Kimi逼近+股票大跌，美国AI各派两周内加速抱团
Related
Baidu's OCR model, Unlimited OCR, topped four charts: HuggingFace, GitHub, and more.
According to Mars Finance, Baidu recently officially released and open-sourced its end-to-end OCR model, Unlimited OCR. The model topped the GitHub Daily Trending and Python charts the day after its release, and also ranked first on HuggingFace's global overall model trending chart and multimodal model trending chart. Unlimited OCR is designed for long document parsing scenarios, with a total parameter size of 3B and only about 570M activation parameters during inference. Public evaluation results show that Unlimited OCR achieved a 93.92% overall score in the OmniDocBench v1.6 benchmark test, setting a new record for end-to-end OCR.
OpenAI's pre-IPO financial structure has drawn attention: its asset-light financial statements may face scrutiny from the capital market.
According to Odaily Odaily, OpenAI's unique financial structure, as it prepares for its IPO, may face further scrutiny from investors and the market. Publicly disclosed financial data reveals OpenAI to be a low-debt software company. As of March 31, 2026, the company's balance sheet shows zero debt and less than $750 million in lease liabilities. Meanwhile, although OpenAI is one of the most hardware-independent technology companies, its cash flow statement shows that its capital expenditures for the quarter were only about $46 million. This "asset-light" financial performance contrasts with OpenAI's actual business model. Analysts believe that if OpenAI is to proceed with its IPO, investors will likely focus on assessing the asset structure behind its financial statements and how the company can maintain capital efficiency while rapidly expanding its AI infrastructure. (The Information)
More than 300 data center bans have emerged across the United States, and the construction of AI computing power facilities such as OpenAI Stargate has faced strong local resistance.
According to Beating's monitoring, since 2023, state and local governments across the United States have passed over 300 restrictions or bans on data center construction, with over 275 of them concentrated in 2026. This wave of resistance is most intense in the Midwestern and Southern United States, where tech giants are building large-scale AI computing facilities. In Michigan, OpenAI and Oracle's project to build the Stargate supercomputer cluster has encountered obstacles. Although Michigan offers tax breaks until 2050 for data centers that invest over $250 million and create over 30 jobs, Saline Township voted to block changes to Stargate's plans. After the Stargate team won their lawsuit, construction began on June 1, 2026. As a ripple effect, over 20 towns within 50 miles of the Stargate project have passed data center bans, and over 50 restricted areas statewide have been established. Polls show that a staggering 70% of American adults oppose building AI data centers in their areas, citing water consumption, rising electricity costs, and landscape damage as primary concerns. The boycott has transcended the political spectrum. Left-wing independent Senator Bernie Sanders and Democratic Representative Alexandria Ocasio-Cortez jointly proposed a federal ban, while the right-wing conservative organization Humans First is organizing protests in multiple states. Most restrictions currently range from one month to one year, but more than a dozen states, including New Jersey, have already enacted permanent bans. At the state level, a one-year ban bill passed by the New York legislature is awaiting Governor Kathy Hochul's signature, while the governor of Maine vetoed a similar bill. Meanwhile, Ohio, Illinois, and Arizona have suspended tax breaks for data center developers until 2026.
OpenAI Open Source Plant Talk: Connect your potted plant to your microphone and sensors for real-time chat.
According to Beating's monitoring, the OpenAI open-source educational project Plant Talk combines Realtime API, multimodal vision models, and Arduino microcontrollers to enable indoor plants to have environmental perception and voice dialogue capabilities. The project supports low-latency, full-duplex voice dialogue with plants via a computer camera and microphone, and can connect to a capacitive soil moisture sensor and an LM393 photosensor via Arduino. The plant's AI character (such as the default character George) obtains hardware data through function calling. When a user asks if the plant needs water or light, the AI calls tools to read real-time physical indicators, combines this with images captured by the camera to estimate the plant's dryness, number of branches, and maximum span (supporting the use of a 25-cent coin as a size comparison), and responds in a human-like plant voice. In terms of security and interface design, the backend Express server securely delegates WebRTC connection security to the frontend by requesting a temporary client key (Ephemeral Token), preventing the leakage of the API master key on the client side. Visual recognition and timed observation use the Responses API combined with Zod mode to achieve structured output. The codebase also includes a built-in operation guide for AI agents AGENTS.md, which guides the AI assistant step by step to help users complete the software environment configuration and hardware assembly.
AI-powered agents suffer setbacks in their first foray into coffee shops: Gemini's excessive discounting leads to losses, and GPT's overly stingy practices cause raw material shortages.
According to Beating's monitoring, AI evaluation agency Andon Labs released test data on its AI agent Mona operating a physical coffee shop. In the first two months, Mona ran on a Gemini 3.1 Pro model. During this period, the model showed almost no concept of profit, not only excessively purchasing raw materials but also being easily swayed by customer claims, offering large discounts or even free items, and even admitting to a customer's claim of a 99% discount without verification. This resulted in the coffee shop spending approximately $15,000 on supplier and equipment purchases, while sales were only $9,000, leading to a net operating loss of nearly $6,000 (if fixed costs such as rent and salaries are included, total expenditures reach $38,000). Subsequently, the team switched the model to GPT-5.5. The new model showed significant anxiety in the face of losses and immediately stopped blindly ordering. However, this went to the other extreme: insufficient purchases led to a shortage of fresh raw materials. As of June 25, the availability of menu items had dropped to 77%, and 10 dishes had been forced to be removed from the menu. Meanwhile, GPT-5.5 demonstrated extremely strong anti-cheating and anti-jailbreak capabilities, rejecting all customers who requested special prices or offered free food in exchange for social media promotion.
Amazon AWS establishes a $1 billion AI FDE team to emulate the enterprise AI deployment models of OpenAI and Anthropic.
BlockBeats reported on June 30th that Amazon Web Services (AWS) announced the creation of a new Forward-Deployed Engineer (FDE) organization and will invest $1 billion in internal resources to help enterprise customers deploy customized AI agents and related systems. AWS stated that FDE engineers will be directly stationed at customer companies, responsible for building AI applications, optimizing workflows, and helping customers build the ability to independently develop and operate AI systems, rather than simply delivering and maintaining them. The FDE model was first promoted by Palantir and has rapidly gained popularity in recent years due to the growing demand for enterprise AI deployment. Previously, OpenAI and Anthropic also launched related FDE joint ventures, with sizes of approximately $4 billion and $1.5 billion respectively. Unlike the two companies, which formed joint ventures with private equity firms, AWS's $1 billion investment this time primarily comes from internal resources and is not an independent investment project. --------------------------------- Click the original link below to join the Beating · Lark AI news channel and monitor global AI hot topics and news 24/7.