Track model releases, AI agents, open-source projects, infrastructure, policy, and product updates as they publish.
A photo circulating on Reddit and X appeared to show Senator Mitch McConnell covered in tubes in a hospital bed. Snopes debunked the image after detecting it contained Google's SynthID watermark, confirming it was AI-generated.
SpaceXAI released Grok 4.5 on Wednesday, the company’s first new model since going public several weeks ago. Elon Musk described it as an Opus-class model that is faster, more token-efficient, and cheaper than Anthropic’s flagship series.
General Intuition believes robotics is approaching a foundation model tipping point, having trained a physical AI model on millions of hours of video game data. The startup’s model powered a quadrupedal robot after just eight minutes of real-world fine-tuning, and the company raised $320 million at a $2.3 billion valuation last month.
OpenAI reversed course on Wednesday, announcing that GPT-5.6 Sol, Terra, and Luna will launch publicly on Thursday after previously delaying the release at the US government's request. The White House then issued a statement denying it had approved or cleared the models, creating confusion about the true nature of American AI oversight.
Growing regulatory pressure is building around AI's impact on children's safety, with policymakers and advocacy groups demanding dedicated AI safety frameworks for minors. Current AI regulations largely lack specific provisions for child protection.
Google Photos is adding a new Video Remix feature that can edit and transform videos in seconds, powered by the Gemini Omni model. Users can apply cinematic relighting, replace backgrounds, or add artistic styles like watercolor and oil painting effects, rolling out today to eligible AI subscribers in select countries.
NVIDIA published an open dataset project called 'Data for Agents' on the Hugging Face blog, aiming to provide high-quality training data for AI agent development. The project was contributed by a team of NVIDIA researchers.
Meta announced a new privacy safeguard for its AI glasses that disables the camera if the recording LED indicator is covered or tampered with. However, the update arrives as the company simultaneously expands AI features that collect more personal data from users.
OpenAI released new voice models on July 8 that enable ChatGPT to speak and listen simultaneously, a key milestone for natural real-time conversation. The GPT-Live-1 series replaces the existing Advanced Voice Mode by default, with a larger model available to paid subscribers.
Prime Intellect, a startup providing computing power and tools for enterprises to build AI agents, raised a $130 million Series A at a $1 billion valuation. Led by Radical Ventures with participation from Nvidia Ventures, Intel Capital, and Dell Technologies Capital, the company has already reached a $100 million annualized revenue run rate.
A TechCrunch analysis reveals that multiple AI startups are reaching revenue milestones in increasingly shorter timeframes. Anthropic crossed a $47 billion revenue run rate, while Mercor doubled its gross annualized revenue from $1 billion to $2 billion in just four months.
NVIDIA announced that its Nemotron 3 Ultra model achieved benchmark-leading accuracy on LangChain's Deep Agents harness, completing more tasks at higher throughput while running at 10x lower cost than top closed-source models.
OpenAI released a policy document on July 8 outlining its principles for partnering with government and national security agencies. The framework centers on responsible AI use, democratic accountability, and public safety.
OpenAI published a deep audit of SWE-Bench Pro, revealing that roughly 30-34% of tasks have fundamental flaws that distort model capability assessments. The company has retracted its earlier recommendation for the industry to adopt SWE-Bench Pro.
The press conference for the 2026 World Robot Congress was held in Beijing, unveiling the event's theme: 'Human-Machine Symbiosis, Industry-Demand Integration.' The conference revealed the direction and organizational progress for this year's event.
Chinese AI unicorn MiniMax saw its shares surge over 20% on the final trading day before its lockup expiry, defying the typical pre-dilution pressure. Three Wall Street investment banks issued bullish ratings simultaneously, while over 80% of shareholders signaled they would hold long-term rather than cash out.
OpenAI Academy partnered with the Walton Family Foundation to launch AI Skills Jams, a hands-on training program designed to help K-12 educators build practical AI skills for the classroom. The initiative targets teachers across the United States.
别让AI一上来就“进厂打螺丝”：智源悟界·Orca要先教模型理解世界如何变化. Hugging Face论文月榜第一
翁荔新博客提出「自进化先从Harness开始」，DeepSeek崔添翼转发附议. 崔添翼：这个方向很容易出成果
Alipay has completed an AI upgrade for its 'Tap' NFC feature, forming an AI-powered offline business network covering over 30 million touchpoints. Merchants can now access AI-driven business analytics, customer management, and marketing tools through simple tap interactions.
ZML, a high-profile French AI startup endorsed by Turing Award winner Yann LeCun, has released ZML/LLMD as a free open-source product designed to speed up AI inference across distributed chip clusters. The software aims to reduce the computational costs of running large language models at scale.
阿里斩获国际AI顶会最佳资源论文奖，提出Agent评测新范式.
Alibaba-owned Amap has released Phys AI Data, the industry's first one-stop spatial data foundation designed for physical AI training and real-world deployment. The system consists of Phys AI Foundry for simulation training and Phys AI Map for practical applications, tackling the twin challenges of robot training data scarcity and spatial awareness.
SambaNova has raised $1 billion at an $11 billion valuation in the first close of its Series F round, just months after Intel was reportedly exploring an acquisition at around $1.6 billion. The massive fundraise underscores continued investor appetite for AI chip startups competing in the inference market.
South Korean AI agent startup LinkAlpha has raised 34 billion won (approximately $24 million) for financial-sector AI agent solutions. The funding round highlights growing investor confidence in vertical AI agent applications within the finance industry.
Ant Group's robotics subsidiary Ant Lingbo has open-sourced LingBot-VLA 2.0, a vision-language-action foundation model that supports over 20 robot configurations across 17 manufacturers. The release marks a significant expansion in generalization capability and motion control precision for embodied AI development.
Nvidia is reportedly adopting a novel strategy to counter growing competition in the AI chip market: partnering with its rivals. The move signals a strategic shift for the dominant GPU maker as pressure mounts from AMD, Intel, and hyperscaler custom chips.
Three former Ideal Auto smart driving executives founded Zhijian Dynamic and delivered the first 100 units of their i7 Pro robot in under a year, setting the fastest delivery record in the embodied AI industry. The company also launched the world's first CNC intelligent embodied robot production line, where robots manufacture their own harmonic reducer components.
Chinese AI startup DeepSeek is pushing to design its own in-house AI inference chips, according to three sources familiar with the matter. The early-stage initiative aims to reduce dependence on Nvidia and Huawei for the silicon needed to run its popular AI models.
Hugging Face announced that its Transformers library vLLM modeling backend now matches or exceeds native vLLM implementations in inference speed. Model authors can automatically get fast vLLM inference from their Transformers code with zero porting work.
OpenAI released GPT-Live on July 8, a new generation of voice models designed for more natural human-AI interaction. The model is now live, powering the ChatGPT Voice experience.
MCloudBridge has launched an AI Agent powered by intent analysis, designed to enhance enterprise decision-making and automated service delivery. The agent understands users' underlying intent rather than surface-level commands, enabling more precise task execution in complex business scenarios.
Chinese authorities have initiated discussions with Alibaba and ByteDance regarding restrictions on the export of advanced AI models. The talks signal China's move to build its own framework for controlling the cross-border flow of AI technology.
Meta has released Muse, a new AI image generation model designed for advertising, home decoration, and creator applications. The launch strengthens Meta's generative AI product lineup and positions the company to compete in the rapidly evolving AI visual tools market.
Elon Musk's xAI has partnered with AI coding tool Cursor to jointly develop their first AI model, with a potential release as early as July 8 local time. The collaboration marks xAI's first foray into developer tooling with a custom-built model.
Hugging Face announced that users can now deploy AI models to Amazon SageMaker Studio with a single click, eliminating the need for manual AWS environment configuration. The integration streamlines the workflow from model discovery to training and inference on the cloud.
Microsoft has become the latest Silicon Valley giant to cut back on AI spending by shifting toward its own in-house models. The strategy reflects a broader industry move from external model procurement to self-developed alternatives as companies seek to control surging AI infrastructure costs.
Discord has confirmed that an AI-powered moderation system bug has been wrongfully banning users since May for posting harmless images. An additional 200 users were hit over the weekend before the company identified and fixed the issue.
Anthropic launched Claude Cowork on mobile and web, enabling cross-device task continuity where users start work at their desk and pick up results on their phone. Alongside the expansion, Anthropic released usage data from 1.2 million Cowork sessions showing that business process operations — not coding — is now the largest use category at 33.4%.
Time Magazine reports that China is considering restricting access to its most powerful AI models. The potential policy shift could affect the global availability of leading Chinese AI models including Qwen and DeepSeek series.
Daily Archive