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
hn
hn-ai
12dShow HN: Bottled Reset Rune – auto-use spare Codex resets before they expire—
↗Not briefed
bluesky
bluesky-ai
12dOpenAI control room can stop model output in 35 secondsOpenAI control room can stop model output in 35 seconds
↗Not briefed
hn
hn-ai
12dAnthropic Prepares to Push Ahead with IPO Despite Market Jitters—
↗Not briefed
bluesky
bluesky-ai
12dBREAKING: Anthropic is planning to continue its IPO later this month 'despite recent market volatility', according to Wall Street Journal r…BREAKING: Anthropic is planning to continue its IPO later this month 'despite recent market volatility', according to Wall Street Jour…
↗Not briefed
hn
hn-ai
12dUSB: A trust layer for AI skills and MCP servers—
↗Not briefed
arxiv
arxiv-ai
12dAn Analytical Model of Compute-limited Multistage Training PipelinesModern foundation models are trained through multiple stages, yet how these stages interact and how compute should be allocated across…
↗Not briefed
arxiv
arxiv-ai
12dOff-policy Learning with Excursion PoliciesOff-policy reinforcement learning (RL) tries to learn the value of a target policy while data comes from a different behavior policy.…
↗Not briefed
hn
hn-ai
12dHow well do current Claude/GPT models play multiplayer Magic the Gathering?—
↗Not briefed
hn
hn-ai
12dShow HN: Agent-email-guard – agent tried to send 400 receipts, stopped at 5—
↗Not briefed
hn
hn-ai
12dShow HN: Emma, a real-time AI video persona from Nuance LabsHi HN! I'm Claudia, first employee and founding researcher at Nuance Labs. We're a research startup based in Seattle and hav…
↗Not briefed
arxiv
arxiv-ai
12dWhat drives the emergence of symbolic mechanisms in large language models?Recent work has identified a set of emergent symbolic mechanisms that support abstract reasoning in large language models, but it rema…
↗Not briefed
bluesky
bluesky-ai
12dBehind-the-meter gas power plants are permitted at the state level, and BAAJA does not intervene in that. Nor does the bill mandate that …Behind-the-meter gas power plants are permitted at the state level, and BAAJA does not intervene in that. Nor does the bill mandate…
↗Not briefed
arxiv
arxiv-ai
12dAdaptive Residual Quantization for Memory-Efficient Temporal Action SegmentationTemporal action segmentation (TAS) is commonly trained on large banks of pre-extracted frame features, but storing and replaying those…
↗Not briefed
bluesky
bluesky-ai
12d#ACR26 Monday PM orals - Methodology I wanted to comment on the 3 orals about LLM/AI#ACR26 Monday PM orals - Methodology I wanted to comment on the 3 orals about LLM/AI
↗Not briefed
arxiv
arxiv-ai
12dWhich Acoustic Features Survive Compression? Tracker-Estimated First-Formant Statistics for Codec-Robust Audio DeepfakeDetectionAudio deepfake detectors that work well on clean lab recordings often degrade after telephone or platform compression. This paper asks…
↗Not briefed
bluesky
bluesky-ai
12dIt's no coincidence that the structure of LLM text is shaped like my bowel movementsIt's no coincidence that the structure of LLM text is shaped like my bowel movements
↗Not briefed
hn
hn-ai
12dTell HN: Some of My Recent IdeasI thought it'll be fun to share some different ideas that I've had AI build in the last couple of weeks or months. Using my…
↗Not briefed
arxiv
arxiv-ai
12dDependable Long-Horizon LLM Agents: A Trajectory-Centered Empirical Study of Where Errors Enter and Which Controls Reach ThemWhen an LLM agent fails part-way through a long task, the usual analysis records a symptom: a failure-mode label, or the agent and ste…
↗Not briefed
hn
hn-ai
12dSoftware squeezes more data out of Webb Telescope for imaging exoplanets—
↗Not briefed
arxiv
arxiv-ai
12dCurie's Jarvis: Autonomous Robotic Chemistry Experiments from Printed ProtocolsA person running a bench experiment works from a printed instruction sheet, tells identical containers apart by the labels written on…
↗Not briefed
arxiv
arxiv-ai
12dDomain-Routed Delta Optimization for Multi-Teacher On-Policy DistillationMulti-teacher on-policy distillation assigns each sample a domain teacher but typically updates shared student parameters. We introduc…
↗Not briefed
arxiv
arxiv-ai
12dExpertCrowd-SWE: Measuring Frontier Coding Agents on Expert-Authored Engineering Tasks in Unfamiliar Codebases🔧 Grok 4🔧 Qwen 3ExpertCrowd-SWE (EC-SWE) is a benchmark of 100 expert-authored, multi-file software engineering tasks for evaluating autonomous coding…
↗Not briefed
bluesky
bluesky-ai
12dIm very excited for more labs to compete with Anthropic and distribute AI research abilties, knowing the safety community and others will b…Im very excited for more labs to compete with Anthropic and distribute AI research abilties, knowing the safety community and others w…
↗Not briefed
hn
hn-ai
12dLorem ipsum's fake history is one Latin professor's honest mistake [video]—
↗Not briefed
bluesky
bluesky-ai
12dExcept for the fact that data centers are building gas plants to generate electricity, not solar farms. What about the water supply? Water …Except for the fact that data centers are building gas plants to generate electricity, not solar farms. What about the water supply? W…
↗Not briefed
bluesky
bluesky-ai
12dExactly. Call your senators in force. Demand no on Schumer. How many progressive campaigns did he sit on his dick and do nothing? Mamdani? …Exactly. Call your senators in force. Demand no on Schumer. How many progressive campaigns did he sit on his dick and do nothing? Mamd…
↗Not briefed
bluesky
bluesky-ai
12dAI infrastructure provider Fireworks says hackers accessed some customers' datasets, incomplete model weights, and billing information over…AI infrastructure provider Fireworks says hackers accessed some customers' datasets, incomplete model weights, and billing information…
↗Not briefed
rss
the-decoder
12dClaude Fable 5.2 and Mythos 5.2 cement Anthropic's AI lead just before its IPOAnthropic is releasing Claude Fable 5.2 and Mythos 5.2, its most powerful models to date, in two versions: one freely available and on…
↗Not briefed
rss
techmeme
12dAI infrastructure provider Fireworks says hackers accessed some customers' datasets, incomplete model weights, and billing information over the past month (Anna Tong/Forbes)Anna Tong / Forbes: AI infrastructure provider Fireworks says hackers accessed some customers' datasets, incomplete model weights, and…
↗Not briefed
arxiv
arxiv-ai
12dGigaEdge Spench: An SDK for Practical Streaming Speech EnhancementWe present GigaEdge Spench, an open-source SDK for streaming speech enhancement on CPU servers, mobile devices, and in browsers. Appli…
↗Not briefed
hn
hn-ai
12dLumnik – read legacy ERPs and exports without touching them—
↗Not briefed
arxiv
arxiv-ai
12dSTAR: Semantic-Criticality Tiered Allocation of Radio Resources via Zero-Shot LearningIn remote inference, edge devices transmit high-resolution images over wireless links to a receiver, where an artificial intelligence…
↗Not briefed
hn
hn-ai
12dWhat does this company do?—
↗Not briefed
hn
hn-ai
12dShow HN: ApiClay – Turn a Website, Google Sheet or Database into a Paid API🔧 ChatGPTHi HN, I’m a reverse engineer, and I’ve seen the same problem repeatedly: people need data, but the path from “I need this” to “my app…
↗Not briefed
bluesky
bluesky-ai
12dWhile I agree with the overall findings, there are still so many questionable "good" cases presented as 'evidence-based. Eg: Waymo's AI is …While I agree with the overall findings, there are still so many questionable "good" cases presented as 'evidence-based. Eg: Waymo's A…
↗Not briefed
arxiv
arxiv-ai
12dNCP: Verifier-Guided Optimization of Weight-Space Control PoliciesWe present the Neural Control Plane (NCP), a closed-loop system that observes language-model execution, synthesizes candidate weight-s…
↗Not briefed
hn
hn-ai
12dShow HN: A Terminal UI for Hacker News, and CLI for Agentshn-cli is a Hacker News client for the terminal. You can use it to browse feeds, stories and comment threads in an elegant full-screen…
↗Not briefed
bluesky
bluesky-ai
12dWhy the fuck would you want an LLM as a proofreading tool. LLMs don't understand the data being fed to them or what they're putting out. An…Why the fuck would you want an LLM as a proofreading tool. LLMs don't understand the data being fed to them or what they're putting ou…
↗Not briefed
bluesky
bluesky-ai
12dI also think most Americans are locked into the duopoly of Anthropic and OpenAI when they build their mental model of what AI is; we gotta …I also think most Americans are locked into the duopoly of Anthropic and OpenAI when they build their mental model of what AI is; we g…
↗Not briefed
bluesky
bluesky-ai
12dI made a little #opensource task management app that interfaces with your AI agents via a tiny MCP. If Claude has a task for you, you'll se…I made a little #opensource task management app that interfaces with your AI agents via a tiny MCP. If Claude has a task for you, you'…
↗Not briefed
bluesky
bluesky-ai
12dgenuinely think this is where a lot of the AI criticism goes wrong: you can’t treat Anthropic’s Claude, which burns wild amounts of fossil …genuinely think this is where a lot of the AI criticism goes wrong: you can’t treat Anthropic’s Claude, which burns wild amounts of fo…
↗Not briefed
arxiv
arxiv-ai
12dDecisions by Discrimination: Differential Evidence for Medical ReasoningLarge language models (LLMs) are increasingly studied for clinical decision support, and medical question answering provides a testbed…
↗Not briefed
arxiv
arxiv-ai
12dW-ISA: Transactional Execution of Weight-Space Intervention ProgramsWe present W-ISA, a software instruction set and transactional runtime for executing weight-space intervention programs on pretrained…
↗Not briefed
bluesky
bluesky-ai
12dOpusに助けを求めることができるJevならOpusと同等の精度を4分の1のコストで実現可能 https://gigazine.net/news/20261110-llm-judges-identity-resolution/Opusに助けを求めることができるJevならOpusと同等の精度を4分の1のコストで実現可能 https://gigazine.net/news/20261110-llm-judges-identity-resolution/
↗Not briefed
rss
techmeme
12dSpaceX plans a gas-fired power plant and super-compute facility in Starbase, TX, with ~$24B of investment for initial phases, per a public hearing notice (Bloomberg)Bloomberg: SpaceX plans a gas-fired power plant and super-compute facility in Starbase, TX, with ~$24B of investment for initial phase…
↗Not briefed
arxiv
arxiv-ai
12dBlock-Sparse Gauss-Newton Optimization of Finite Basis Physics-Informed Neural Networks: Sparse Automatic Differentiation ApproachIn this work, we develop an efficient Gauss-Newton optimization approach for finite basis physics-informed neural networks (FBPINNs).…
↗Not briefed
arxiv
arxiv-ai
12dContinuous Horizon ForecastingMulti-horizon forecasting has been dominated by discrete forecast strategies that treat model outputs as fixed-vectors or recurrently…
↗Not briefed
bluesky
bluesky-ai
12dThis analysis is top-down, not bottom-up, so it's unique. But still: the author uses Anthropic, one of the most aggressively anti-climate a…This analysis is top-down, not bottom-up, so it's unique. But still: the author uses Anthropic, one of the most aggressively anti-clim…
↗Not briefed
arxiv
arxiv-ai
12dDeletion Sensitivity Is a Property of the Model: A Three-Condition Measurement across Seven Open Reasoning ModelsReasoning models write a reasoning trace at every turn of a multi-turn tool session, and later turns read this history. Across 94k Sha…
↗Not briefed
bluesky
bluesky-ai
12dPavement AI reaches near-perfect accuracy 🛣️ A new hybrid AI model developed by researchers detects pavement distress with 99.02% precisi…Pavement AI reaches near-perfect accuracy 🛣️ A new hybrid AI model developed by researchers detects pavement distress with 99.02% pr…
↗Not briefed