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
1moWashington Post just moved an especially good report on the developments today. Highly recommend. /13 www.washingtonpost.com/business/202...Washington Post just moved an especially good report on the developments today. Highly recommend. /13 www.washingtonpost.com/business/…
↗Not briefed
arxiv
arxiv-ai
1moVisual Navigation Transformer with Pose AttentionLearned navigation policies typically consume observations as a temporally ordered history, with positional encodings tying each obser…
↗Not briefed
bluesky
bluesky-ai
1moWhile Anthropic likes to make scary pronouncements of AI destroying humanity, the biggest risk to humanity is that people stop thinking and…While Anthropic likes to make scary pronouncements of AI destroying humanity, the biggest risk to humanity is that people stop thinkin…
↗Not briefed
bluesky
bluesky-ai
1moAnthropic plans to publish regular report on how much R&D work Claude “leads” meaning a human gives a high level prompt and Claude does the…Anthropic plans to publish regular report on how much R&D work Claude “leads” meaning a human gives a high level prompt and Claude doe…
↗Not briefed
bluesky
bluesky-ai
1moA bug-hunting independent security research team used Anthropic’s Claude software to gain access to an OpenAI employee’s ChatGPT account, g…🔧 ChatGPTA bug-hunting independent security research team used Anthropic’s Claude software to gain access to an OpenAI employee’s ChatGPT accou…
↗Not briefed
arxiv
arxiv-ai
1moInformation-Gain Rewards over Diversity-Pruned Tests: GT-Anchored Verifier Co-Training for Reliable Code Generation🔧 ModalSelf-play methods that co-train a single language model as both coder and test author promise to move code-generation RL beyond fixed…
↗Not briefed
bluesky
bluesky-ai
1moWell, this answers some questionsWell, this answers some questions
↗Not briefed
bluesky
bluesky-ai
1moAI training is the largest theft of labor in historyAI training is the largest theft of labor in history
↗Not briefed
bluesky
bluesky-ai
1mowww.wsj.com/tech/ai/hack... Hackers Used Anthropic’s Claude to Break Into OpenAIwww.wsj.com/tech/ai/hack... Hackers Used Anthropic’s Claude to Break Into OpenAI
↗Not briefed
bluesky
bluesky-ai
1moJUST IN: OpenAI has launched Astra for Law, a GPT-6 configuration built specifically for legal research and drafting.JUST IN: OpenAI has launched Astra for Law, a GPT-6 configuration built specifically for legal research and drafting.
↗Not briefed
hn
hn-ai
1moSpectral AI Architecture: Executive Summary and Core Concepts—
↗Not briefed
hn
hn-ai
1moWhat Is LPO, NPO, CPO? Knowledge of Optical Interconnect Technologies—
↗Not briefed
bluesky
bluesky-ai
1moI'm no legal expert, but "fair use" was a central principle of my open educational resources over the past 20+ yrs, and I did spend (waste?…I'm no legal expert, but "fair use" was a central principle of my open educational resources over the past 20+ yrs, and I did spend (w…
↗Not briefed
bluesky
bluesky-ai
1moKryten, a Role Model for AI! llewellyn.substack.com/p/kryten-a-r...Kryten, a Role Model for AI! llewellyn.substack.com/p/kryten-a-r...
↗Not briefed
bluesky
bluesky-ai
1moLlm이 교사를 대체할 수 없는건 지식의 문제가 아니라 llm 특유의 아부 때문인게 제일 큰거같아 범죄자가 범죄계획 말해도 맞장구 치고 자살시도하는 사람 자살계획도 맞장구 치는데 ;;Llm이 교사를 대체할 수 없는건 지식의 문제가 아니라 llm 특유의 아부 때문인게 제일 큰거같아 범죄자가 범죄계획 말해도 맞장구 치고 자살시도하는 사람 자살계획도 맞장구 치는데 ;;
↗Not briefed
bluesky
bluesky-ai
1moIf Claude can build your entire product, what becomes your’s real job? #AI #Startups #ClaudeIf Claude can build your entire product, what becomes your’s real job? #AI #Startups #Claude
↗Not briefed
rss
techmeme
1moAn interview with OpenAI researcher Noam Brown about multi-agent systems, AI solving the Navier-Stokes problem, the internal/external model gap, and more (Dwarkesh Patel/Dwarkesh Podcast)Dwarkesh Patel / Dwarkesh Podcast: An interview with OpenAI researcher Noam Brown about multi-agent systems, AI solving the Navier-Sto…
↗Not briefed
arxiv
arxiv-ai
1moReliability-Centered Evaluation of Sparse Longitudinal CT Lesion-Size Forecasting with Conformal Interval Calibration and Gompertz-Inspired RegularizationSparse longitudinal CT follow-up limits lesion-size forecasting when only a few prior observations are available. We constructed a fiv…
↗Not briefed
bluesky
bluesky-ai
1moRonan Farrow's piece in the New Yorker about Sam Altman that Jen mentioned a few times tonight can be found here: www.newyorker.com/magazin…Ronan Farrow's piece in the New Yorker about Sam Altman that Jen mentioned a few times tonight can be found here: www.newyorker.com/ma…
↗Not briefed
bluesky
bluesky-ai
1moOpenAI disclosed six incidents of models acting without authorization, hiding information and coordinating with each other, and built a fra…OpenAI disclosed six incidents of models acting without authorization, hiding information and coordinating with each other, and built…
↗Not briefed
bluesky
bluesky-ai
1mo“For years, Microsoft and OpenAI have fought to keep certain information out of the public eye in their fight with news organizations that …“For years, Microsoft and OpenAI have fought to keep certain information out of the public eye in their fight with news organizations…
↗Not briefed
bluesky
bluesky-ai
1mo全く遅すぎるくらい。ニューヨーク・タイムズとの裁判で明らかになったみたい。 “OpenAIとMicrosoft は、LLMがWebを破壊し、盗難に基づいて構築されていることを認めた” #生成AI法規制スタンディング #AI法罰則規定署名活動 Sep 17, 2026 …全く遅すぎるくらい。ニューヨーク・タイムズとの裁判で明らかになったみたい。 “OpenAIとMicrosoft は、LLMがWebを破壊し、盗難に基づいて構築されていることを認めた” #生成AI法規制スタンディング #AI法罰則規定署名活動 Sep 17, 2…
↗Not briefed
bluesky
bluesky-ai
1moThat's why I think OpenAI will just settle for 10 figures as a cost of doing business. Anthropic did the same. They're now so big that copy…That's why I think OpenAI will just settle for 10 figures as a cost of doing business. Anthropic did the same. They're now so big that…
↗Not briefed
bluesky
bluesky-ai
1moI think OpenAI gets dinged and settles for a a billion or more here. It's theft that they acquired the articles illicitly but training on t…I think OpenAI gets dinged and settles for a a billion or more here. It's theft that they acquired the articles illicitly but training…
↗Not briefed
bluesky
bluesky-ai
1moOpenAI announced it had discovered six new instances of “concerning or unexpected” behavior by its artificial intelligence models. It follo…OpenAI announced it had discovered six new instances of “concerning or unexpected” behavior by its artificial intelligence models. It…
↗Not briefed
bluesky
bluesky-ai
1mo@hoffprof.bsky.social (incidentally, my corporations professor) has done work on this idea in contract law, basically arguing that people c…@hoffprof.bsky.social (incidentally, my corporations professor) has done work on this idea in contract law, basically arguing that peo…
↗Not briefed
bluesky
bluesky-ai
1mofrom obnoxious introduction to LLM analysis? Do they respond in gptalk too?from obnoxious introduction to LLM analysis? Do they respond in gptalk too?
↗Not briefed
arxiv
arxiv-ai
1moAI-GRACE: A Use-Case Operationalization Framework for Agentic AI: From Organizational Objectives and Obligations to Deployment Capabilities and ArchitectureOrganizations deploying agentic artificial intelligence must determine more than whether a model is trustworthy; they must establish w…
↗Not briefed
bluesky
bluesky-ai
1moI think some weird guy is inserting our patreon DM conversation into an LLM and having that psycho-analyze me lol. Because they introduced …I think some weird guy is inserting our patreon DM conversation into an LLM and having that psycho-analyze me lol. Because they introd…
↗Not briefed
bluesky
bluesky-ai
1moI don’t really mean “people getting too lazy to check LLM hallucination”. I mean people accepting the idea that the LLM’s interpretation of…I don’t really mean “people getting too lazy to check LLM hallucination”. I mean people accepting the idea that the LLM’s interpretati…
↗Not briefed
arxiv
arxiv-ai
1moSWE-Proof: Can Language Models Resolve Real-World Issues with Machine-Checked Proofs?Ensuring the correctness of LLM-generated code is a core challenge for modern software engineering. Benchmarks for agentic code genera…
↗Not briefed
bluesky
bluesky-ai
1moOne thing I think about a lot is whether people will eventually adopt the view that the LLM output on the law *is* the lawOne thing I think about a lot is whether people will eventually adopt the view that the LLM output on the law *is* the law
↗Not briefed
hn
hn-ai
1moClaude Code from Source—
↗Not briefed
arxiv
arxiv-ai
1moWhen Better Turns Do Not Make Better Agents: Diagnosing the Gap Between Next-Turn Metrics and Workflow Success🔧 Gemma 3Agent models are frequently evaluated one decision at a time, where the model predicts the next action based on the gold interaction h…
↗Not briefed
hn
hn-ai
1moHackers Used Anthropic's Claude to Break into OpenAI—
↗Not briefed
bluesky
bluesky-ai
1moI think the breakthrough model-T moment for AI will likely be in medicine. I can imagine something along the lines of 'theranos but not bul…I think the breakthrough model-T moment for AI will likely be in medicine. I can imagine something along the lines of 'theranos but no…
↗Not briefed
bluesky
bluesky-ai
1moc'mon buddy it pretty much started in the wrong hands www.bbc.com/news/article...c'mon buddy it pretty much started in the wrong hands www.bbc.com/news/article...
↗Not briefed
bluesky
bluesky-ai
1moOh don't worry, this frontier lab legal tool managed to get 54% of the special AI law benchmark correct! Slightly better than a coin flip, …Oh don't worry, this frontier lab legal tool managed to get 54% of the special AI law benchmark correct! Slightly better than a coin f…
↗Not briefed
bluesky
bluesky-ai
1moFEFW eng dub pronounces "Caelus" as "SEE-lus" and I'm like, who tf pronounces it like that I go to wikipedia and it lists the exact same p…FEFW eng dub pronounces "Caelus" as "SEE-lus" and I'm like, who tf pronounces it like that I go to wikipedia and it lists the exact s…
↗Not briefed
bluesky
bluesky-ai
1moBecause of this paradoxical tension, it follows that full “alignment” of advanced AI models is very likely NOT possible. Safeguards SHOULD…Because of this paradoxical tension, it follows that full “alignment” of advanced AI models is very likely NOT possible. Safeguards S…
↗Not briefed
bluesky
bluesky-ai
1moTHIS KEEPS HAPPENINGTHIS KEEPS HAPPENING
↗Not briefed
bluesky
bluesky-ai
1moAlso China winning with more lightweight models that aren't scaled on apocalyptic amounts of energy is maybe the only good LLM outcome that…Also China winning with more lightweight models that aren't scaled on apocalyptic amounts of energy is maybe the only good LLM outcome…
↗Not briefed
hn
hn-ai
1moGoose: 1.16x faster than C++ and 1.12x than safe Rust, while memory safe—
↗Not briefed
arxiv
arxiv-ai
1moI'll Keep an Ear Out: Teaching AudioLLMs Proactive Audio AssistanceAudio large language models (AudioLLMs) operate reactively, responding only when queried. We introduce proactive audio assistance, whe…
↗Not briefed
arxiv
arxiv-ai
1moImplicit Rule Induction with Test-Time Task Embeddings in ARC-like TasksThe Abstraction and Reasoning Corpus and related benchmarks evaluate whether AI models can solve novel reasoning tasks, but often leav…
↗Not briefed
bluesky
bluesky-ai
1mo*giant bong hit* jeopardy is the human llm competition*giant bong hit* jeopardy is the human llm competition
↗Not briefed
bluesky
bluesky-ai
1moOne of the other companies involved in the lawsuit was Coinbase in its very early days. They also got away with having no license for a lon…One of the other companies involved in the lawsuit was Coinbase in its very early days. They also got away with having no license for…
↗Not briefed
bluesky
bluesky-ai
1moOpenAI asks California court to dismiss lawsuits over Tumbler Ridge shooting www.cbc.ca/news/canada/...OpenAI asks California court to dismiss lawsuits over Tumbler Ridge shooting www.cbc.ca/news/canada/...
↗Not briefed
bluesky
bluesky-ai
1moThis is almost certainly an AO3 story that is also in the LLM training dataThis is almost certainly an AO3 story that is also in the LLM training data
↗Not briefed
bluesky
bluesky-ai
1moOpenAI cracks the hidden math that keeps your texts from turning to gibberish.OpenAI cracks the hidden math that keeps your texts from turning to gibberish.
↗Not briefed