The day's stories
OtherThe agent evaluation gap: Enterprise AI organizations have a reality-alignment problem, not a coverage problem — and most are shipping to production anywayA recent survey conducted among 157 enterprises highlights a critical issue in the deployment of AI agents within organizations. The findings reveal a significaVentureBeat AIJul 16· 63%SafetyThe agent security gap: 54% of enterprises have already had an AI agent incident, and most still let agents share credentialsA recent survey has uncovered a troubling trend in enterprise security regarding AI agents, revealing that 54% of organizations have experienced at least one seVentureBeat AIJul 16· 63%ResearchThe AI context gap: Enterprise AI organizations have a trust problem, not a retrieval problem — and most are still building the fixA recent study involving 101 enterprises has highlighted a critical issue in the realm of enterprise AI: a trust problem stemming from inconsistent or missing bVentureBeat AIJul 16· 62%InfrastructureThe AI compute gap: Enterprises are buying infrastructure faster than they can measure what it costsA recent survey of 107 enterprises highlights a growing concern in the AI landscape: organizations are investing in AI infrastructure at a rapid pace, yet they VentureBeat AIJul 16· 61%FundingApplied Computing wants to give oil and gas operators an AI model for the entire plantApplied Computing, a startup based in London, has successfully raised $20 million to enhance its AI model named Orbital. This innovative model is designed speciTechCrunch AIJul 16· 60%Model ReleaseNVIDIA Nemotron 3 Embed Ranks #1 Overall on RTEB, Advancing Agentic RetrievalNVIDIA has introduced the Nemotron 3 Embed, a new suite of embedding models that have achieved the top ranking on the Retrieval Task Evaluation Benchmark (RTEB)Hugging Face BlogJul 16· 59%FundingHow a former DeepMind researcher raised at a $300M pre-seed valuation before launching a productAndrew Dai, a former researcher at Google DeepMind, has successfully raised a $55 million seed funding round for his startup, Elorian, which is valued at an impTechCrunch AIJul 16· 59%Product LaunchYes, you can now order DoorDash from the command lineDoorDash has introduced a new command-line interface (CLI) tool named 'dd-cli' that allows developers to place food orders directly from their AI agents. This lTechCrunch AIJul 16· 59%Model ReleaseMoonshot's upcoming Kimi 3 is expected to close the gap with Anthropic's Opus 4.8Moonshot AI is gearing up to release its latest model, Kimi 3, which is expected to rival Anthropic's Opus 4.8 in performance. This new model will be notable foTechCrunch AIJul 16· 59%ResearchWhy AMI Labs’ Alexandre LeBrun won't call his AI 'AGI' or 'superintelligence'Alexandre LeBrun, the CEO of AMI Labs, has chosen to avoid the terms 'AGI' (Artificial General Intelligence) and 'superintelligence' when discussing the capabilTechCrunch AIJul 16· 58%Product LaunchRoblox launches an AI-powered game-creation feature in its mobile appRoblox has introduced a new AI-powered feature named 'Build' within its mobile app, enabling users to create games through simple text prompts. This innovation TechCrunch AIJul 16· 58%Product LaunchGoogle Vids now lets you star in your own AI videosGoogle has launched an update to its Google Vids platform, introducing features that allow users to create custom digital avatars and generate AI-driven videos.TechCrunch AIJul 16· 57%RegulationEU forces Google to share its toys with the other AI and search kidsThe European Union has taken significant steps to curb Google's dominance in the search and AI sectors by mandating the company to share its search data with coThe Register AIJul 16· 57%ResearchNewer Models, Same AdvantageRecent developments in optical character recognition (OCR) have highlighted the advantages of specialized training approaches, particularly in the case of DharmHugging Face BlogJul 16· 57%Product LaunchGoogle's AI Mode now lets you link and interact with select appsGoogle has rolled out a significant update to its AI Mode, allowing users to link and interact with select applications such as Instacart, Canva, and YouTube diTechCrunch AIJul 16· 57%FundingEnergy IPOs surge as investors hunt for ways to play AI boomThe energy sector is witnessing a remarkable surge in initial public offerings (IPOs), with companies raising a staggering $12.6 billion in the first half of thArs Technica AIJul 16· 56%Product LaunchWhy is OpenAI selling a ChatGPT basketball?OpenAI has introduced a unique product: a ChatGPT-branded basketball, priced at $70. This launch is part of their 'Pause. Play. Prompt.' campaign, which aims toTechCrunch AIJul 16· 56%FundingAI vendors have found someone to pay their infrastructure bills: YouRecent insights from Forrester indicate a significant shift in the financial landscape for AI vendors and their customers. As the costs associated with AI infraThe Register AIJul 16· 55%RegulationHP fined 1.4 billion rupees for “cartelization” of ink cartridges, toner, PCsHP India has been penalized with a fine of 1.4 billion rupees by the Indian government due to its involvement in cartelization practices. The Competition CommisArs Technica AIJul 16· 52%ResearchImproving Molecular Property Prediction in Small Language Models Using Graph-based ToolsRecent research has focused on enhancing the capabilities of small language models (SLMs) for predicting molecular properties by integrating graph-based tools. arXiv cs.AIJul 16· 64%ResearchLearning Safe Agent Behaviour from Human Preferences and Justifications via World ModelsA recent study introduced DROPJ, an innovative method aimed at enhancing the safety of agent policies in unpredictable environments. By integrating human preferarXiv cs.AIJul 16· 64%ResearchUncertainty-Aware Sequential Decision Rules for Event-Triggered LLM Invocation in Streaming SystemsA recent study introduced a novel framework aimed at optimizing the invocation of Large Language Models (LLMs) within streaming inference systems. The researchearXiv cs.LGJul 16· 63%ResearchInterventional Grounding Audits: Black-Box Premise-Dependency Tests for LLM Chain-of-Thought via Predicate SubstitutionRecent research has introduced a new method called interventional grounding audits, aimed at evaluating the premise dependency of large language models (LLMs) darXiv cs.AIJul 16· 63%ResearchOracle Agent Memory as an Enterprise Memory Substrate for Long-Horizon AI AgentsOracle has unveiled a new solution called Oracle Agent Memory, aimed at improving memory management for long-horizon AI agents. This innovative system utilizes arXiv cs.AIJul 16· 63%ResearchSTKAN: Kolmogorov-Arnold Networks for Spatio-Temporal ForecastingA new architecture called STKAN has been introduced for spatio-temporal forecasting, leveraging Taylor-polynomial Kolmogorov-Arnold Network modules. This innovaarXiv cs.LGJul 16· 62%ResearchBeyond Backbone Backpropagation: A Decoupled Strategy for Efficient Transfer LearningRecent research has introduced a new training strategy for deep learning models that enhances the efficiency of transfer learning. This approach decouples featuarXiv cs.LGJul 16· 62%ResearchAutomatic Differentiation from Scratch: How PyTorch Computes Gradients in Physics-Informed Neural NetworksA recent paper published on arXiv delves into the intricacies of PyTorch's automatic differentiation engine, particularly in the context of Physics-Informed NeuarXiv cs.LGJul 16· 62%ResearchTSSM: Triaxial State Space Model for Global Station Weather Forecasting with Temporal-Variable-Historical ModelingA new model called the Triaxial State Space Model (TSSM) has been developed for global weather forecasting, particularly focusing on extreme weather events. ThiarXiv cs.LGJul 16· 62%ResearchSelf-Improvements in Modern Agentic Systems: A SurveyA recent survey published on arXiv outlines a comprehensive framework for self-improving autonomous agents. This research highlights the evolution of these systarXiv cs.AIJul 16· 62%ResearchTargeted Recovery of Weight-Space Mechanisms From Neural NetworksA recent study has introduced a novel approach called targeted parameter decomposition (tPD) that aims to enhance the efficiency of neural network analysis. ThiarXiv cs.LGJul 16· 62%ResearchSPINE: Bridging the Cyber-Physical Gap with Agentic AIResearchers have introduced a new framework called SPINE, which aims to bridge the gap between cyber and physical systems by utilizing agentic AI. This innovatiarXiv cs.AIJul 16· 61%ResearchOriginBlame: Record- and Token-Level Data Provenance for AI Training DatasetsThe recent introduction of OriginBlame marks a significant advancement in the field of AI training data management. This innovative system provides record- and arXiv cs.AIJul 16· 61%ResearchFederated Explainable Artificial Intelligence: Roles, Architectures, Evaluation, and Open ChallengesA recent systematic review has been published on Federated Explainable Artificial Intelligence (FedXAI), which merges the concepts of Federated Learning and ExparXiv cs.LGJul 16· 61%ResearchNetworked Intelligence: Active Shared Context Graphs for Human-AI Team ScienceA recent study has introduced Mycelium, a groundbreaking system designed to enhance collaboration between human researchers and AI agents in scientific researcharXiv cs.AIJul 16· 60%ResearchA Hybrid Mamba for Audio-Visual NavigationThe recent paper titled 'A Hybrid Mamba for Audio-Visual Navigation' presents a novel architecture called Samba, which innovatively integrates a Mamba State EncarXiv cs.LGJul 16· 60%ResearchProbabilistic Extension of Neuro-Symbolic AGI Robots based on Belnap's Typed Intensional FOLA recent paper has introduced a significant enhancement to neuro-symbolic AI by integrating probabilistic reasoning with Belnap's Typed Intensional First-Order arXiv cs.AIJul 16· 60%ResearchWhat Your Model Threw Away and Why You'll Want It Back: Masking, Fingerprinting, and Privacy from Discarded GeometryRecent research has unveiled a novel framework that delves into the information discarded by machine learning models, particularly those influenced by Lie grouparXiv cs.LGJul 16· 59%