The day's stories
Product LaunchAnthropic launches Claude Sonnet 5 as a cheaper way to run agentsAnthropic has unveiled Claude Sonnet 5, a new AI model aimed at enhancing the efficiency and cost-effectiveness of running agentic tasks. This model is designedTechCrunch AIJun 30· 61%FundingThe DeepMind trio who built a poker AI are now making money for quant hedge fundsThree former researchers from DeepMind have founded EquiLibre Technologies, an AI lab that leverages reinforcement learning techniques originally developed for TechCrunch AIJun 30· 61%Product LaunchAnthropic’s Claude Science bets on workflow, not a new model, to win over scientistsAnthropic has introduced Claude Science, a new AI workbench aimed at scientists to enhance their research workflows. This platform integrates various tools and TechCrunch AIJun 30· 60%Product LaunchAmazon launches new $1 billion FDE org, following OpenAI and AnthropicAmazon Web Services (AWS) has announced the launch of a new internal organization dedicated to forward-deployed engineers (FDE), backed by a substantial investmTechCrunch AIJun 30· 60%Product LaunchLumo, Proton's privacy-focused AI chatbot, gets an upgradeProton has announced an upgrade to its privacy-focused AI chatbot, Lumo, with the release of version 2.0. This new version introduces several advanced features,TechCrunch AIJun 30· 59%BenchmarkScarfBench: Benchmarking AI Agents for Enterprise Java Framework MigrationThe introduction of ScarfBench marks a significant step forward in the evaluation of AI agents specifically designed for enterprise Java framework migration. ThHugging Face BlogJun 30· 59%Product LaunchVibe coding platform Base44 launches own model as AI startups seek defensibilityBase44, a vibe coding platform recently acquired by Wix, has introduced its own AI model named Base1. This new model is designed to enhance the app creation proTechCrunch AIJun 30· 58%FundingNvidia competitor Etched hits $5B valuation, $1B in sales for AI chipEtched, a burgeoning competitor in the AI chip industry, has achieved significant milestones by reaching a $5 billion valuation and generating $1 billion in salTechCrunch AIJun 30· 58%Product LaunchGoogle introduces a faster, cheaper image generator with Nano Banana 2 LiteGoogle has unveiled Nano Banana 2 Lite, a new iteration of its AI image generator designed to be both faster and more cost-effective. This latest version can geTechCrunch AIJun 30· 58%Product LaunchActi puts AI agents directly into your smartphone keyboardActi, a startup based in Singapore, has unveiled an innovative AI-powered keyboard designed for both iOS and Android devices. This new keyboard integrates AI fuTechCrunch AIJun 30· 58%SafetyNew attack provides one more reason why AI browsers are a bad ideaRecent research has unveiled significant vulnerabilities in AI browsers, demonstrating how attackers can exploit these weaknesses to circumvent safety protocolsArs Technica AIJun 30· 57%Product LaunchCrypto exchange OKX wants AI agents to hire and pay each otherCrypto exchange OKX has introduced a groundbreaking marketplace named OKX AI, designed to facilitate autonomous transactions among AI agents. This innovative plTechCrunch AIJun 30· 57%BenchmarkFeaturing Every Eval Ever Results on Hugging Face Model PagesHugging Face has announced the integration of Every Eval Ever (EEE) with its Community Evals, marking a significant step towards standardizing the reporting of Hugging Face BlogJun 30· 57%OtherThe AI jobs debate just got messierThe ongoing debate surrounding the impact of AI on employment has intensified, particularly as companies announce significant layoffs. Projections indicate thatTechCrunch AIJun 30· 56%Product LaunchX now offers an MCP server to make its platform easier for AI tools to useX has introduced a hosted Model Context Protocol (MCP) server aimed at simplifying the integration of AI tools with its platform. This new service allows develoTechCrunch AIJun 30· 56%Product LaunchMicrosoft builds a bouncer to keep bots out of Teams meetingsMicrosoft has introduced a new feature for Teams designed to enhance security by preventing unauthorized bots from joining meetings. This 'bouncer' technology rThe Register AIJun 30· 56%ResearchWhy Specialization Is InevitableThe Hugging Face Blog discusses the inevitability of specialization in AI systems, arguing that focused systems tend to outperform their general counterparts. THugging Face BlogJun 30· 56%Product LaunchPodcasting platform Riverside enters the newsletter publishing gameRiverside, a prominent platform for video and podcast recording, has expanded its offerings by launching a new newsletter feature. This innovative tool enables TechCrunch AIJun 30· 55%ResearchATOD: Annealed Turn-aware On-policy Distillation for Multi-turn Autonomous AgentsThe recent paper on ATOD (Annealed Turn-aware On-policy Distillation) presents a novel approach to training language-model agents for multi-turn interactions. TarXiv cs.AIJun 30· 63%ResearchAgent-Native Immune System: Architecture, Taxonomy, and EngineeringResearchers have unveiled the Agent-Native Immune System (ANIS), a groundbreaking defense architecture designed specifically for autonomous AI agents. This innoarXiv cs.AIJun 30· 63%ResearchSingular Learning and Occam's Razor in Deep Monomial NetworksRecent research has delved into the dynamics of deep fully-connected networks that utilize monomial activations. The study highlights how critical points in thearXiv cs.LGJun 30· 63%ResearchLearning to Distributedly Estimate under Partially Known Dynamics: A Covariance-Agnostic Neural Kalman Consensus FilterA new research paper presents the Covariance-Agnostic Neural Kalman Consensus Filter (CA-NKCF), an innovative online distributed sensing framework designed for arXiv cs.LGJun 30· 63%ResearchAn Agentic AI Pipeline for Appliance-Level Energy Anomaly Detection and LLM-Driven RecommendationsA recent paper introduces an innovative agentic AI pipeline specifically designed for detecting energy anomalies at the appliance level within office buildings.arXiv cs.LGJun 30· 62%ResearchVerifiable Geometry Problem Solving: Solver-Driven Autoformalization and Theorem ProposingRecent advancements in geometry problem solving have been showcased through the introduction of a new framework called Solver-Driven Geometry Problem Solving (SarXiv cs.AIJun 30· 62%ResearchCounterfactual Residual Data Augmentation for RegressionA recent study has introduced a new technique called Counterfactual Residual Data Augmentation (CRDA) aimed at improving regression tasks in machine learning. TarXiv cs.LGJun 30· 62%ResearchOntology-Guided Evidence Path Inference for Multi-hop Knowledge Graph Question AnsweringThe recent paper on Ontology-Guided Evidence Path Inference (OPI) introduces a novel framework aimed at improving multi-hop knowledge graph question answering (arXiv cs.AIJun 30· 62%Product LaunchJD Oxygen AI Item Center (Oxygen AIIC) V1: An Industrial-Scale LLM/VLM-Centric Solution for Item Understanding, Management, and ApplicationsJD.com has officially launched the JD Oxygen AI Item Center (Oxygen AIIC), a cutting-edge platform aimed at revolutionizing item understanding and management tharXiv cs.AIJun 30· 61%BenchmarkCan AI Draw Science? A Benchmark for Evaluating Scientific Figure Generation by Text-to-Image and Multimodal ModelsA new benchmark called SciDraw-Bench has been introduced to evaluate the capabilities of text-to-image and multimodal models in generating scientific figures. TarXiv cs.LGJun 30· 61%ResearchLifted Causal InferenceRecent research has introduced Lifted Causal Inference (LCI), a novel algorithm designed to enhance the efficiency of calculating causal effects in relational darXiv cs.AIJun 30· 61%ResearchOn the Necessity of a Liquid Substrate for Mesh IntelligenceRecent research has highlighted the importance of a liquid substrate for enhancing mesh intelligence among autonomous agents. The study identifies two critical arXiv cs.LGJun 30· 61%BenchmarkNormAct: A Benchmark for Hidden Social Norm Compliance in Embodied PlanningResearchers have introduced NormAct, a new benchmark designed to evaluate how multimodal large language models (MLLMs) comply with hidden social norms during tharXiv cs.AIJun 30· 61%ResearchRelBall: Relation Ball with Quaternion Rotation for Knowledge Graph CompletionThe recent paper on RelBall presents a novel approach to Knowledge Graph Completion (KGC) by introducing a model that leverages quaternion rotation and a tail-carXiv cs.AIJun 30· 61%ResearchS-GAI: Spectral Geometry-Aware Initialization for Sigmoidal MLPs -- From Dataset Geometry to Network WeightsA new research framework called Spectral Geometry-Aware Initialization (S-GAI) has been introduced for sigmoidal multilayer perceptrons (MLPs). This innovative arXiv cs.LGJun 30· 60%ResearchModelling Emotional Memory in Children with Tensor NetworksA recent study published on arXiv investigates the impact of emotional valence on children's recognition memory using tensor networks. The research highlights harXiv cs.LGJun 30· 60%ResearchTandem Reinforcement Learning with Verifiable RewardsRecent research has introduced a new method called Tandem Reinforcement Learning (TRL), which integrates a more powerful senior model with a static junior modelarXiv cs.AIJun 30· 60%ResearchPosition: RL Researchers Need to Distinguish Between Solving Simulators and Using Simulators as a ProxyA recent paper highlights a critical distinction that reinforcement learning (RL) researchers must make between solving simulators and using them as proxies forarXiv cs.LGJun 30· 60%ResearchAI-Driven Synthesis for High-Tech System Design: Automating InnovationRecent research has introduced a novel approach to high-tech system design called automation-in-design (AiD). This paradigm integrates computational design syntarXiv cs.AIJun 30· 58%