The day's stories
Model ReleaseGoogle's Gemini Omni turns images, audio, and text into video — and that's just the startGoogle has unveiled Gemini Omni, a groundbreaking family of multimodal AI models designed to generate high-quality videos from a combination of images, audio, aTechCrunch AIMay 19· 61%Model ReleaseWith Gemini 3.5 Flash, Google bets its next AI wave on agents, not chatbotsGoogle has unveiled its latest AI model, Gemini 3.5 Flash, which signifies a pivotal shift in the company's approach to artificial intelligence. Unlike previousTechCrunch AIMay 19· 60%Model ReleaseAgentic app coding gets an upgrade with Google's release of Android CLIAt the recent Google I/O developer conference, Google unveiled the stable release of its Android Command Line Interface (CLI), marking a significant advancementTechCrunch AIMay 19· 60%OtherGoogle’s Genie world model can now simulate real streets with Street ViewGoogle DeepMind has made a significant advancement in its Project Genie by integrating Google Street View, allowing for the creation of immersive simulations ofTechCrunch AIMay 19· 60%Model ReleaseGoogle introduces Gemini Spark, a 24/7 agentic assistant with Gmail integration, at IO 2026At the Google I/O 2026 conference, Google unveiled Gemini Spark, a groundbreaking personal AI assistant designed to operate around the clock. This new assistantTechCrunch AIMay 19· 59%OtherGoogle Search as you know it is overGoogle has unveiled a major transformation of its Search functionality, marking a departure from the traditional search experience. The new AI-powered intelligeTechCrunch AIMay 19· 59%OtherGoogle just redesigned the search box for the first time in 25 years — here’s why it matters more than you think.Google has unveiled a significant redesign of its search box, marking the first major update in 25 years. This new interface is powered by AI and supports multiVentureBeat AIMay 19· 59%OtherOlmoEarth v1.1: A more efficient family of modelsThe recent release of OlmoEarth v1.1 marks a significant advancement in transformer-based models, particularly in the realm of remote sensing. This new family oHugging Face BlogMay 19· 59%FundingFrom teen hacker to Iron Dome researcher, this founder raised $28M to fight AI phishingShay Shwartz, a former hacker and cybersecurity expert, has launched a new startup named Ocean that aims to tackle the growing threat of AI-driven phishing attaTechCrunch AIMay 19· 59%Model ReleaseGoogle updates its Gemini app to take on ChatGPT and Claude at IO 2026At the Google I/O 2026 event, Google unveiled major updates to its Gemini app, aiming to enhance its competitiveness against leading AI platforms like ChatGPT aTechCrunch AIMay 19· 58%OtherHow to use Google’s new information agentsAt the 2026 Google I/O keynote, Google unveiled its latest innovation: AI information agents designed to enhance user engagement with personalized information. TechCrunch AIMay 19· 58%OtherGoogle's AI Studio now lets anyone build Android apps in minutesGoogle has introduced new features in its AI Studio that empower users to create Android applications in a matter of minutes. This advancement is designed to stTechCrunch AIMay 19· 58%OtherOpenAI is making it easier to check if an image was made by their modelsOpenAI has taken significant steps to improve the transparency of AI-generated images by adopting the C2PA standard for metadata signaling and collaborating witTechCrunch AIMay 19· 57%OtherGoogle's new Universal Cart wants to follow your entire shopping journey across the internetGoogle has unveiled its Universal Cart, a new feature designed to streamline the online shopping experience by allowing users to manage their purchases across dTechCrunch AIMay 19· 57%Model ReleaseGoogle launches Antigravity 2.0 with an updated desktop app and CLI toolGoogle has officially launched Antigravity 2.0, an upgraded version of its platform designed to enhance agentic coding capabilities. This new iteration includesTechCrunch AIMay 19· 57%OtherGoogle adds voice-based prompting to Docs and KeepAt the recent Google I/O developer conference, Google unveiled new voice-based prompting features for its Workspace applications, including Docs, Keep, and GmaiTechCrunch AIMay 19· 57%OtherGoogle takes a page out of Meta's book, announces new audio-powered smart glassesAt the recent Google I/O event, Google unveiled a new line of AI-powered audio glasses, developed in collaboration with Warby Parker and Gentle Monster. These iTechCrunch AIMay 19· 55%OtherGoogle just declared itself a contender in AI design at IO 2026At the recent Google I/O 2026 event, Google unveiled Pics, an innovative AI-powered design and image-generation application tailored for Google Workspace users.TechCrunch AIMay 19· 55%OtherGoogle's AI now lets you talk to your Gmail inboxGoogle has unveiled a significant enhancement to its Gmail service by introducing a feature called Gmail Live, which leverages conversational AI to allow users TechCrunch AIMay 19· 55%SafetyElon Musk said Sam Altman 'stole' a non-profit — but the trial showed he had similar aimsIn a recent trial, Elon Musk's lawsuit against OpenAI and Microsoft was dismissed by a jury, which found no merit in his accusations of 'breach of charitable trTechCrunch AIMay 19· 54%OtherSolvita: Enhancing Large Language Models for Competitive Programming via Agentic EvolutionRecent research has unveiled Solvita, an innovative framework designed to enhance large language models specifically for competitive programming. This frameworkarXiv cs.AIMay 19· 64%OtherZero-Shot Goal Recognition with Large Language ModelsRecent research has explored the capabilities of large language models (LLMs) in zero-shot goal recognition, specifically using classical Planning Domain DefiniarXiv cs.AIMay 19· 64%OtherCAPS: Cascaded Adaptive Pairwise Selection for Efficient Parallel ReasoningThe recent introduction of CAPS (Cascaded Adaptive Pairwise Selection) marks a significant advancement in the field of AI, particularly in enhancing the efficiearXiv cs.AIMay 19· 63%OtherContext Pruning for Coding Agents via Multi-Rubric Latent ReasoningRecent research has unveiled a new framework called LaMR (Latent Multi-Rubric) aimed at enhancing the performance of coding agents through a method known as conarXiv cs.AIMay 19· 63%OtherReducing Credit Assignment Variance via Counterfactual Reasoning PathsRecent advancements in reinforcement learning have led to the introduction of a novel framework known as Implicit Behavior Policy Optimization (IBPO). This apprarXiv cs.LGMay 19· 63%OtherAdaGraph: A Graph-Native Clustering Algorithm That Overcomes the Curse of Dimensionality and Enables Scientific DiscoveryAdaGraph is an innovative clustering algorithm designed to tackle the challenges posed by high-dimensional data in unsupervised learning. By employing a graph-narXiv cs.LGMay 19· 63%OtherBelief Engine: Configurable and Inspectable Stance Dynamics in Multi-Agent LLM DeliberationResearchers have unveiled the Belief Engine (BE), a groundbreaking framework aimed at improving the management and inspection of stance dynamics in multi-agent arXiv cs.AIMay 19· 63%OtherLanguage Game: Talking to Non-Human SystemsResearchers have introduced a novel framework designed to facilitate dialogue with non-human systems, including complex entities like gene regulatory networks aarXiv cs.LGMay 19· 63%ResearchSignMuon: Communication-Efficient Distributed Muon OptimizationResearchers have introduced SignMuon, a novel optimizer designed to improve the efficiency of distributed training for large neural networks. This approach levearXiv cs.LGMay 19· 62%OtherSystematic Optimization of Real-Time Diffusion Model Inference on Apple M3 UltraRecent research has focused on optimizing real-time image generation using diffusion models specifically on the Apple M3 Ultra chip. The study emphasizes the unarXiv cs.LGMay 19· 62%OtherFrom LLM-Generated Conjectures to Lean Formalizations: Automated Polynomial Inequality Proving via Sum-of-Squares CertificatesRecent advancements in the intersection of artificial intelligence and formal mathematics have been highlighted by the introduction of NSPI, a neuro-symbolic frarXiv cs.AIMay 19· 62%SafetyEnsemble Monitoring for AI Control: Diverse Signals Outweigh More ComputeRecent research has highlighted the effectiveness of ensemble monitoring in AI systems, demonstrating that incorporating diverse monitoring signals can significarXiv cs.AIMay 19· 62%OtherBeyond Partner Diversity: An Influence-Based Team Steering Framework for Zero-Shot Human-Machine TeamingRecent research has introduced a novel framework called Influence-Based Team Steering (IBTS), aimed at enhancing zero-shot human-machine teaming. This frameworkarXiv cs.AIMay 19· 62%BenchmarkX-SYNTH: Beyond Retrieval -- Enterprise Context Synthesis from Observed Human AttentionResearchers have unveiled X-SYNTH, a groundbreaking framework designed to enhance AI agents' performance by leveraging human attention data. This innovative apparXiv cs.AIMay 19· 61%OtherSMCEvolve: Principled Scientific Discovery via Sequential Monte Carlo EvolutionResearchers have unveiled SMCEvolve, an innovative framework designed to advance automated scientific discovery through the application of Sequential Monte CarlarXiv cs.AIMay 19· 61%OtherMirror Descent-Type Algorithms for the Variational Inequality Problem with Functional ConstraintsRecent advancements in optimization techniques have led to the development of mirror descent-type algorithms tailored for variational inequality problems that iarXiv cs.LGMay 19· 61%OtherWhen Actions Disappear: Adversarial Action Removal in Self-Play Reinforcement LearningRecent research has delved into the implications of adversarial action removal in self-play reinforcement learning, highlighting a critical vulnerability in howarXiv cs.LGMay 19· 60%OtherInvestigating Action Encodings in Recurrent Neural Networks in Reinforcement LearningRecent research has focused on enhancing the performance of recurrent neural networks (RNNs) in reinforcement learning (RL) by integrating action information inarXiv cs.LGMay 19· 60%RegulationA Structural Threshold in Decision Capacity Governs Collapse in Self-Play Reinforcement LearningRecent research has uncovered a critical structural threshold in decision capacity that plays a significant role in the stability of self-play reinforcement leaarXiv cs.LGMay 19· 60%