The day's stories
FundingGroq raises $350M to fuel its pivot from AI chips to neocloudGroq has successfully raised $350 million in funding to support its strategic shift from being an AI chip manufacturer to becoming a neocloud service provider. TechCrunch AIAug 17· 59%RegulationAmazon, which started off selling books, is destroying rare texts to train AIRecent reports indicate that Amazon has been acquiring rare books, cutting them into pieces, and scanning them to create datasets for training AI models. This pTechCrunch AIAug 17· 59%FundingNvidia investing $1.5B in SoftBank data center developer behind OpenAI projectNvidia has announced a significant investment of $1.5 billion in SB Energy, a data center developer linked to SoftBank and OpenAI. This strategic move positionsTechCrunch AIAug 17· 59%Product LaunchMicrosoft blames AI for delayed Exchange update, can’t say when it will arriveMicrosoft has announced a delay in the release of the Cumulative Update 1 (CU1) for Exchange Server Subscription Edition, attributing the postponement to additiThe Register AIAug 17· 58%Product LaunchExcel's Copilot function is headed for the Recycle BinMicrosoft has announced the retirement of its COPILOT() function in Excel, which will officially be removed on September 14, 2026. This decision comes less thanThe Register AIAug 17· 56%BenchmarkSame Cluster, 33 Points More Utilization: What Changed Was the OrderA recent development in GPU scheduling has led to a remarkable improvement in resource utilization and output. Researchers introduced a constraint-aware GPU allHugging Face BlogAug 17· 56%FundingAnthropic's annualized revenue surges to $65BAnthropic has reported a remarkable surge in its annualized revenue, reaching $65 billion as of July 2026. This figure marks a substantial increase from $47 bilTechCrunch AIAug 17· 56%Product LaunchMozilla adds ad blocking to Firefox for iOSMozilla has launched an optional ad blocking feature for its Firefox browser on iOS, enhancing user experience by reducing unwanted advertisements. This new funThe Register AIAug 17· 54%OtherAI automation startup Relay shuts down, staff joins Google's Chrome teamAI automation startup Relay has officially shut down, marking a significant shift in the landscape of AI-driven automation solutions. The closure has led to theTechCrunch AIAug 17· 54%FundingNvidia discloses $21B stake in SpaceXNvidia has disclosed a significant financial investment in SpaceX, owning nearly 123 million shares valued at approximately $21 billion as of June. However, theArs Technica AIAug 17· 54%BenchmarkImport AI 469: Science AI; RSI simulator; and Zuck's technological pessimismRecent advancements in AI research have led to the introduction of DiG-bench, a novel benchmark aimed at assessing AI's ability to discover rules within interacImport AIAug 17· 68%ResearchDepth-Aware Sensitivity Analysis of Mixture-of-Experts Models via Magnitude-Based Expert MaskingA recent paper has conducted a depth-aware sensitivity analysis on the Qwen3.6-35B-A3B Mixture-of-Experts model, focusing on how different layers respond to exparXiv cs.AIAug 17· 64%ResearchTraining-Free Knowledge Transfer Across Model Scales through Activation-Guided PruningRecent research has introduced a novel framework called Activation-Prune-Merge (APM), which facilitates the transfer of knowledge from larger language models toarXiv cs.LGAug 17· 64%ResearchA Year in LLM Serving: Workload Evolution, Caching and Load-BalancingA recent study published on arXiv has provided an in-depth analysis of the serving workloads associated with Large Language Models (LLMs) over the past year. TharXiv cs.AIAug 17· 63%ResearchFrom BERT to Frontier Agents: Eight Years of Language-Model Progress, the Collapse of the Capability-Cost Curve, and the Rise of Task-Targeted ModelsRecent research has documented the rapid evolution of language models over the past eight years, tracing their development from BERT to advanced agents capable arXiv cs.LGAug 17· 63%ResearchAgentao: A Governed Local-First Runtime for Tool-Using LLM AgentsThe recent paper on Agentao presents a novel governed local-first runtime specifically designed for tool-using large language model (LLM) agents. This frameworkarXiv cs.AIAug 17· 63%ResearchRobust Dual-Model Collaborative Random Vector Functional Link NetworkA new research paper presents the KRPRVFL model, which enhances the traditional Random Vector Functional Link (RVFL) networks. This innovative model addresses sarXiv cs.LGAug 17· 62%BenchmarkDon't Claim Benchmark-Oriented Optimization Improves General Coding Capability -- Diverse Evaluation Is RequiredA recent paper published on arXiv critiques the prevalent practice of using limited coding benchmarks to evaluate the general coding capabilities of AI models. arXiv cs.LGAug 17· 62%ResearchStable Miscalibration in Large Language Models: A Practical View of High-Confidence ErrorsA recent study published on arXiv delves into the phenomenon of stable miscalibration in large language models (LLMs), where these models exhibit high-confidencarXiv cs.AIAug 17· 62%ResearchEEG-PRISM: Physiologically-Grounded Interpretability of Predictions by EEG Foundation ModelsResearchers have introduced EEG-PRISM, a novel approach aimed at improving the interpretability of EEG foundation models. This method translates time-channel atarXiv cs.LGAug 17· 62%ResearchMeasuring Cross-Task Behavioral Consistency in Language Model AgentsRecent research has introduced the Behavioral Consistency Metric (BCM), which aims to evaluate the performance consistency of language model agents across multiarXiv cs.AIAug 17· 62%ResearchInducing Reward-Free Judging Rubrics that Reduce Over-Crediting in Agent EvaluationRecent research has introduced a novel method called RubricForge, designed to improve the evaluation of AI agents by creating reward-free judging rubrics. This arXiv cs.AIAug 17· 62%ResearchContrastive Learning for Interpretable Anomaly Detection at Collider ExperimentsResearchers have introduced a novel framework named ORCA, designed specifically for anomaly detection in collider physics. This two-stage framework leverages suarXiv cs.LGAug 17· 61%ResearchL-FNO: Lorentzian Fourier Neural Operator for Stochastic Event DynamicsThe introduction of the Lorentzian Fourier Neural Operator (L-FNO) marks a significant advancement in the field of stochastic modeling. This new neural operatorarXiv cs.LGAug 17· 61%ResearchActive Perception for Embodied DisambiguationRecent research has introduced a novel framework for active perception in robotics, focusing on embodied disambiguation. This framework allows robots to dynamicarXiv cs.AIAug 17· 60%ResearchCross-Disciplinary Taxonomy and Modeling of Misunderstanding Generation, Amplification, and Detection, from Pragmatics to AI AgentsA recent paper published on arXiv presents a comprehensive analysis of misunderstandings in communication, especially in the context of AI-mediated interactionsarXiv cs.AIAug 17· 60%ResearchRobust XGBoosting for RegressionA recent paper on arXiv presents a novel approach to enhancing the robustness of XGBoost, a widely used machine learning algorithm for regression tasks. The autarXiv cs.LGAug 17· 59%ResearchThe Query Knows What to Forget: A Second Erase Direction for Linear AttentionA new research paper presents an innovative method called Query-derived Erase Direction (QED) aimed at improving linear attention models. This method introducesarXiv cs.LGAug 17· 59%ResearchAI Evaluation Should Work With HumansA recent paper emphasizes the need to shift the focus of AI evaluation from merely assessing superhuman autonomous performance to evaluating the effectiveness oarXiv cs.AIAug 17· 58%