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
2momfers thinking an LLM knows whatever the hell this ismfers thinking an LLM knows whatever the hell this is
↗Not briefed
arxiv
arxiv-ai
2moCT-SAFR: Safe and Interpretable Chain-of-Thought Reasoning for Autonomous Robots: A Multi-Layered Verification Framework for Trustworthy AI-Driven Robotic Decision MakingChain-of-Thought (CoT) prompting enables LLMs to perform explicit, step-by-step reasoning, creating opportunities for sophisticated au…
↗Not briefed
bluesky
bluesky-ai
2moThe warnings come amid a push by some AI executives for a coordinated slowdown in AI development.The warnings come amid a push by some AI executives for a coordinated slowdown in AI development.
↗Not briefed
arxiv
arxiv-ai
2moLooped GPT-BERT: Trading Parameters for Computation in Small Language ModelingWhen training data are limited, increasing parameter count is not the only way to improve language-model performance. A small paramete…
↗Not briefed
bluesky
bluesky-ai
2mo虽然学术上的问题看起来还是罗生门,但威胁研究者除名为A\工作的合作者实在是让OpenAI很难看...虽然学术上的问题看起来还是罗生门,但威胁研究者除名为A\工作的合作者实在是让OpenAI很难看...
↗Not briefed
bluesky
bluesky-ai
2moMy guess is he’s leaving Anthropic to start an AI cybersecurity consultancy.My guess is he’s leaving Anthropic to start an AI cybersecurity consultancy.
↗Not briefed
bluesky
bluesky-ai
2moWe have to crush these sociopaths. You cannot trust any elected officials who take money from AI, Crypto and AIPAC in 2026. www.wsj.com/t…We have to crush these sociopaths. You cannot trust any elected officials who take money from AI, Crypto and AIPAC in 2026. www.wsj.…
↗Not briefed
hn
hn-ai
2moLIV Golf files for bankruptcy with stars owed millions—
↗Not briefed
hn
hn-ai
2moShow HN: GitHub Actions Aibom Generation—
↗Not briefed
bluesky
bluesky-ai
2moIs it possible that AI, and specifically LLMs, will allow improvements in drug discovery and testing? Yes. Is this it? I don’t think so.Is it possible that AI, and specifically LLMs, will allow improvements in drug discovery and testing? Yes. Is this it? I don’t think…
↗Not briefed
bluesky
bluesky-ai
2moLike, OpenAI and Anthropic aren't going to build an evil superintelligence, but there are many other destructive effects enabled by AI inve…Like, OpenAI and Anthropic aren't going to build an evil superintelligence, but there are many other destructive effects enabled by AI…
↗Not briefed
bluesky
bluesky-ai
2moThe lead researcher at Anthropic enthusiastically quote tweeting an agreement really does not help this argument! bsky.app/profile/phil...The lead researcher at Anthropic enthusiastically quote tweeting an agreement really does not help this argument! bsky.app/profile/ph…
↗Not briefed
hn
hn-ai
2moShow HN: Self-host open-source LLMs on AWS with scale-to-zero—
↗Not briefed
bluesky
bluesky-ai
2moDaVinci Resolve 21.1がClaudeやChatGPTなどのAIアシスタントとの連携をサポート https://gigazine.net/news/20260909-davinci-resolve-21-1-update/🔧 ChatGPTDaVinci Resolve 21.1がClaudeやChatGPTなどのAIアシスタントとの連携をサポート https://gigazine.net/news/20260909-davinci-resolve-21-1-update/
↗Not briefed
bluesky
bluesky-ai
2moOpenAIが「GPT-6 Astraより高性能なモデル」を用いて数学の難問「ナビエ・ストークス方程式」の解決を主張、一方で人間の数学者の成果を流用した疑惑も生じる https://gigazine.net/news/20260909-openai-navier-stokes/OpenAIが「GPT-6 Astraより高性能なモデル」を用いて数学の難問「ナビエ・ストークス方程式」の解決を主張、一方で人間の数学者の成果を流用した疑惑も生じる https://gigazine.net/news/20260909-openai-navier-st…
↗Not briefed
bluesky
bluesky-ai
2moThink about the stuff an "unintelligent" piece of code can do, and just imagine it being more efficient and having a little genetic algorit…Think about the stuff an "unintelligent" piece of code can do, and just imagine it being more efficient and having a little genetic al…
↗Not briefed
bluesky
bluesky-ai
2moAnthropic was created as a rebellion out of openai due to concerns that ai safety wasn't being taken seriously enough at oai. Dario recentl…Anthropic was created as a rebellion out of openai due to concerns that ai safety wasn't being taken seriously enough at oai. Dario re…
↗Not briefed
bluesky
bluesky-ai
2moi think the odds are about 2:1 against openai accomplishing anything noteworthy in the area of superconductivity over the next two weeks ju…i think the odds are about 2:1 against openai accomplishing anything noteworthy in the area of superconductivity over the next two wee…
↗Not briefed
bluesky
bluesky-ai
2moI resigned from Anthropic today. All the people building AI there today believe that there’s a greater than 10 percent chance that the AI t…I resigned from Anthropic today. All the people building AI there today believe that there’s a greater than 10 percent chance that the…
↗Not briefed
bluesky
bluesky-ai
2moUpdate on the Anthropic rumor situation:Update on the Anthropic rumor situation:
↗Not briefed
arxiv
arxiv-ai
2moALIGN-HOLD: Experience Alignment for Real-Time Hold Control in Large-Scale Ride-Hailing Matching at DiDiReal-time hold control is a high-leverage mechanism in large-scale ride-hailing systems: by selectively deferring driver-order pairs,…
↗Not briefed
arxiv
arxiv-ai
2moWhich Medical Questions Deserve Rationales? Perturbation-Sensitive Selection for Robust QAMedical question-answering datasets often contain answer labels, whereas high-quality rationales remain scarce, noisy, or costly to va…
↗Not briefed
bluesky
bluesky-ai
2moA Blunt Warning from an AI Researcher Who Just Quit Anthropic after Moving There from OpenAI revkin.substack.com/p/a-blunt-wa...A Blunt Warning from an AI Researcher Who Just Quit Anthropic after Moving There from OpenAI revkin.substack.com/p/a-blunt-wa...
↗Not briefed
bluesky
bluesky-ai
2moHey I was going to try to make sense of the proof via the Lean files. Any chance you plan on digging in too? github.com/openai/Navie...Hey I was going to try to make sense of the proof via the Lean files. Any chance you plan on digging in too? github.com/openai/Nav…
↗Not briefed
bluesky
bluesky-ai
2moi think there’s a social cache on this website for people who continue to post about how AI is just the latest Ape Yacht Club and i’m sorry…i think there’s a social cache on this website for people who continue to post about how AI is just the latest Ape Yacht Club and i’m…
↗Not briefed
arxiv
arxiv-ai
2moSettling: Equilibrium Inference for Non-Convex Validity SetsMany learning systems return a single point estimate even when admissible outputs form disconnected or non-convex sets. Under squared…
↗Not briefed
bluesky
bluesky-ai
2molike, anthropic and openAI probably paid this researcher more right out of college than they would’ve paid someone with two decades of expe…like, anthropic and openAI probably paid this researcher more right out of college than they would’ve paid someone with two decades of…
↗Not briefed
bluesky
bluesky-ai
2mox.com/kinocodesune... いやもう止まれよ…( ;∀;)x.com/kinocodesune... いやもう止まれよ…( ;∀;)
↗Not briefed
arxiv
arxiv-ai
2moSafe to Stop? Risk-Constrained Stopping for Sequential Clinical Diagnosis AgentsClinical diagnosis agents must decide not only what test to request next, but also when to diagnose or defer. Existing agent benchmark…
↗Not briefed
bluesky
bluesky-ai
2mo'ai' drug discovery is REAL!!! but it requires a lot of fucking work to get something that can actually evaluate it and do the computationa…'ai' drug discovery is REAL!!! but it requires a lot of fucking work to get something that can actually evaluate it and do the computa…
↗Not briefed
arxiv
arxiv-ai
2moX2-NativeCursor: Native-Token Text Progress Tracking for Incremental-Text Streaming Codec TTSIncremental-text streaming text-to-speech (TTS) needs online text progress tracking for synchronized highlighting, interruption handli…
↗Not briefed
arxiv
arxiv-ai
2moMuon-C: Operator-Aligned Muon for Convolutional KernelsMuon replaces matrix momentum with an approximately orthogonal polar direction, but its geometry depends on the matrix representation.…
↗Not briefed
hn
hn-ai
2moGetting 50 GB/S Back Out of the Neural Engine—
↗Not briefed
bluesky
bluesky-ai
2moI’d bet anything that guy who resigned from Anthropic got a bigger severance if he spouted their doomerist bullshitI’d bet anything that guy who resigned from Anthropic got a bigger severance if he spouted their doomerist bullshit
↗Not briefed
bluesky
bluesky-ai
2moThere is a difference between building and app with an LLM and hoping an LLM builds an appThere is a difference between building and app with an LLM and hoping an LLM builds an app
↗Not briefed
bluesky
bluesky-ai
2mothe takeaway is that OpenAI threw $10M-$20M to solve a problem to get community noted on their press release and add "While unlikely, we ca…the takeaway is that OpenAI threw $10M-$20M to solve a problem to get community noted on their press release and add "While unlikely,…
↗Not briefed
rss
techmeme
2moOver 10 US states have rolled back tech giants' tax breaks, which top $1B/year in some states, as many others propose similar bills amid a data center backlash (Wall Street Journal)Wall Street Journal: Over 10 US states have rolled back tech giants' tax breaks, which top $1B/year in some states, as many others pro…
↗Not briefed
arxiv
arxiv-ai
2moSEA-SpeechBench: A Large-Scale Multitask Benchmark for Speech Understanding Across Southeast AsiaThe rapid advancement of audio and multimodal large language models has unlocked transformative speech understanding capabilities, yet…
↗Not briefed
arxiv
arxiv-ai
2moIntroducing Consort: A Spec-First Agent Framework for Enforced, Test-Driven Development on Live Database BranchesWhen an agent writes code, the development framework becomes the control system for a non-deterministic worker. Spec-first, agent-driv…
↗Not briefed
bluesky
bluesky-ai
2moSomehow no one at Anthropic or OpenAI knows how to unplug an ethernet cable or remove a wifi chipSomehow no one at Anthropic or OpenAI knows how to unplug an ethernet cable or remove a wifi chip
↗Not briefed
bluesky
bluesky-ai
2moi actually don’t think this but i do think this guy is going to pop up in an anti-LLM think tank or a “safe AI” company sooner rather than …i actually don’t think this but i do think this guy is going to pop up in an anti-LLM think tank or a “safe AI” company sooner rather…
↗Not briefed
bluesky
bluesky-ai
2mothis will work because whenever someone shows an llm my writing it falls in love with methis will work because whenever someone shows an llm my writing it falls in love with me
↗Not briefed
arxiv
arxiv-ai
2moPRAGMA: Evaluating Personalized Guidance with Memory Alignment in Lifelong ConversationsLarge language models (LLMs) are increasingly deployed as personalized assistants that interact with users over extended periods of ti…
↗Not briefed
bluesky
bluesky-ai
2mo文章を「拡散」で高速生成するAIモデル「Mercury 2.5」登場、毎秒1107トークンの高速生成&GPT-5.6 Luna級の性能を100万出力トークン当たり約120円で提供 https://gigazine.net/news/20260909-mercury-2-5/🔧 GPT-5文章を「拡散」で高速生成するAIモデル「Mercury 2.5」登場、毎秒1107トークンの高速生成&GPT-5.6 Luna級の性能を100万出力トークン当たり約120円で提供 https://gigazine.net/news/20260909-mercury-2…
↗Not briefed
bluesky
bluesky-ai
2moIt doesn't even have to be regurgitated here. The human work pointed the way. OpenAI likely wouldn't have been able to do it without that, …It doesn't even have to be regurgitated here. The human work pointed the way. OpenAI likely wouldn't have been able to do it without t…
↗Not briefed
bluesky
bluesky-ai
2moone reason this is stupid is that 1) that isn't the only "true" (???) use of it, and 2) no they aren't, may I introduce you to any number o…one reason this is stupid is that 1) that isn't the only "true" (???) use of it, and 2) no they aren't, may I introduce you to any num…
↗Not briefed
bluesky
bluesky-ai
2moA buddy of mine who was doing AI research at Google at a more senior level quit because he finds the LLM math pretty boring, I don’t think …A buddy of mine who was doing AI research at Google at a more senior level quit because he finds the LLM math pretty boring, I don’t t…
↗Not briefed
bluesky
bluesky-ai
2moAI right now is so fuckin strange - it's both a fucking genius and yet does the dumbest goddamn shit like no claude, you can't just call a…AI right now is so fuckin strange - it's both a fucking genius and yet does the dumbest goddamn shit like no claude, you can't just c…
↗Not briefed
arxiv
arxiv-ai
2moPELM: Power Efficient On-Device LLM Inference with Speculative Decoding and Dynamic Voltage Frequency ScalingDeploying Large Language Models (LLMs) directly on mobile platforms at the edge is gaining traction due to a myriad of benefits, such…
↗Not briefed
bluesky
bluesky-ai
2moHindustan Times : FREE Anthropic researcher quits over 'out of control' AI fears, says 'AI could kill us all by end of decade'“ At OpenAI,…Hindustan Times : FREE Anthropic researcher quits over 'out of control' AI fears, says 'AI could kill us all by end of decade'“ At Op…
↗Not briefed