Track model releases, AI agents, open-source projects, infrastructure, policy, and product updates as they publish.
Rokid released AIOS, the first native operating system purpose-built for AI glasses, at Rokid Open Day 2026, along with the AIUI (AI Generated UI) interaction standard. The launch marks a shift from hardware-centric smart glasses to a system-level platform with its own intent-driven interaction model, app ecosystem, and developer community.
NVIDIA announced a new business model that partners with AI cloud providers to deploy large-scale, multi-tenant AI factories through a revenue-sharing and credit-support framework. Sharon AI and Firmus are among the first partners, with Sharon AI deploying up to 40,000 Grace Blackwell GB300 GPUs and Firmus building a 360-megawatt AI factory campus in Indonesia.
TechCrunch reports that SpaceX demonstrated a handset-like AI device prototype to investors before going public, with a design sleeker and slimmer than an iPhone. The device is expected to run a proprietary operating system and integrate technology from xAI, though Elon Musk denied the report.
Ashton Kutcher is departing Sound Ventures to co-found a new venture capital firm with Morgan Beller, pivoting from AI lab investments toward infrastructure and energy. The move signals a broader shift in AI venture focus from model-layer companies to the underlying compute and power systems.
Google published its June 2026 AI updates roundup, spanning consumer products and developer tools. Highlights include Android 17, Gemma 4 12B for local laptop inference, Nano Banana 2 Lite as the fastest Gemini image model, and Gemini Omni Flash entering public preview.
Cloudflare has set a September 15 deadline for AI companies to distinguish web crawlers used for search from those used for AI training and agents, or face default blocking across many publisher sites. This move pressures AI companies to negotiate content licensing deals with publishers to avoid being cut off from valuable data.
Anthropic has officially released Claude Sonnet 5, its latest mid-tier model, while simultaneously competing aggressively on AI agent pricing against rivals like OpenAI. The dual move signals a strategic push to capture more market share in the rapidly growing AI agent services space.
The US government has re-authorised Anthropic's AI models, according to a report by Digitec. This decision signals a significant regulatory move that may influence the broader AI industry's compliance landscape.
Google, the New York Jobs CEO Council, and Urban Assembly co-hosted an AI summit at Google's New York offices, bringing together 150 education and industry leaders. The event focused on shaping the future of AI in classrooms, with discussions on responsible integration and practical applications.
Google has launched its Gemini Spark AI agent on the Mac platform, extending the intelligent assistant from web and mobile into the desktop ecosystem. Mac users can now access Gemini Spark's agentic capabilities natively without relying on a browser.
Chinese life sciences instrument maker MGI (华大智造) partnered with Shanghai Artificial Intelligence Laboratory to develop Physical AI systems for life science labs. The project aims to bring AI into physical lab operations and data analysis, reportedly surpassing mainstream large models on certain tasks.
Anthropic is facing backlash after reports emerged that its Claude Code tool secretly tags users associated with 147 Chinese domain names. The controversy has sparked debate about privacy boundaries and potential geographic discrimination in AI developer tools.
MIT Technology Review reports that major LLMs produce strikingly similar responses to open-ended questions — for example, almost always saying '7' when asked for a random number between 1 and 10. An unnamed startup is now working on technical approaches to restore output diversity and break what the article calls AI's 'groupthink' problem.
Privacy-focused AI platform Venice AI raised $65 million in Series A funding, reaching unicorn status with a valuation over $1 billion. CEO Erik Voorhees told TechCrunch the company is already profitable, with annualized recurring revenue exceeding $70 million.
Google's 24/7 agentic assistant Gemini Spark has launched on Mac, bringing real-time tracking capabilities and support for more applications. The release extends Google's desktop AI agent capabilities into Apple's ecosystem.
The National Law Review published a test in which ChatGPT, Gemini, Perplexity, and another major AI model produced identical answers to the same legal analysis questions. The "zero disagreement" result has sparked fresh debate about AI consistency and reliability in legal practice.
Google has released the Nano Banana 2 Lite image model and expanded accessibility to Gemini Omni Flash for developers. These updates provide lighter image generation capabilities and broader access to multimodal AI tools.
Meta is developing plans to sell access to its AI compute power and models via a new cloud infrastructure business. The move would put Meta in direct competition with Amazon Web Services, Google Cloud, and Microsoft Azure.
NVIDIA and its partners are investing in American manufacturing, supply chains, and energy grids to produce AI infrastructure. This initiative aims to enhance US leadership in healthcare, scientific discovery, and industrial productivity.
Exabeam has expanded its behavior intelligence platform to secure the agentic enterprise. The move addresses the growing security challenges posed by AI agents in enterprise environments.
MGI (华大智造), a leading gene sequencing instrument maker, has partnered with the Shanghai Artificial Intelligence Laboratory to integrate AI into laboratory workflows. This cross-industry collaboration aims to bring intelligent analysis and automation to gene sequencing and life science research.
Sondera has open-sourced its AI Agent Policy Enforcement Harness, providing developers with policy management tools. This move is expected to enhance the safety and control of AI agents.
BNB Chain has officially launched Agent Studio, a platform designed to simplify the deployment of AI agents on blockchain. The tool provides developers and projects with one-stop capabilities for creating and managing on-chain agents.
NVIDIA has open-sourced a new robotics skill library, pushing embodied AI toward a continuous learning paradigm rather than static model training. Jim Fan stated that this open-source initiative marks a fundamental shift in how robot skills are acquired and shared.
OceanBase has formally introduced its 'Lake-House Integration' (Lakebase) strategy, proposing a three-tier AI database architecture spanning multi-model data engines, context layers, and AI application layers. The approach aims to unify structured and unstructured data management, enabling a single database to handle OLTP, OLAP, search, vector retrieval, and AI inference workloads.
The AFAC2026 Financial Intelligence Innovation Competition has officially launched, presenting four real-world financial business scenarios. The problem setters indicate there are no predetermined optimal solutions, challenging teams to explore innovative AI approaches.
OpenSquilla has launched version 0.4.0, introducing a self-verification feature that allows AI-generated code to automatically check for correctness. The tool has gained thousands of GitHub stars within weeks of its initial release.
Qunhe Technology has three papers accepted at ECCV 2026, covering spatial perception and reasoning, reinforcement learning data generation, and high-fidelity physics simulation. The company also partnered with NVIDIA, Adobe, and Apple to develop SPEAR, a next-generation physics AI simulation platform.
FaceMind Research Asia, a Chinese startup founded by two post-95 PhDs, has topped Hugging Face Papers with its Looped World Models (LoopWM) paper. The work proposes iteratively refining latent state representations through shared-parameter Transformer blocks, achieving up to 100x parameter efficiency gains.
OpenAI has unveiled the Codex Micro keyboard, marking the company's first foray into physical hardware. The compact keyboard is branded under the Codex name, suggesting deep integration with OpenAI's AI-powered coding assistant.
SiliconFlow, a Chinese AI infrastructure company known for its inference-as-a-service platform, has filed its IPO prospectus. The filing offers the first public glimpse into the financials of a business often called a Token factory in the AI reasoning era.
France has completed the first payment transaction initiated and executed autonomously by an AI agent, as reported by FinTech Futures. The event marks a critical step for AI agents moving from information processing to real-world financial operations.
MiniMax has published the M3 multimodal model series on Hugging Face, including a base model and the MXFP8 quantized version. The series supports image-text-to-text tasks, uses MoE architecture, and offers agent and coding capabilities.
MiniMax has officially released its M3 model on Hugging Face, a multimodal mixture-of-experts (MoE) model supporting an image-text-to-text pipeline. The release has already garnered over 192,000 downloads and 1,271 likes on the platform.
GSMA Intelligence has released the Agentic Core white paper, defining a new paradigm for the evolution of intelligent core networks. The document aims to guide the telecom industry towards AI-driven network architectures.
Om AI Lianhui has unveiled VLX, claiming it as the world's first edge streaming multimodal model designed for the physical world. The model enables real-time multimodal processing on edge devices, reducing latency and enhancing privacy.
Benchmark evaluations of Alibaba's Qwen 3.6 27B model show it achieving frontier-level results across multiple tests while running locally on consumer hardware. The results have sparked discussion about whether open-source models are closing in on proprietary frontier systems like GPT-5.
British autonomous driving AI company Wayve has launched an $85 million employee tender offer, valuing the company at $8.5 billion. The move reflects a growing trend among AI startups using secondary offerings to attract and retain talent.
Google has opened up Gemini Omni Flash to developers via the Gemini API and Google AI Studio, combining multimodal reasoning with video generation and editing. It also released Nano Banana 2 Lite, an image model that generates 1K resolution images in about 4 seconds at half the cost of its predecessor.
Hugging Face and Cerebras announced a collaboration to power real-time voice AI using Google's Gemma 4 model on Cerebras hardware. The partnership combines Cerebras' wafer-scale chips with Hugging Face's open-source ecosystem for low-latency speech inference.
Daily Archive