The day's stories
OtherOpenAI claims it solved an 80-year-old math problem — for real this timeOpenAI has made a significant announcement regarding its advancements in mathematical reasoning, claiming to have solved a long-standing math problem that has pTechCrunch AIMay 20· 60%Model ReleaseStability AI releases a new audio model that can create 6-minute songsStability AI has unveiled its latest audio model, Stability Audio 3.0, which is designed to generate professional-grade music compositions that can last up to sTechCrunch AIMay 20· 59%InfrastructureNvidia posts another record quarter, reveals $43B of holdings in startupsNvidia has reported a record revenue of $81.6 billion for the quarter ending April 26, showcasing the company's robust performance primarily driven by its data TechCrunch AIMay 20· 58%OtherGoogle publishes exploit code threatening millions of Chromium usersGoogle has taken a controversial step by publishing exploit code that targets a long-standing vulnerability in the Chromium browser codebase. This vulnerabilityArs Technica AIMay 20· 57%Product LaunchFigma adds an AI assistant to its collaborative canvasFigma has launched an innovative AI assistant integrated into its collaborative design canvas, enabling users to leverage natural language prompts for generatinTechCrunch AIMay 20· 56%OtherAI search startups are blowing upThe landscape of search technology is undergoing a significant transformation as AI search startups gain momentum. Exa Labs has recently secured $250 million inTechCrunch AIMay 20· 56%FundingNanoClaw creator turns down $20M buyout offer, raises $12M seed insteadNanoCo, the innovative company behind NanoClaw, has made headlines by rejecting a lucrative $20 million buyout offer in favor of raising $12 million in a seed fTechCrunch AIMay 20· 56%FundingxAI burned $6.4B last year — SpaceX’s IPO filing shows why the spending is far from overElon Musk's xAI experienced significant financial challenges in 2025, reporting a staggering loss of $6.4 billion against revenues of only $3.2 billion. This inTechCrunch AIMay 20· 56%Product LaunchIrisGo, a startup backed by Andrew Ng, looks to become the AI desktop buddy you never knew you neededIrisGo, a new startup co-founded by Jeffrey Lai, a former Apple engineer, and supported by Andrew Ng's AI Fund, is set to revolutionize the way knowledge workerTechCrunch AIMay 20· 56%InfrastructureAnthropic will pay xAI $1.25B per month for computeAnthropic has secured a major agreement with xAI, committing to pay $1.25 billion per month for 300 megawatts of compute power until May 2029. This deal not onlTechCrunch AIMay 20· 55%InfrastructureMusk’s xAI is being sued over its data center generators — now it’s buying $2.8B moreElon Musk's xAI is currently embroiled in a lawsuit initiated by the NAACP, which alleges that the company's use of gas turbines at its Tennessee data center coTechCrunch AIMay 20· 54%FundingOpenAI barrels toward IPO that may happen in SeptemberOpenAI is making significant strides toward its initial public offering (IPO), which is anticipated to take place as early as September. This development comes TechCrunch AIMay 20· 53%ResearchPROWL: Prioritized Regret-Driven Optimization for World Model LearningThe recent introduction of PROWL (Prioritized Regret-Driven Optimization for World Model Learning) presents a significant advancement in the field of action-conarXiv cs.LGMay 20· 64%OtherTheory-optimal Quantization Based on FlatnessA recent paper introduces a novel post-training quantization framework called Bidirectional Diagonal Quantization (BDQ), which aims to enhance the performance oarXiv cs.LGMay 20· 64%ResearchLearn-by-Wire Training Control Governance: Bounded Autonomous Training Under Stress for Stability and EfficiencyRecent advancements in AI training methodologies have led to the introduction of the Learn-by-Wire Guard (LBW-Guard), a novel governance layer aimed at enhancinarXiv cs.AIMay 20· 64%OtherTrustworthy Agent Network: Trust in Agent Networks Must Be Baked In, Not Bolted OnRecent research has focused on the development of Agent-to-Agent (A2A) networks, which are comprised of autonomous agents utilizing large language models (LLMs)arXiv cs.AIMay 20· 64%BenchmarkDimensional Balance Improves Large Scale Spatiotemporal Prediction PerformanceRecent research has unveiled a novel framework aimed at enhancing spatiotemporal prediction accuracy by addressing the discrepancies between spatial and temporaarXiv cs.LGMay 20· 64%OtherInterference-Aware Multi-Task UnlearningRecent research has introduced an innovative interference-aware framework designed to enhance multi-task unlearning in machine learning. This new approach addrearXiv cs.AIMay 20· 64%OtherEmbedding by Elicitation: Dynamic Representations for Bayesian Optimization of System PromptsRecent research has unveiled a novel framework called ReElicit, which employs Bayesian optimization to enhance the tuning of AI system prompts. This innovative arXiv cs.AIMay 20· 63%ResearchDecisionBench: A Benchmark for Emergent Delegation in Long-Horizon Agentic WorkflowsResearchers have unveiled DecisionBench, a novel benchmark aimed at assessing emergent delegation within long-horizon agentic workflows. This benchmark consistsarXiv cs.AIMay 20· 63%OtherHELLoRA: Hot Experts Layer-Level Low-Rank Adaptation for Mixture-of-Experts ModelsResearchers have unveiled HELLoRA, a novel adaptation technique designed specifically for Mixture-of-Experts (MoE) models. This method enhances the efficiency oarXiv cs.LGMay 20· 63%Product LaunchOperationalizing Document AI: A Microservice Architecture for OCR and LLM Pipelines in ProductionRecent research has introduced a microservice architecture aimed at enhancing the operationalization of Document AI, focusing on the integration of Optical ChararXiv cs.AIMay 20· 63%FundingPosition: Let's Develop Data Probes to Fundamentally Understand How Data Affects LLM PerformanceA recent position paper emphasizes the need for 'data probes' to systematically investigate how different characteristics of data impact the performance of largarXiv cs.AIMay 20· 63%OtherSimply Stabilizing the Loop via Fully Looped TransformerRecent research has unveiled the Fully Looped Transformer, a novel architecture designed to improve the stability of training in transformer models. This innovaarXiv cs.LGMay 20· 62%OtherAgentNLQ: A General-Purpose Agent for Natural Language to SQLThe introduction of AgentNLQ marks a significant advancement in the field of natural language processing, specifically in converting natural language queries inarXiv cs.AIMay 20· 62%OtherUCCI: Calibrated Uncertainty for Cost-Optimal LLM Cascade RoutingThe introduction of UCCI, a calibrated router designed for large language model (LLM) cascades, marks a significant advancement in optimizing inference costs. BarXiv cs.LGMay 20· 62%OtherAdaptive Multi-Scale Goodness Aggregation for Forward-Forward LearningRecent research has unveiled a new method called Adaptive Multi-Scale Goodness Aggregation (AMSGA), which enhances the Forward-Forward learning algorithm. This arXiv cs.LGMay 20· 62%OtherReCrit: Transition-Aware Reinforcement Learning for Scientific Critic ReasoningThe recent introduction of ReCrit, a transition-aware reinforcement learning framework, marks a significant advancement in the field of scientific reasoning witarXiv cs.LGMay 20· 62%OtherRobust Basis Spline Decoupling for the Compression of Transformer ModelsRecent advancements in AI model compression have been highlighted by the introduction of the R-CMTF-BSD algorithm, which employs a B-spline-based decoupling fraarXiv cs.LGMay 20· 62%FundingEvaluating the Utility of Personal Health Records in Personalized Health AIA recent study has assessed the role of Personal Health Records (PHRs) in enhancing the capabilities of large language models (LLMs) for health-related inquiriearXiv cs.AIMay 20· 60%FundingAccurate Evaluation of Quickest Changepoint Detectors via Non-parametric Survival AnalysisRecent research has introduced innovative non-parametric estimators, KM-ARL and KM-ADD, aimed at enhancing the evaluation of quickest changepoint detectors. ThearXiv cs.LGMay 20· 60%