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
4dI've been looking at this all day trying to see the part where it explains why detectors cannot work (They can, because it's a stylometric …I've been looking at this all day trying to see the part where it explains why detectors cannot work (They can, because it's a stylome…
↗Not briefed
arxiv
arxiv-ai
4dEfficient Modeling of Multi-Agent Systems via Structure-Aware Physics-Informed LearningModeling large multi-agent systems poses challenges for trajectory inference and parameter identification, as network size and couplin…
↗Not briefed
hn
hn-ai
4dLogitech Harmony IR code online database archive—
↗Not briefed
rss
the-decoder
4dAlibaba's Qwen4-Max takes on Opus 5.5, GPT-6, and Kimi K3.1 Pro at much lower pricesAlibaba launches Qwen4-Max, a new flagship model with 2.4 trillion parameters that activates only 71 billion per token. A hybrid archi…
↗Not briefed
rss
techmeme
4dNoeri, which is developing AI software to review and suggest improvements for its clients' AI models and tools, raised $110M led by Menlo Ventures (Dina Bass/Bloomberg)Dina Bass / Bloomberg: Noeri, which is developing AI software to review and suggest improvements for its clients' AI models and tools,…
↗Not briefed
bluesky
bluesky-ai
4dAnthropic is going public this week with a $1 trillion plus valuation after stealing all our work to train its plagiarism machine btw No o…Anthropic is going public this week with a $1 trillion plus valuation after stealing all our work to train its plagiarism machine btw…
↗Not briefed
bluesky
bluesky-ai
4dWhen Pennsylvania became the first state to run a pilot program with OpenAI, SEIU local 668 fought back—and won. Now PA has a worker board…When Pennsylvania became the first state to run a pilot program with OpenAI, SEIU local 668 fought back—and won. Now PA has a worker…
↗Not briefed
arxiv
arxiv-ai
4dMegNIST, A Benchmark for Non-Invasive Inner Speech DecodingMegNIST is a benchmark dataset for developing and evaluating machine learning methods for inner speech decoding from magnetoencephalog…
↗Not briefed
bluesky
bluesky-ai
4dPlus LangChain, from LLM glue code to enterprise agent management. One founder got hired after a single night of cramming. Full episode and…🔧 LangChainPlus LangChain, from LLM glue code to enterprise agent management. One founder got hired after a single night of cramming. Full episod…
↗Not briefed
arxiv
arxiv-ai
4dSelf-attention is a Dyson equation for many-body machine learningWhy does symmetric sum-pooling (DeepSets) predict ground-state potentials to sub-meV accuracy, yet fail for correlated wavefunctions a…
↗Not briefed
bluesky
bluesky-ai
4dOpenAI canceled their robotics program just 5 months in. Robotics + AI is a waaaaaaay harder problem than a LLM. Thats also how you know we…OpenAI canceled their robotics program just 5 months in. Robotics + AI is a waaaaaaay harder problem than a LLM. Thats also how you kn…
↗Not briefed
bluesky
bluesky-ai
4dWell, my portion of the Anthropic settlement was paid in a timely manner, by today's due date, a cool $200,000, accompanied by a handwritte…Well, my portion of the Anthropic settlement was paid in a timely manner, by today's due date, a cool $200,000, accompanied by a handw…
↗Not briefed
arxiv
arxiv-ai
4dBeyond Bilingual: Studying Quadrilingual Code-Switching with Conversational AIMultilingual speakers choose languages by topic and conversation partner. Large language models (LLMs) support these languages unequal…
↗Not briefed
arxiv
arxiv-ai
4dTrain After You Measure: a Pre-Registered Learning Curve for Fine-Tuning Typed Decision Models on One User's LabelsFine-tuning a small typed decision model on one user's email labels did not clear the margin we pre-registered over a logistic regress…
↗Not briefed
hn
hn-ai
4dShow HN: Near-Realtime LLM Steering—
↗Not briefed
arxiv
arxiv-ai
4dStability-Compensated Threshold Calibration for Selection-Conditioned Risk Control in Large Language ModelsLarge language models often produce hallucinated answers, motivating selective prediction policies that abstain on uncertain queries.…
↗Not briefed
bluesky
bluesky-ai
4dIt really is like Twitter over here. We even got moderators that hate trans people, won't ban obvious bigotry, and keeps around LLM argue b…It really is like Twitter over here. We even got moderators that hate trans people, won't ban obvious bigotry, and keeps around LLM ar…
↗Not briefed
rss
techmeme
4dHarvard study: just 403 US data centers operational between May 2024 and April 2025 consumed as much electricity annually as the state of Michigan (Molly Taft/Wired)Molly Taft / Wired: Harvard study: just 403 US data centers operational between May 2024 and April 2025 consumed as much electricity a…
↗Not briefed
bluesky
bluesky-ai
4dLOL this was just this year cybernews.com/ai-news/clau...LOL this was just this year cybernews.com/ai-news/clau...
↗Not briefed
arxiv
arxiv-ai
4dThe Mask Is Not the Object: Volumetric Supervision for 3D Gaussian Segmentation3D Gaussian Splatting enables object-level operations on Gaussian primitives, yet existing segmentation methods are largely trained an…
↗Not briefed
arxiv
arxiv-ai
4dVisualArtifactSpec 1.0: Evaluating Visual Artifacts under Task SpecificationsVisual artifact tasks test agents' ability to translate task specifications and source materials into outputs that meet practical need…
↗Not briefed
bluesky
bluesky-ai
4dShould have been the big publishers suing all the 'AI' companies into oblivion for breaching the copyrights they are supposed to protect. I…Should have been the big publishers suing all the 'AI' companies into oblivion for breaching the copyrights they are supposed to prote…
↗Not briefed
hn
hn-ai
4dPgvectorfast: Speeding up pgvector with a fast buffer pool—
↗Not briefed
bluesky
bluesky-ai
4dAnthropic: $64 billion IPO, writers suffering the torments of the damned trying to get paid a pittance for stolen work @flaminghydra.com: …Anthropic: $64 billion IPO, writers suffering the torments of the damned trying to get paid a pittance for stolen work @flaminghydra.…
↗Not briefed
bluesky
bluesky-ai
4dAnthropic has commitments to data centres ("infrastructure") of nearly one trillion dollars, re. updated pre-IPO filing (Q3 figures were ha…Anthropic has commitments to data centres ("infrastructure") of nearly one trillion dollars, re. updated pre-IPO filing (Q3 figures we…
↗Not briefed
hn
hn-ai
4dSlate Screenless GPS—
↗Not briefed
bluesky
bluesky-ai
4dgenuinely I’d rather have someone use an LLM to write code and then document it themselves vs the other way aroundgenuinely I’d rather have someone use an LLM to write code and then document it themselves vs the other way around
↗Not briefed
bluesky
bluesky-ai
4dI looked into but didn’t pursue the Anthropic settlement; it felt like payouts would be inadequate and tenuous. And… yep. I also had one o…I looked into but didn’t pursue the Anthropic settlement; it felt like payouts would be inadequate and tenuous. And… yep. I also had…
↗Not briefed
bluesky
bluesky-ai
4dI am now being served Anthropic IPO ads every two minutes in Finnish. In the pine forests of Central Finland. There is no escape. YOUR CAPI…I am now being served Anthropic IPO ads every two minutes in Finnish. In the pine forests of Central Finland. There is no escape. YOUR…
↗Not briefed
arxiv
arxiv-ai
4dA Crowd-Audited Atlas of Implicit Bias in LLM-Powered AgentsPrior work has shown that large language models (LLMs) express implicit biases in high-publicity, one-off decision scenarios such as h…
↗Not briefed
arxiv
arxiv-ai
4dProceedings Eighth International Workshop on Formal Methods for Autonomous SystemsThis EPTCS volume contains the papers from the Eighth International Workshop on Formal Methods for Autonomous Systems (FMAS 2026), whi…
↗Not briefed
arxiv
arxiv-ai
4dWeVisForensics: Self-Evolving Agent for Omni AI-Generated Video DetectionRecent advances in video generation call for reliable detection grounded in explicit forensic evidence. However, current methods prima…
↗Not briefed
arxiv
arxiv-ai
4dLearning What Pairs Miss for Higher-Order Multimodal AlignmentContrastive learning beyond two modalities now scores the whole tuple with multilinear products, volumes, and fusion heads, critics th…
↗Not briefed
bluesky
bluesky-ai
4dHere's how anyone can easily use private, open weight models micahflee.com/heres-how-an...Here's how anyone can easily use private, open weight models micahflee.com/heres-how-an...
↗Not briefed
arxiv
arxiv-ai
4dFederated TD Learning in Heterogeneous Environments with Average Rewards: A Two-timescale Approach with Polyak-Ruppert AveragingFederated Reinforcement Learning (FRL) provides a promising way to speedup training in reinforcement learning when one has access to m…
↗Not briefed
hn
hn-ai
4dShow HN: A linter and GitHub Action for CLAUDE.md, AGENTS.md and SKILL.md—
↗Not briefed
arxiv
arxiv-ai
4dPremise Recovery in Software Design: What Does This Decision Rest On?Design decisions shape software architecture, and writing them down preserves the reasons behind it. Architecture decision records (AD…
↗Not briefed
hn
hn-ai
4dPrefer synthetic traffic over LLM as a judge—
↗Not briefed
arxiv
arxiv-ai
4dWhat You Post Is What You Symptom: Estimating Psychometric Networks from Social Media Discourse🔧 ReplicateSymptom networks in psychopathology are estimated almost exclusively from self-report questionnaires, restricting inference to whateve…
↗Not briefed
bluesky
bluesky-ai
4dMooi schrikbarend uitgelegd door Jonathan Witteman, met dank aan gore metaforen: www.volkskrant.nl/economie/ant...Mooi schrikbarend uitgelegd door Jonathan Witteman, met dank aan gore metaforen: www.volkskrant.nl/economie/ant...
↗Not briefed
bluesky
bluesky-ai
4dMy search consulting moved away from LLM judges. Now I: * Vibecode a facsimile of my clients application * Let agents act as users to use …My search consulting moved away from LLM judges. Now I: * Vibecode a facsimile of my clients application * Let agents act as users to…
↗Not briefed
rss
gary-marcus
4dGame over, man?Looks like the bubble has finally popped
↗Not briefed
arxiv
arxiv-ai
4dSpecificity-Guided Knowledge Diversion for Decomposable Generative ModelsGenerative model pre-training typically optimizes a model as a whole without explicitly organizing its parameters for selective reuse.…
↗Not briefed
hn
hn-ai
4dSwiftWin – build and run SwiftUI iPhone apps from Windows, no Mac—
↗Not briefed
arxiv
arxiv-ai
4dRepresentational Change Under Self-Consuming Training: A Sparse Autoencoder AnalysisLarge language models are increasingly exposed to model-generated text, raising the possibility that future models will be trained rec…
↗Not briefed
hn
hn-ai
4dProject Gutenberg temporarily down for maintenancehttps:&#x2F;&#x2F;mastodon.social&#x2F;@gutenberg_new&#x2F;117624704231883347Project<p>Gutenberg is currently running on a backup data…
↗Not briefed
hn
hn-ai
4dShow HN: DockShip42, small browser game to land StarshipFun 15 minute project with Claude code
↗Not briefed
bluesky
bluesky-ai
4dWhat if I told you that accusing Black creators of being AI, developing AI personal assistants that look like Black people, & using DEI as …What if I told you that accusing Black creators of being AI, developing AI personal assistants that look like Black people, & using DE…
↗Not briefed
hn
hn-ai
4dShow HN: Meridian, a network engineer's Swiss army knife—
↗Not briefed
bluesky
bluesky-ai
4dShuxian Liu and David Gray Widder at the University of Texas at Austin compared how the US government, Chinese state media and online publi…Shuxian Liu and David Gray Widder at the University of Texas at Austin compared how the US government, Chinese state media and online…
↗Not briefed