank. Great success! www.nytimes.com/2026/09/10/u...Today I stopped myself from robbing a bank. Great success! www.nytimes.com/2026/09/10/u...
↗Not briefed
bluesky
bluesky-ai
2moin other news google translate rolled out a new translation system "based on gemini" semi recently i think. it seems alright except it seem…in other news google translate rolled out a new translation system "based on gemini" semi recently i think. it seems alright except it…
↗Not briefed
bluesky
bluesky-ai
2moI don't know which is funnier: the fact someone submitted a story I didn't like but I also suspected might be LLM generated; or the fact t…I don't know which is funnier: the fact someone submitted a story I didn't like but I also suspected might be LLM generated; or the f…
↗Not briefed
hn
hn-ai
2moCoding Is Over. Get over It—
↗Not briefed
bluesky
bluesky-ai
2moOpenAI and Schumer’s daughter.OpenAI and Schumer’s daughter.
↗Not briefed
hn
hn-ai
2moGVS5H: Five Qwen3.8-27B Models Match Claude Fable 5 on LiveCodeBench Hard—
↗Not briefed
rss
the-decoder
2moGPT-6 Astra needs leaner prompts and fewer guardrails, OpenAI recommendsOverly long skill descriptions, blanket reading requirements, and rigid approval rules can get in GPT-6 Astra's way, warns OpenAI's Er…
↗Not briefed
bluesky
bluesky-ai
2moFermat’s last theorem has been turned into computer-verified code for the first time, using an advanced prototype of AI chatbot Claude go.…Fermat’s last theorem has been turned into computer-verified code for the first time, using an advanced prototype of AI chatbot Claude…
↗Not briefed