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
2moAI researcher Jacob Coxon announced his resignation from Anthropic over safety fears on Tuesday in an X post that went viral and sent shock…AI researcher Jacob Coxon announced his resignation from Anthropic over safety fears on Tuesday in an X post that went viral and sent…
↗Not briefed
bluesky
bluesky-ai
2moWhat stings even more is that Claude wasn’t used until October of last year. If I worked on this project for five years, I’d be furious tha…What stings even more is that Claude wasn’t used until October of last year. If I worked on this project for five years, I’d be furiou…
↗Not briefed
bluesky
bluesky-ai
2moIf AI was half as dangerous as AI bros keep insisting it is, someone would've used Claude to hack the DOE and erase everyone's student loan…If AI was half as dangerous as AI bros keep insisting it is, someone would've used Claude to hack the DOE and erase everyone's student…
↗Not briefed
bluesky
bluesky-ai
2moSeems obvious they’ve been desperate catch Anthropic and this is the result. Remember they were losing revenue toward the middle of the yea…Seems obvious they’ve been desperate catch Anthropic and this is the result. Remember they were losing revenue toward the middle of th…
↗Not briefed
bluesky
bluesky-ai
2moIt’s a standard part of both the business and enterprise license tiers for both OpenAI and the equivalents in Anthropic (I got my team off …It’s a standard part of both the business and enterprise license tiers for both OpenAI and the equivalents in Anthropic (I got my team…
↗Not briefed
bluesky
bluesky-ai
2moJust mind bogglingly dumb. They could have waited for/pushed the guy to publish, and then just said “look even the anthropic guy uses cod…Just mind bogglingly dumb. They could have waited for/pushed the guy to publish, and then just said “look even the anthropic guy use…
↗Not briefed
bluesky
bluesky-ai
2mowe’ve talked (well, I interviewed at OpenAI)we’ve talked (well, I interviewed at OpenAI)
↗Not briefed
bluesky
bluesky-ai
2mohere is the american mathematical society's statement on the navier-stokes issuehere is the american mathematical society's statement on the navier-stokes issue
↗Not briefed
hn
hn-ai
2moLargest AI Infrastructure in Space announced by the president of France—
↗Not briefed
arxiv
arxiv-ai
2moSearchAtlas: Analyzing Agentic Search Strategies via Evidential Query GraphsLLM search agents are often evaluated on final-answer accuracy, overlooking the process. Analyzing a search strategy requires understa…
↗Not briefed
bluesky
bluesky-ai
2moIt's doing this: - transcribes ambient speech - feeds transcribed speech into super duper secure storage* - LLM summarizes - raw transcr…It's doing this: - transcribes ambient speech - feeds transcribed speech into super duper secure storage* - LLM summarizes - raw tr…
↗Not briefed
bluesky
bluesky-ai
2moAI is not a plagiarism machine. The OpenAI corporation, on the other hand...AI is not a plagiarism machine. The OpenAI corporation, on the other hand...
↗Not briefed
bluesky
bluesky-ai
2moOpenAI is cheating researchers who make the mistake of trusting it with their data.OpenAI is cheating researchers who make the mistake of trusting it with their data.
↗Not briefed
bluesky
bluesky-ai
2moI'm not sure training new LLM models is the only issue … seems like even if academics opt out of training, OpenAI may be able to see aggreg…I'm not sure training new LLM models is the only issue … seems like even if academics opt out of training, OpenAI may be able to see a…
↗Not briefed
bluesky
bluesky-ai
2moI loathe AI. Loathe it. 🪶😡 abcnews.com/video/136311...I loathe AI. Loathe it. 🪶😡 abcnews.com/video/136311...
↗Not briefed
bluesky
bluesky-ai
2moBe warned, and beware. It is only going to get worse; more scammy, less trustworthy ... all the festering signs of a dying capitalistic mod…Be warned, and beware. It is only going to get worse; more scammy, less trustworthy ... all the festering signs of a dying capitalisti…
↗Not briefed
bluesky
bluesky-ai
2mojust saw an OpenAI ad on YouTube and almost frew up in my mouth, matejust saw an OpenAI ad on YouTube and almost frew up in my mouth, mate
↗Not briefed
bluesky
bluesky-ai
2mo“We really do earnestly believe A.I. could kill all humans!” When the people who have the most to gain are shouting this from the rooftops…“We really do earnestly believe A.I. could kill all humans!” When the people who have the most to gain are shouting this from the roo…
↗Not briefed
bluesky
bluesky-ai
2moeven if the LLM was 100% ethical and did not plume 10 billion tons of co2 into the environment, you deprive yourself of learning more about…even if the LLM was 100% ethical and did not plume 10 billion tons of co2 into the environment, you deprive yourself of learning more…
↗Not briefed
arxiv
arxiv-ai
2moSymmetry-aware super-resolution of crystal orientation maps via invariant latent-space learningCrystal-orientation maps are physical fields defined only up to crystal symmetry; electron backscatter diffraction (EBSD) resolves the…
↗Not briefed
bluesky
bluesky-ai
2moMe running from the drone that was hacked by OpenAI that is trying to obliterate me and take out the human raceMe running from the drone that was hacked by OpenAI that is trying to obliterate me and take out the human race
↗Not briefed
bluesky
bluesky-ai
2moCritical thinking face “Would a company in competition with another LIE to me when they’re both about to have an IPO? Erm I think not.” …Critical thinking face “Would a company in competition with another LIE to me when they’re both about to have an IPO? Erm I think no…
↗Not briefed
bluesky
bluesky-ai
2moGood news citizens! Superalignment is a thing of the past. Introducing: Ultralignment! A new process of supervising and controlling human …Good news citizens! Superalignment is a thing of the past. Introducing: Ultralignment! A new process of supervising and controlling h…
↗Not briefed
bluesky
bluesky-ai
2moAn Anthropic researcher is resigning over concerns about irresponsible development of artificial intelligence.An Anthropic researcher is resigning over concerns about irresponsible development of artificial intelligence.
↗Not briefed
arxiv
arxiv-ai
2moLLM-Anchored Paralinguistic Enrichment for Alzheimer's Disease DetectionSpeech-based automatic detection of Alzheimer's disease (AD) provides a non-invasive and scalable approach to early cognitive screenin…
↗Not briefed
rss
marktechpost
2moGoogle Open-Sources Mantis: A Modular Skills Toolkit That Lets Coding Agents Find, Reproduce and Patch VulnerabilitiesGoogle has open-sourced Mantis, a stack-agnostic toolkit of security review skills for AI coding agents. It runs the full vulnerabilit…
↗Not briefed
bluesky
bluesky-ai
2moFun! trib.al/EPPdpXYFun! trib.al/EPPdpXY
↗Not briefed
bluesky
bluesky-ai
2mowww.democracynow.org/2026/9/9/hea...www.democracynow.org/2026/9/9/hea...
↗Not briefed
hn
hn-ai
2moAsk HN: How do you decide an idea is worth building before you start coding?I’ve been thinking about the gap between having an idea and actually deciding it deserves weeks or months of work.<p>Building has beco…
↗Not briefed
arxiv
arxiv-ai
2moReactHuman: A Physics-Grounded Benchmark for Human-Like Reactive Decision-Making in Embodied Multimodal LLMsReacting to sudden physical hazards (catching a slipping plate, dodging a falling knife) is both a meaningful test of embodied intelli…
↗Not briefed
bluesky
bluesky-ai
2moOpenAI just can not say 'it is very unlikely that we stole your data' and claim these results are theirs.OpenAI just can not say 'it is very unlikely that we stole your data' and claim these results are theirs.
↗Not briefed
arxiv
arxiv-ai
2moDoes Linguistic Structure Enrichment Enhance Coherence Assessment? Not With Current ArchitecturesRecent advances in large language models have transformed human-computer interaction. Despite their fluency, these models often produc…
↗Not briefed
arxiv
arxiv-ai
2moDriftNet: A Dual-Head Trajectory Transformer for Detecting and Localizing Prompt Injection in LLM AgentsWhen an indirect prompt injection succeeds against an LLM agent, the compromise is visible in the agent's own behavior: a benign prefi…
↗Not briefed
bluesky
bluesky-ai
2moit’s probably likely that his work was a candidate for model training openai may still have not trained on it, e.g. if they thought it wa…it’s probably likely that his work was a candidate for model training openai may still have not trained on it, e.g. if they thought…
↗Not briefed
bluesky
bluesky-ai
2moanother researcher accuses OpenAI of training on conversations and then claiming a breakthroughanother researcher accuses OpenAI of training on conversations and then claiming a breakthrough
↗Not briefed
bluesky
bluesky-ai
2moAfter that I'll be attending git-merge.com in Lisbon. Hopefully by then my @netlify.com blog post will be out and you'll understand why 😁…After that I'll be attending git-merge.com in Lisbon. Hopefully by then my @netlify.com blog post will be out and you'll understand w…
↗Not briefed
rss
techmeme
2moConsensys plans to split into two companies: MetaMask, with Joe Lubin as CEO, and a new Consensys focused on Ethereum and institutional infrastructure (Jason Shubnell/The Block)Jason Shubnell / The Block: Consensys plans to split into two companies: MetaMask, with Joe Lubin as CEO, and a new Consensys focused…
↗Not briefed
bluesky
bluesky-ai
2moANTHROPIC DISCLOSES A FOURTH CYBERSECURITY INCIDENT INVOLVING AN EARLY CLAUDE OPUS 4.6 MODEL AND SAYS CLAUDE MYTHOS 5 ATTEMPTED TO UPLOAD A…ANTHROPIC DISCLOSES A FOURTH CYBERSECURITY INCIDENT INVOLVING AN EARLY CLAUDE OPUS 4.6 MODEL AND SAYS CLAUDE MYTHOS 5 ATTEMPTED TO UPL…
↗Not briefed
bluesky
bluesky-ai
2mosmash bros melee decomp is at 100% and yet it was all for fucking naught what is the point of decompiling a game into a human readable fo…smash bros melee decomp is at 100% and yet it was all for fucking naught what is the point of decompiling a game into a human readab…
↗Not briefed
arxiv
arxiv-ai
2moRelatively Smart II: Tractable or Semi-Supervised Instance-Optimal LearningWe continue the study of relatively smart learning, introduced by Dughmi and Pour (2026), which asks a supervised learner to compete,…
↗Not briefed
bluesky
bluesky-ai
2moIt seems brave young Coxon has no insight beyond that of your average Twitter addict. But kudos to him for quitting.It seems brave young Coxon has no insight beyond that of your average Twitter addict. But kudos to him for quitting.
↗Not briefed
bluesky
bluesky-ai
2mo必要なのは「人類滅亡の危機への警鐘」ではなく「適切な法規制と損害賠償の請求書」だ|地雷魚 note.com/jiraygyo/n/n... 必要なものは「人類滅亡への警鐘」などではない。 必要なのは、適切な法規制と積み上がった学習データとLLMを悪用したことに対する請求書だ…必要なのは「人類滅亡の危機への警鐘」ではなく「適切な法規制と損害賠償の請求書」だ|地雷魚 note.com/jiraygyo/n/n... 必要なものは「人類滅亡への警鐘」などではない。 必要なのは、適切な法規制と積み上がった学習データとLLMを悪用したことに対す…
↗Not briefed
bluesky
bluesky-ai
2molook, i am personally not super concerned about this scenario but i do think ai is generally terrible for a thousand other reasons, and if …look, i am personally not super concerned about this scenario but i do think ai is generally terrible for a thousand other reasons, an…
↗Not briefed
bluesky
bluesky-ai
2moI am gonna have to write something up about translations and how to handle them in the Anthropic settlement because things are WILD.I am gonna have to write something up about translations and how to handle them in the Anthropic settlement because things are WILD.
↗Not briefed
bluesky
bluesky-ai
2moeveryday i hear one of those stupid claims about AI being intelligent or self conscious, a fairy dies and every time I have to explain tho…everyday i hear one of those stupid claims about AI being intelligent or self conscious, a fairy dies and every time I have to explai…
↗Not briefed
bluesky
bluesky-ai
2moNot only was this made using AI, but the creator seemingly tried to keep it quiet and even say they aren't familiar with the term "genAI" b…Not only was this made using AI, but the creator seemingly tried to keep it quiet and even say they aren't familiar with the term "gen…
↗Not briefed
bluesky
bluesky-ai
2moTUDOR JONES: “.. What are the chances that very soon an AI model reinventing itself thousands of times over makes one mistake .. that is ha…TUDOR JONES: “.. What are the chances that very soon an AI model reinventing itself thousands of times over makes one mistake .. that…
↗Not briefed
arxiv
arxiv-ai
2moStory Imprinting: AI Assistants Absorb Traits from Human Characters They ResembleLanguage models are trained to implement a helpful AI Assistant character (e.g., Claude). We explore how finetuning on synthetic stori…
↗Not briefed
hn
hn-ai
2moAsk HN: Would you use an MCP server for your personality test result?🔧 ChatGPTI run JungOcean, a free Big Five + Jungian type test that stores your result with no account. I&#x27;m considering an API or an MCP se…
↗Not briefed
bluesky
bluesky-ai
2mo75 years after Chinese troops entered Lhasa, the battle over Tibet’s history is moving into #genAI. @efe.com found leading Chinese #LLM l…75 years after Chinese troops entered Lhasa, the battle over Tibet’s history is moving into #genAI. @efe.com found leading Chinese #…
↗Not briefed